]>
Commit | Line | Data |
---|---|---|
29714574 TF |
1 | '\" te |
2 | .\" Copyright (c) 2013 by Turbo Fredriksson <turbo@bayour.com>. All rights reserved. | |
d4a72f23 | 3 | .\" Copyright (c) 2017 Datto Inc. |
917f475f | 4 | .\" Copyright (c) 2018 by Delphix. All rights reserved. |
29714574 TF |
5 | .\" The contents of this file are subject to the terms of the Common Development |
6 | .\" and Distribution License (the "License"). You may not use this file except | |
7 | .\" in compliance with the License. You can obtain a copy of the license at | |
8 | .\" usr/src/OPENSOLARIS.LICENSE or http://www.opensolaris.org/os/licensing. | |
9 | .\" | |
10 | .\" See the License for the specific language governing permissions and | |
11 | .\" limitations under the License. When distributing Covered Code, include this | |
12 | .\" CDDL HEADER in each file and include the License file at | |
13 | .\" usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this | |
14 | .\" CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your | |
15 | .\" own identifying information: | |
16 | .\" Portions Copyright [yyyy] [name of copyright owner] | |
d4a72f23 | 17 | .TH ZFS-MODULE-PARAMETERS 5 "Oct 28, 2017" |
29714574 TF |
18 | .SH NAME |
19 | zfs\-module\-parameters \- ZFS module parameters | |
20 | .SH DESCRIPTION | |
21 | .sp | |
22 | .LP | |
23 | Description of the different parameters to the ZFS module. | |
24 | ||
25 | .SS "Module parameters" | |
26 | .sp | |
27 | .LP | |
28 | ||
de4f8d5d BB |
29 | .sp |
30 | .ne 2 | |
31 | .na | |
32 | \fBdbuf_cache_max_bytes\fR (ulong) | |
33 | .ad | |
34 | .RS 12n | |
35 | Maximum size in bytes of the dbuf cache. When \fB0\fR this value will default | |
36 | to \fB1/2^dbuf_cache_shift\fR (1/32) of the target ARC size, otherwise the | |
37 | provided value in bytes will be used. The behavior of the dbuf cache and its | |
38 | associated settings can be observed via the \fB/proc/spl/kstat/zfs/dbufstats\fR | |
39 | kstat. | |
40 | .sp | |
41 | Default value: \fB0\fR. | |
42 | .RE | |
43 | ||
2e5dc449 MA |
44 | .sp |
45 | .ne 2 | |
46 | .na | |
47 | \fBdbuf_metadata_cache_max_bytes\fR (ulong) | |
48 | .ad | |
49 | .RS 12n | |
50 | Maximum size in bytes of the metadata dbuf cache. When \fB0\fR this value will | |
51 | default to \fB1/2^dbuf_cache_shift\fR (1/16) of the target ARC size, otherwise | |
52 | the provided value in bytes will be used. The behavior of the metadata dbuf | |
53 | cache and its associated settings can be observed via the | |
54 | \fB/proc/spl/kstat/zfs/dbufstats\fR kstat. | |
55 | .sp | |
56 | Default value: \fB0\fR. | |
57 | .RE | |
58 | ||
de4f8d5d BB |
59 | .sp |
60 | .ne 2 | |
61 | .na | |
62 | \fBdbuf_cache_hiwater_pct\fR (uint) | |
63 | .ad | |
64 | .RS 12n | |
65 | The percentage over \fBdbuf_cache_max_bytes\fR when dbufs must be evicted | |
66 | directly. | |
67 | .sp | |
68 | Default value: \fB10\fR%. | |
69 | .RE | |
70 | ||
71 | .sp | |
72 | .ne 2 | |
73 | .na | |
74 | \fBdbuf_cache_lowater_pct\fR (uint) | |
75 | .ad | |
76 | .RS 12n | |
77 | The percentage below \fBdbuf_cache_max_bytes\fR when the evict thread stops | |
78 | evicting dbufs. | |
79 | .sp | |
80 | Default value: \fB10\fR%. | |
81 | .RE | |
82 | ||
83 | .sp | |
84 | .ne 2 | |
85 | .na | |
86 | \fBdbuf_cache_shift\fR (int) | |
87 | .ad | |
88 | .RS 12n | |
89 | Set the size of the dbuf cache, \fBdbuf_cache_max_bytes\fR, to a log2 fraction | |
90 | of the target arc size. | |
91 | .sp | |
92 | Default value: \fB5\fR. | |
93 | .RE | |
94 | ||
2e5dc449 MA |
95 | .sp |
96 | .ne 2 | |
97 | .na | |
98 | \fBdbuf_metadata_cache_shift\fR (int) | |
99 | .ad | |
100 | .RS 12n | |
101 | Set the size of the dbuf metadata cache, \fBdbuf_metadata_cache_max_bytes\fR, | |
102 | to a log2 fraction of the target arc size. | |
103 | .sp | |
104 | Default value: \fB6\fR. | |
105 | .RE | |
106 | ||
6d836e6f RE |
107 | .sp |
108 | .ne 2 | |
109 | .na | |
110 | \fBignore_hole_birth\fR (int) | |
111 | .ad | |
112 | .RS 12n | |
113 | When set, the hole_birth optimization will not be used, and all holes will | |
114 | always be sent on zfs send. Useful if you suspect your datasets are affected | |
115 | by a bug in hole_birth. | |
116 | .sp | |
9ea9e0b9 | 117 | Use \fB1\fR for on (default) and \fB0\fR for off. |
6d836e6f RE |
118 | .RE |
119 | ||
29714574 TF |
120 | .sp |
121 | .ne 2 | |
122 | .na | |
123 | \fBl2arc_feed_again\fR (int) | |
124 | .ad | |
125 | .RS 12n | |
83426735 D |
126 | Turbo L2ARC warm-up. When the L2ARC is cold the fill interval will be set as |
127 | fast as possible. | |
29714574 TF |
128 | .sp |
129 | Use \fB1\fR for yes (default) and \fB0\fR to disable. | |
130 | .RE | |
131 | ||
132 | .sp | |
133 | .ne 2 | |
134 | .na | |
135 | \fBl2arc_feed_min_ms\fR (ulong) | |
136 | .ad | |
137 | .RS 12n | |
83426735 D |
138 | Min feed interval in milliseconds. Requires \fBl2arc_feed_again=1\fR and only |
139 | applicable in related situations. | |
29714574 TF |
140 | .sp |
141 | Default value: \fB200\fR. | |
142 | .RE | |
143 | ||
144 | .sp | |
145 | .ne 2 | |
146 | .na | |
147 | \fBl2arc_feed_secs\fR (ulong) | |
148 | .ad | |
149 | .RS 12n | |
150 | Seconds between L2ARC writing | |
151 | .sp | |
152 | Default value: \fB1\fR. | |
153 | .RE | |
154 | ||
155 | .sp | |
156 | .ne 2 | |
157 | .na | |
158 | \fBl2arc_headroom\fR (ulong) | |
159 | .ad | |
160 | .RS 12n | |
83426735 D |
161 | How far through the ARC lists to search for L2ARC cacheable content, expressed |
162 | as a multiplier of \fBl2arc_write_max\fR | |
29714574 TF |
163 | .sp |
164 | Default value: \fB2\fR. | |
165 | .RE | |
166 | ||
167 | .sp | |
168 | .ne 2 | |
169 | .na | |
170 | \fBl2arc_headroom_boost\fR (ulong) | |
171 | .ad | |
172 | .RS 12n | |
83426735 D |
173 | Scales \fBl2arc_headroom\fR by this percentage when L2ARC contents are being |
174 | successfully compressed before writing. A value of 100 disables this feature. | |
29714574 | 175 | .sp |
be54a13c | 176 | Default value: \fB200\fR%. |
29714574 TF |
177 | .RE |
178 | ||
29714574 TF |
179 | .sp |
180 | .ne 2 | |
181 | .na | |
182 | \fBl2arc_noprefetch\fR (int) | |
183 | .ad | |
184 | .RS 12n | |
83426735 D |
185 | Do not write buffers to L2ARC if they were prefetched but not used by |
186 | applications | |
29714574 TF |
187 | .sp |
188 | Use \fB1\fR for yes (default) and \fB0\fR to disable. | |
189 | .RE | |
190 | ||
191 | .sp | |
192 | .ne 2 | |
193 | .na | |
194 | \fBl2arc_norw\fR (int) | |
195 | .ad | |
196 | .RS 12n | |
197 | No reads during writes | |
198 | .sp | |
199 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
200 | .RE | |
201 | ||
202 | .sp | |
203 | .ne 2 | |
204 | .na | |
205 | \fBl2arc_write_boost\fR (ulong) | |
206 | .ad | |
207 | .RS 12n | |
603a1784 | 208 | Cold L2ARC devices will have \fBl2arc_write_max\fR increased by this amount |
83426735 | 209 | while they remain cold. |
29714574 TF |
210 | .sp |
211 | Default value: \fB8,388,608\fR. | |
212 | .RE | |
213 | ||
214 | .sp | |
215 | .ne 2 | |
216 | .na | |
217 | \fBl2arc_write_max\fR (ulong) | |
218 | .ad | |
219 | .RS 12n | |
220 | Max write bytes per interval | |
221 | .sp | |
222 | Default value: \fB8,388,608\fR. | |
223 | .RE | |
224 | ||
99b14de4 ED |
225 | .sp |
226 | .ne 2 | |
227 | .na | |
228 | \fBmetaslab_aliquot\fR (ulong) | |
229 | .ad | |
230 | .RS 12n | |
231 | Metaslab granularity, in bytes. This is roughly similar to what would be | |
232 | referred to as the "stripe size" in traditional RAID arrays. In normal | |
233 | operation, ZFS will try to write this amount of data to a top-level vdev | |
234 | before moving on to the next one. | |
235 | .sp | |
236 | Default value: \fB524,288\fR. | |
237 | .RE | |
238 | ||
f3a7f661 GW |
239 | .sp |
240 | .ne 2 | |
241 | .na | |
242 | \fBmetaslab_bias_enabled\fR (int) | |
243 | .ad | |
244 | .RS 12n | |
245 | Enable metaslab group biasing based on its vdev's over- or under-utilization | |
246 | relative to the pool. | |
247 | .sp | |
248 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
249 | .RE | |
250 | ||
d830d479 MA |
251 | .sp |
252 | .ne 2 | |
253 | .na | |
254 | \fBmetaslab_force_ganging\fR (ulong) | |
255 | .ad | |
256 | .RS 12n | |
257 | Make some blocks above a certain size be gang blocks. This option is used | |
258 | by the test suite to facilitate testing. | |
259 | .sp | |
260 | Default value: \fB16,777,217\fR. | |
261 | .RE | |
262 | ||
4e21fd06 DB |
263 | .sp |
264 | .ne 2 | |
265 | .na | |
266 | \fBzfs_metaslab_segment_weight_enabled\fR (int) | |
267 | .ad | |
268 | .RS 12n | |
269 | Enable/disable segment-based metaslab selection. | |
270 | .sp | |
271 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
272 | .RE | |
273 | ||
274 | .sp | |
275 | .ne 2 | |
276 | .na | |
277 | \fBzfs_metaslab_switch_threshold\fR (int) | |
278 | .ad | |
279 | .RS 12n | |
280 | When using segment-based metaslab selection, continue allocating | |
321204be | 281 | from the active metaslab until \fBzfs_metaslab_switch_threshold\fR |
4e21fd06 DB |
282 | worth of buckets have been exhausted. |
283 | .sp | |
284 | Default value: \fB2\fR. | |
285 | .RE | |
286 | ||
29714574 TF |
287 | .sp |
288 | .ne 2 | |
289 | .na | |
aa7d06a9 | 290 | \fBmetaslab_debug_load\fR (int) |
29714574 TF |
291 | .ad |
292 | .RS 12n | |
aa7d06a9 GW |
293 | Load all metaslabs during pool import. |
294 | .sp | |
295 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
296 | .RE | |
297 | ||
298 | .sp | |
299 | .ne 2 | |
300 | .na | |
301 | \fBmetaslab_debug_unload\fR (int) | |
302 | .ad | |
303 | .RS 12n | |
304 | Prevent metaslabs from being unloaded. | |
29714574 TF |
305 | .sp |
306 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
307 | .RE | |
308 | ||
f3a7f661 GW |
309 | .sp |
310 | .ne 2 | |
311 | .na | |
312 | \fBmetaslab_fragmentation_factor_enabled\fR (int) | |
313 | .ad | |
314 | .RS 12n | |
315 | Enable use of the fragmentation metric in computing metaslab weights. | |
316 | .sp | |
317 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
318 | .RE | |
319 | ||
b8bcca18 MA |
320 | .sp |
321 | .ne 2 | |
322 | .na | |
d2734cce | 323 | \fBvdev_max_ms_count\fR (int) |
b8bcca18 MA |
324 | .ad |
325 | .RS 12n | |
e4e94ca3 | 326 | When a vdev is added target this number of metaslabs per top-level vdev. |
b8bcca18 MA |
327 | .sp |
328 | Default value: \fB200\fR. | |
329 | .RE | |
330 | ||
d2734cce SD |
331 | .sp |
332 | .ne 2 | |
333 | .na | |
334 | \fBvdev_min_ms_count\fR (int) | |
335 | .ad | |
336 | .RS 12n | |
337 | Minimum number of metaslabs to create in a top-level vdev. | |
338 | .sp | |
339 | Default value: \fB16\fR. | |
340 | .RE | |
341 | ||
e4e94ca3 DB |
342 | .sp |
343 | .ne 2 | |
344 | .na | |
345 | \fBvdev_ms_count_limit\fR (int) | |
346 | .ad | |
347 | .RS 12n | |
348 | Practical upper limit of total metaslabs per top-level vdev. | |
349 | .sp | |
350 | Default value: \fB131,072\fR. | |
351 | .RE | |
352 | ||
f3a7f661 GW |
353 | .sp |
354 | .ne 2 | |
355 | .na | |
356 | \fBmetaslab_preload_enabled\fR (int) | |
357 | .ad | |
358 | .RS 12n | |
359 | Enable metaslab group preloading. | |
360 | .sp | |
361 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
362 | .RE | |
363 | ||
364 | .sp | |
365 | .ne 2 | |
366 | .na | |
367 | \fBmetaslab_lba_weighting_enabled\fR (int) | |
368 | .ad | |
369 | .RS 12n | |
370 | Give more weight to metaslabs with lower LBAs, assuming they have | |
371 | greater bandwidth as is typically the case on a modern constant | |
372 | angular velocity disk drive. | |
373 | .sp | |
374 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
375 | .RE | |
376 | ||
29714574 TF |
377 | .sp |
378 | .ne 2 | |
379 | .na | |
380 | \fBspa_config_path\fR (charp) | |
381 | .ad | |
382 | .RS 12n | |
383 | SPA config file | |
384 | .sp | |
385 | Default value: \fB/etc/zfs/zpool.cache\fR. | |
386 | .RE | |
387 | ||
e8b96c60 MA |
388 | .sp |
389 | .ne 2 | |
390 | .na | |
391 | \fBspa_asize_inflation\fR (int) | |
392 | .ad | |
393 | .RS 12n | |
394 | Multiplication factor used to estimate actual disk consumption from the | |
395 | size of data being written. The default value is a worst case estimate, | |
396 | but lower values may be valid for a given pool depending on its | |
397 | configuration. Pool administrators who understand the factors involved | |
398 | may wish to specify a more realistic inflation factor, particularly if | |
399 | they operate close to quota or capacity limits. | |
400 | .sp | |
83426735 | 401 | Default value: \fB24\fR. |
e8b96c60 MA |
402 | .RE |
403 | ||
6cb8e530 PZ |
404 | .sp |
405 | .ne 2 | |
406 | .na | |
407 | \fBspa_load_print_vdev_tree\fR (int) | |
408 | .ad | |
409 | .RS 12n | |
410 | Whether to print the vdev tree in the debugging message buffer during pool import. | |
411 | Use 0 to disable and 1 to enable. | |
412 | .sp | |
413 | Default value: \fB0\fR. | |
414 | .RE | |
415 | ||
dea377c0 MA |
416 | .sp |
417 | .ne 2 | |
418 | .na | |
419 | \fBspa_load_verify_data\fR (int) | |
420 | .ad | |
421 | .RS 12n | |
422 | Whether to traverse data blocks during an "extreme rewind" (\fB-X\fR) | |
423 | import. Use 0 to disable and 1 to enable. | |
424 | ||
425 | An extreme rewind import normally performs a full traversal of all | |
426 | blocks in the pool for verification. If this parameter is set to 0, | |
427 | the traversal skips non-metadata blocks. It can be toggled once the | |
428 | import has started to stop or start the traversal of non-metadata blocks. | |
429 | .sp | |
83426735 | 430 | Default value: \fB1\fR. |
dea377c0 MA |
431 | .RE |
432 | ||
433 | .sp | |
434 | .ne 2 | |
435 | .na | |
436 | \fBspa_load_verify_metadata\fR (int) | |
437 | .ad | |
438 | .RS 12n | |
439 | Whether to traverse blocks during an "extreme rewind" (\fB-X\fR) | |
440 | pool import. Use 0 to disable and 1 to enable. | |
441 | ||
442 | An extreme rewind import normally performs a full traversal of all | |
1c012083 | 443 | blocks in the pool for verification. If this parameter is set to 0, |
dea377c0 MA |
444 | the traversal is not performed. It can be toggled once the import has |
445 | started to stop or start the traversal. | |
446 | .sp | |
83426735 | 447 | Default value: \fB1\fR. |
dea377c0 MA |
448 | .RE |
449 | ||
450 | .sp | |
451 | .ne 2 | |
452 | .na | |
453 | \fBspa_load_verify_maxinflight\fR (int) | |
454 | .ad | |
455 | .RS 12n | |
456 | Maximum concurrent I/Os during the traversal performed during an "extreme | |
457 | rewind" (\fB-X\fR) pool import. | |
458 | .sp | |
83426735 | 459 | Default value: \fB10000\fR. |
dea377c0 MA |
460 | .RE |
461 | ||
6cde6435 BB |
462 | .sp |
463 | .ne 2 | |
464 | .na | |
465 | \fBspa_slop_shift\fR (int) | |
466 | .ad | |
467 | .RS 12n | |
468 | Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space | |
469 | in the pool to be consumed. This ensures that we don't run the pool | |
470 | completely out of space, due to unaccounted changes (e.g. to the MOS). | |
471 | It also limits the worst-case time to allocate space. If we have | |
472 | less than this amount of free space, most ZPL operations (e.g. write, | |
473 | create) will return ENOSPC. | |
474 | .sp | |
83426735 | 475 | Default value: \fB5\fR. |
6cde6435 BB |
476 | .RE |
477 | ||
0dc2f70c MA |
478 | .sp |
479 | .ne 2 | |
480 | .na | |
481 | \fBvdev_removal_max_span\fR (int) | |
482 | .ad | |
483 | .RS 12n | |
484 | During top-level vdev removal, chunks of data are copied from the vdev | |
485 | which may include free space in order to trade bandwidth for IOPS. | |
486 | This parameter determines the maximum span of free space (in bytes) | |
487 | which will be included as "unnecessary" data in a chunk of copied data. | |
488 | ||
489 | The default value here was chosen to align with | |
490 | \fBzfs_vdev_read_gap_limit\fR, which is a similar concept when doing | |
491 | regular reads (but there's no reason it has to be the same). | |
492 | .sp | |
493 | Default value: \fB32,768\fR. | |
494 | .RE | |
495 | ||
29714574 TF |
496 | .sp |
497 | .ne 2 | |
498 | .na | |
499 | \fBzfetch_array_rd_sz\fR (ulong) | |
500 | .ad | |
501 | .RS 12n | |
27b293be | 502 | If prefetching is enabled, disable prefetching for reads larger than this size. |
29714574 TF |
503 | .sp |
504 | Default value: \fB1,048,576\fR. | |
505 | .RE | |
506 | ||
507 | .sp | |
508 | .ne 2 | |
509 | .na | |
7f60329a | 510 | \fBzfetch_max_distance\fR (uint) |
29714574 TF |
511 | .ad |
512 | .RS 12n | |
7f60329a | 513 | Max bytes to prefetch per stream (default 8MB). |
29714574 | 514 | .sp |
7f60329a | 515 | Default value: \fB8,388,608\fR. |
29714574 TF |
516 | .RE |
517 | ||
518 | .sp | |
519 | .ne 2 | |
520 | .na | |
521 | \fBzfetch_max_streams\fR (uint) | |
522 | .ad | |
523 | .RS 12n | |
27b293be | 524 | Max number of streams per zfetch (prefetch streams per file). |
29714574 TF |
525 | .sp |
526 | Default value: \fB8\fR. | |
527 | .RE | |
528 | ||
529 | .sp | |
530 | .ne 2 | |
531 | .na | |
532 | \fBzfetch_min_sec_reap\fR (uint) | |
533 | .ad | |
534 | .RS 12n | |
27b293be | 535 | Min time before an active prefetch stream can be reclaimed |
29714574 TF |
536 | .sp |
537 | Default value: \fB2\fR. | |
538 | .RE | |
539 | ||
25458cbe TC |
540 | .sp |
541 | .ne 2 | |
542 | .na | |
543 | \fBzfs_arc_dnode_limit\fR (ulong) | |
544 | .ad | |
545 | .RS 12n | |
546 | When the number of bytes consumed by dnodes in the ARC exceeds this number of | |
9907cc1c | 547 | bytes, try to unpin some of it in response to demand for non-metadata. This |
627791f3 | 548 | value acts as a ceiling to the amount of dnode metadata, and defaults to 0 which |
9907cc1c G |
549 | indicates that a percent which is based on \fBzfs_arc_dnode_limit_percent\fR of |
550 | the ARC meta buffers that may be used for dnodes. | |
25458cbe TC |
551 | |
552 | See also \fBzfs_arc_meta_prune\fR which serves a similar purpose but is used | |
553 | when the amount of metadata in the ARC exceeds \fBzfs_arc_meta_limit\fR rather | |
554 | than in response to overall demand for non-metadata. | |
555 | ||
556 | .sp | |
9907cc1c G |
557 | Default value: \fB0\fR. |
558 | .RE | |
559 | ||
560 | .sp | |
561 | .ne 2 | |
562 | .na | |
563 | \fBzfs_arc_dnode_limit_percent\fR (ulong) | |
564 | .ad | |
565 | .RS 12n | |
566 | Percentage that can be consumed by dnodes of ARC meta buffers. | |
567 | .sp | |
568 | See also \fBzfs_arc_dnode_limit\fR which serves a similar purpose but has a | |
569 | higher priority if set to nonzero value. | |
570 | .sp | |
be54a13c | 571 | Default value: \fB10\fR%. |
25458cbe TC |
572 | .RE |
573 | ||
574 | .sp | |
575 | .ne 2 | |
576 | .na | |
577 | \fBzfs_arc_dnode_reduce_percent\fR (ulong) | |
578 | .ad | |
579 | .RS 12n | |
580 | Percentage of ARC dnodes to try to scan in response to demand for non-metadata | |
6146e17e | 581 | when the number of bytes consumed by dnodes exceeds \fBzfs_arc_dnode_limit\fR. |
25458cbe TC |
582 | |
583 | .sp | |
be54a13c | 584 | Default value: \fB10\fR% of the number of dnodes in the ARC. |
25458cbe TC |
585 | .RE |
586 | ||
49ddb315 MA |
587 | .sp |
588 | .ne 2 | |
589 | .na | |
590 | \fBzfs_arc_average_blocksize\fR (int) | |
591 | .ad | |
592 | .RS 12n | |
593 | The ARC's buffer hash table is sized based on the assumption of an average | |
594 | block size of \fBzfs_arc_average_blocksize\fR (default 8K). This works out | |
595 | to roughly 1MB of hash table per 1GB of physical memory with 8-byte pointers. | |
596 | For configurations with a known larger average block size this value can be | |
597 | increased to reduce the memory footprint. | |
598 | ||
599 | .sp | |
600 | Default value: \fB8192\fR. | |
601 | .RE | |
602 | ||
ca0bf58d PS |
603 | .sp |
604 | .ne 2 | |
605 | .na | |
606 | \fBzfs_arc_evict_batch_limit\fR (int) | |
607 | .ad | |
608 | .RS 12n | |
8f343973 | 609 | Number ARC headers to evict per sub-list before proceeding to another sub-list. |
ca0bf58d PS |
610 | This batch-style operation prevents entire sub-lists from being evicted at once |
611 | but comes at a cost of additional unlocking and locking. | |
612 | .sp | |
613 | Default value: \fB10\fR. | |
614 | .RE | |
615 | ||
29714574 TF |
616 | .sp |
617 | .ne 2 | |
618 | .na | |
619 | \fBzfs_arc_grow_retry\fR (int) | |
620 | .ad | |
621 | .RS 12n | |
ca85d690 | 622 | If set to a non zero value, it will replace the arc_grow_retry value with this value. |
d4a72f23 | 623 | The arc_grow_retry value (default 5) is the number of seconds the ARC will wait before |
ca85d690 | 624 | trying to resume growth after a memory pressure event. |
29714574 | 625 | .sp |
ca85d690 | 626 | Default value: \fB0\fR. |
29714574 TF |
627 | .RE |
628 | ||
629 | .sp | |
630 | .ne 2 | |
631 | .na | |
7e8bddd0 | 632 | \fBzfs_arc_lotsfree_percent\fR (int) |
29714574 TF |
633 | .ad |
634 | .RS 12n | |
7e8bddd0 BB |
635 | Throttle I/O when free system memory drops below this percentage of total |
636 | system memory. Setting this value to 0 will disable the throttle. | |
29714574 | 637 | .sp |
be54a13c | 638 | Default value: \fB10\fR%. |
29714574 TF |
639 | .RE |
640 | ||
641 | .sp | |
642 | .ne 2 | |
643 | .na | |
7e8bddd0 | 644 | \fBzfs_arc_max\fR (ulong) |
29714574 TF |
645 | .ad |
646 | .RS 12n | |
83426735 D |
647 | Max arc size of ARC in bytes. If set to 0 then it will consume 1/2 of system |
648 | RAM. This value must be at least 67108864 (64 megabytes). | |
649 | .sp | |
650 | This value can be changed dynamically with some caveats. It cannot be set back | |
651 | to 0 while running and reducing it below the current ARC size will not cause | |
652 | the ARC to shrink without memory pressure to induce shrinking. | |
29714574 | 653 | .sp |
7e8bddd0 | 654 | Default value: \fB0\fR. |
29714574 TF |
655 | .RE |
656 | ||
ca85d690 | 657 | .sp |
658 | .ne 2 | |
659 | .na | |
660 | \fBzfs_arc_meta_adjust_restarts\fR (ulong) | |
661 | .ad | |
662 | .RS 12n | |
663 | The number of restart passes to make while scanning the ARC attempting | |
664 | the free buffers in order to stay below the \fBzfs_arc_meta_limit\fR. | |
665 | This value should not need to be tuned but is available to facilitate | |
666 | performance analysis. | |
667 | .sp | |
668 | Default value: \fB4096\fR. | |
669 | .RE | |
670 | ||
29714574 TF |
671 | .sp |
672 | .ne 2 | |
673 | .na | |
674 | \fBzfs_arc_meta_limit\fR (ulong) | |
675 | .ad | |
676 | .RS 12n | |
2cbb06b5 BB |
677 | The maximum allowed size in bytes that meta data buffers are allowed to |
678 | consume in the ARC. When this limit is reached meta data buffers will | |
679 | be reclaimed even if the overall arc_c_max has not been reached. This | |
9907cc1c G |
680 | value defaults to 0 which indicates that a percent which is based on |
681 | \fBzfs_arc_meta_limit_percent\fR of the ARC may be used for meta data. | |
29714574 | 682 | .sp |
83426735 | 683 | This value my be changed dynamically except that it cannot be set back to 0 |
9907cc1c | 684 | for a specific percent of the ARC; it must be set to an explicit value. |
83426735 | 685 | .sp |
29714574 TF |
686 | Default value: \fB0\fR. |
687 | .RE | |
688 | ||
9907cc1c G |
689 | .sp |
690 | .ne 2 | |
691 | .na | |
692 | \fBzfs_arc_meta_limit_percent\fR (ulong) | |
693 | .ad | |
694 | .RS 12n | |
695 | Percentage of ARC buffers that can be used for meta data. | |
696 | ||
697 | See also \fBzfs_arc_meta_limit\fR which serves a similar purpose but has a | |
698 | higher priority if set to nonzero value. | |
699 | ||
700 | .sp | |
be54a13c | 701 | Default value: \fB75\fR%. |
9907cc1c G |
702 | .RE |
703 | ||
ca0bf58d PS |
704 | .sp |
705 | .ne 2 | |
706 | .na | |
707 | \fBzfs_arc_meta_min\fR (ulong) | |
708 | .ad | |
709 | .RS 12n | |
710 | The minimum allowed size in bytes that meta data buffers may consume in | |
711 | the ARC. This value defaults to 0 which disables a floor on the amount | |
712 | of the ARC devoted meta data. | |
713 | .sp | |
714 | Default value: \fB0\fR. | |
715 | .RE | |
716 | ||
29714574 TF |
717 | .sp |
718 | .ne 2 | |
719 | .na | |
720 | \fBzfs_arc_meta_prune\fR (int) | |
721 | .ad | |
722 | .RS 12n | |
2cbb06b5 BB |
723 | The number of dentries and inodes to be scanned looking for entries |
724 | which can be dropped. This may be required when the ARC reaches the | |
725 | \fBzfs_arc_meta_limit\fR because dentries and inodes can pin buffers | |
726 | in the ARC. Increasing this value will cause to dentry and inode caches | |
727 | to be pruned more aggressively. Setting this value to 0 will disable | |
728 | pruning the inode and dentry caches. | |
29714574 | 729 | .sp |
2cbb06b5 | 730 | Default value: \fB10,000\fR. |
29714574 TF |
731 | .RE |
732 | ||
bc888666 BB |
733 | .sp |
734 | .ne 2 | |
735 | .na | |
ca85d690 | 736 | \fBzfs_arc_meta_strategy\fR (int) |
bc888666 BB |
737 | .ad |
738 | .RS 12n | |
ca85d690 | 739 | Define the strategy for ARC meta data buffer eviction (meta reclaim strategy). |
740 | A value of 0 (META_ONLY) will evict only the ARC meta data buffers. | |
d4a72f23 | 741 | A value of 1 (BALANCED) indicates that additional data buffers may be evicted if |
ca85d690 | 742 | that is required to in order to evict the required number of meta data buffers. |
bc888666 | 743 | .sp |
ca85d690 | 744 | Default value: \fB1\fR. |
bc888666 BB |
745 | .RE |
746 | ||
29714574 TF |
747 | .sp |
748 | .ne 2 | |
749 | .na | |
750 | \fBzfs_arc_min\fR (ulong) | |
751 | .ad | |
752 | .RS 12n | |
ca85d690 | 753 | Min arc size of ARC in bytes. If set to 0 then arc_c_min will default to |
754 | consuming the larger of 32M or 1/32 of total system memory. | |
29714574 | 755 | .sp |
ca85d690 | 756 | Default value: \fB0\fR. |
29714574 TF |
757 | .RE |
758 | ||
759 | .sp | |
760 | .ne 2 | |
761 | .na | |
d4a72f23 | 762 | \fBzfs_arc_min_prefetch_ms\fR (int) |
29714574 TF |
763 | .ad |
764 | .RS 12n | |
d4a72f23 | 765 | Minimum time prefetched blocks are locked in the ARC, specified in ms. |
2b84817f | 766 | A value of \fB0\fR will default to 1000 ms. |
d4a72f23 TC |
767 | .sp |
768 | Default value: \fB0\fR. | |
769 | .RE | |
770 | ||
771 | .sp | |
772 | .ne 2 | |
773 | .na | |
774 | \fBzfs_arc_min_prescient_prefetch_ms\fR (int) | |
775 | .ad | |
776 | .RS 12n | |
777 | Minimum time "prescient prefetched" blocks are locked in the ARC, specified | |
778 | in ms. These blocks are meant to be prefetched fairly aggresively ahead of | |
2b84817f | 779 | the code that may use them. A value of \fB0\fR will default to 6000 ms. |
29714574 | 780 | .sp |
83426735 | 781 | Default value: \fB0\fR. |
29714574 TF |
782 | .RE |
783 | ||
6cb8e530 PZ |
784 | .sp |
785 | .ne 2 | |
786 | .na | |
787 | \fBzfs_max_missing_tvds\fR (int) | |
788 | .ad | |
789 | .RS 12n | |
790 | Number of missing top-level vdevs which will be allowed during | |
791 | pool import (only in read-only mode). | |
792 | .sp | |
793 | Default value: \fB0\fR | |
794 | .RE | |
795 | ||
ca0bf58d PS |
796 | .sp |
797 | .ne 2 | |
798 | .na | |
c30e58c4 | 799 | \fBzfs_multilist_num_sublists\fR (int) |
ca0bf58d PS |
800 | .ad |
801 | .RS 12n | |
802 | To allow more fine-grained locking, each ARC state contains a series | |
803 | of lists for both data and meta data objects. Locking is performed at | |
804 | the level of these "sub-lists". This parameters controls the number of | |
c30e58c4 MA |
805 | sub-lists per ARC state, and also applies to other uses of the |
806 | multilist data structure. | |
ca0bf58d | 807 | .sp |
c30e58c4 | 808 | Default value: \fB4\fR or the number of online CPUs, whichever is greater |
ca0bf58d PS |
809 | .RE |
810 | ||
811 | .sp | |
812 | .ne 2 | |
813 | .na | |
814 | \fBzfs_arc_overflow_shift\fR (int) | |
815 | .ad | |
816 | .RS 12n | |
817 | The ARC size is considered to be overflowing if it exceeds the current | |
818 | ARC target size (arc_c) by a threshold determined by this parameter. | |
819 | The threshold is calculated as a fraction of arc_c using the formula | |
820 | "arc_c >> \fBzfs_arc_overflow_shift\fR". | |
821 | ||
822 | The default value of 8 causes the ARC to be considered to be overflowing | |
823 | if it exceeds the target size by 1/256th (0.3%) of the target size. | |
824 | ||
825 | When the ARC is overflowing, new buffer allocations are stalled until | |
826 | the reclaim thread catches up and the overflow condition no longer exists. | |
827 | .sp | |
828 | Default value: \fB8\fR. | |
829 | .RE | |
830 | ||
728d6ae9 BB |
831 | .sp |
832 | .ne 2 | |
833 | .na | |
834 | ||
835 | \fBzfs_arc_p_min_shift\fR (int) | |
836 | .ad | |
837 | .RS 12n | |
ca85d690 | 838 | If set to a non zero value, this will update arc_p_min_shift (default 4) |
839 | with the new value. | |
d4a72f23 | 840 | arc_p_min_shift is used to shift of arc_c for calculating both min and max |
ca85d690 | 841 | max arc_p |
728d6ae9 | 842 | .sp |
ca85d690 | 843 | Default value: \fB0\fR. |
728d6ae9 BB |
844 | .RE |
845 | ||
62422785 PS |
846 | .sp |
847 | .ne 2 | |
848 | .na | |
849 | \fBzfs_arc_p_dampener_disable\fR (int) | |
850 | .ad | |
851 | .RS 12n | |
852 | Disable arc_p adapt dampener | |
853 | .sp | |
854 | Use \fB1\fR for yes (default) and \fB0\fR to disable. | |
855 | .RE | |
856 | ||
29714574 TF |
857 | .sp |
858 | .ne 2 | |
859 | .na | |
860 | \fBzfs_arc_shrink_shift\fR (int) | |
861 | .ad | |
862 | .RS 12n | |
ca85d690 | 863 | If set to a non zero value, this will update arc_shrink_shift (default 7) |
864 | with the new value. | |
29714574 | 865 | .sp |
ca85d690 | 866 | Default value: \fB0\fR. |
29714574 TF |
867 | .RE |
868 | ||
03b60eee DB |
869 | .sp |
870 | .ne 2 | |
871 | .na | |
872 | \fBzfs_arc_pc_percent\fR (uint) | |
873 | .ad | |
874 | .RS 12n | |
875 | Percent of pagecache to reclaim arc to | |
876 | ||
877 | This tunable allows ZFS arc to play more nicely with the kernel's LRU | |
878 | pagecache. It can guarantee that the arc size won't collapse under scanning | |
879 | pressure on the pagecache, yet still allows arc to be reclaimed down to | |
880 | zfs_arc_min if necessary. This value is specified as percent of pagecache | |
881 | size (as measured by NR_FILE_PAGES) where that percent may exceed 100. This | |
882 | only operates during memory pressure/reclaim. | |
883 | .sp | |
be54a13c | 884 | Default value: \fB0\fR% (disabled). |
03b60eee DB |
885 | .RE |
886 | ||
11f552fa BB |
887 | .sp |
888 | .ne 2 | |
889 | .na | |
890 | \fBzfs_arc_sys_free\fR (ulong) | |
891 | .ad | |
892 | .RS 12n | |
893 | The target number of bytes the ARC should leave as free memory on the system. | |
894 | Defaults to the larger of 1/64 of physical memory or 512K. Setting this | |
895 | option to a non-zero value will override the default. | |
896 | .sp | |
897 | Default value: \fB0\fR. | |
898 | .RE | |
899 | ||
29714574 TF |
900 | .sp |
901 | .ne 2 | |
902 | .na | |
903 | \fBzfs_autoimport_disable\fR (int) | |
904 | .ad | |
905 | .RS 12n | |
27b293be | 906 | Disable pool import at module load by ignoring the cache file (typically \fB/etc/zfs/zpool.cache\fR). |
29714574 | 907 | .sp |
70081096 | 908 | Use \fB1\fR for yes (default) and \fB0\fR for no. |
29714574 TF |
909 | .RE |
910 | ||
80d52c39 TH |
911 | .sp |
912 | .ne 2 | |
913 | .na | |
914 | \fBzfs_checksums_per_second\fR (int) | |
915 | .ad | |
916 | .RS 12n | |
917 | Rate limit checksum events to this many per second. Note that this should | |
918 | not be set below the zed thresholds (currently 10 checksums over 10 sec) | |
919 | or else zed may not trigger any action. | |
920 | .sp | |
921 | Default value: 20 | |
922 | .RE | |
923 | ||
2fe61a7e PS |
924 | .sp |
925 | .ne 2 | |
926 | .na | |
927 | \fBzfs_commit_timeout_pct\fR (int) | |
928 | .ad | |
929 | .RS 12n | |
930 | This controls the amount of time that a ZIL block (lwb) will remain "open" | |
931 | when it isn't "full", and it has a thread waiting for it to be committed to | |
932 | stable storage. The timeout is scaled based on a percentage of the last lwb | |
933 | latency to avoid significantly impacting the latency of each individual | |
934 | transaction record (itx). | |
935 | .sp | |
be54a13c | 936 | Default value: \fB5\fR%. |
2fe61a7e PS |
937 | .RE |
938 | ||
0dc2f70c MA |
939 | .sp |
940 | .ne 2 | |
941 | .na | |
942 | \fBzfs_condense_indirect_vdevs_enable\fR (int) | |
943 | .ad | |
944 | .RS 12n | |
945 | Enable condensing indirect vdev mappings. When set to a non-zero value, | |
946 | attempt to condense indirect vdev mappings if the mapping uses more than | |
947 | \fBzfs_condense_min_mapping_bytes\fR bytes of memory and if the obsolete | |
948 | space map object uses more than \fBzfs_condense_max_obsolete_bytes\fR | |
949 | bytes on-disk. The condensing process is an attempt to save memory by | |
950 | removing obsolete mappings. | |
951 | .sp | |
952 | Default value: \fB1\fR. | |
953 | .RE | |
954 | ||
955 | .sp | |
956 | .ne 2 | |
957 | .na | |
958 | \fBzfs_condense_max_obsolete_bytes\fR (ulong) | |
959 | .ad | |
960 | .RS 12n | |
961 | Only attempt to condense indirect vdev mappings if the on-disk size | |
962 | of the obsolete space map object is greater than this number of bytes | |
963 | (see \fBfBzfs_condense_indirect_vdevs_enable\fR). | |
964 | .sp | |
965 | Default value: \fB1,073,741,824\fR. | |
966 | .RE | |
967 | ||
968 | .sp | |
969 | .ne 2 | |
970 | .na | |
971 | \fBzfs_condense_min_mapping_bytes\fR (ulong) | |
972 | .ad | |
973 | .RS 12n | |
974 | Minimum size vdev mapping to attempt to condense (see | |
975 | \fBzfs_condense_indirect_vdevs_enable\fR). | |
976 | .sp | |
977 | Default value: \fB131,072\fR. | |
978 | .RE | |
979 | ||
3b36f831 BB |
980 | .sp |
981 | .ne 2 | |
982 | .na | |
983 | \fBzfs_dbgmsg_enable\fR (int) | |
984 | .ad | |
985 | .RS 12n | |
986 | Internally ZFS keeps a small log to facilitate debugging. By default the log | |
987 | is disabled, to enable it set this option to 1. The contents of the log can | |
988 | be accessed by reading the /proc/spl/kstat/zfs/dbgmsg file. Writing 0 to | |
989 | this proc file clears the log. | |
990 | .sp | |
991 | Default value: \fB0\fR. | |
992 | .RE | |
993 | ||
994 | .sp | |
995 | .ne 2 | |
996 | .na | |
997 | \fBzfs_dbgmsg_maxsize\fR (int) | |
998 | .ad | |
999 | .RS 12n | |
1000 | The maximum size in bytes of the internal ZFS debug log. | |
1001 | .sp | |
1002 | Default value: \fB4M\fR. | |
1003 | .RE | |
1004 | ||
29714574 TF |
1005 | .sp |
1006 | .ne 2 | |
1007 | .na | |
1008 | \fBzfs_dbuf_state_index\fR (int) | |
1009 | .ad | |
1010 | .RS 12n | |
83426735 D |
1011 | This feature is currently unused. It is normally used for controlling what |
1012 | reporting is available under /proc/spl/kstat/zfs. | |
29714574 TF |
1013 | .sp |
1014 | Default value: \fB0\fR. | |
1015 | .RE | |
1016 | ||
1017 | .sp | |
1018 | .ne 2 | |
1019 | .na | |
1020 | \fBzfs_deadman_enabled\fR (int) | |
1021 | .ad | |
1022 | .RS 12n | |
b81a3ddc | 1023 | When a pool sync operation takes longer than \fBzfs_deadman_synctime_ms\fR |
8fb1ede1 BB |
1024 | milliseconds, or when an individual I/O takes longer than |
1025 | \fBzfs_deadman_ziotime_ms\fR milliseconds, then the operation is considered to | |
1026 | be "hung". If \fBzfs_deadman_enabled\fR is set then the deadman behavior is | |
1027 | invoked as described by the \fBzfs_deadman_failmode\fR module option. | |
1028 | By default the deadman is enabled and configured to \fBwait\fR which results | |
1029 | in "hung" I/Os only being logged. The deadman is automatically disabled | |
1030 | when a pool gets suspended. | |
29714574 | 1031 | .sp |
8fb1ede1 BB |
1032 | Default value: \fB1\fR. |
1033 | .RE | |
1034 | ||
1035 | .sp | |
1036 | .ne 2 | |
1037 | .na | |
1038 | \fBzfs_deadman_failmode\fR (charp) | |
1039 | .ad | |
1040 | .RS 12n | |
1041 | Controls the failure behavior when the deadman detects a "hung" I/O. Valid | |
1042 | values are \fBwait\fR, \fBcontinue\fR, and \fBpanic\fR. | |
1043 | .sp | |
1044 | \fBwait\fR - Wait for a "hung" I/O to complete. For each "hung" I/O a | |
1045 | "deadman" event will be posted describing that I/O. | |
1046 | .sp | |
1047 | \fBcontinue\fR - Attempt to recover from a "hung" I/O by re-dispatching it | |
1048 | to the I/O pipeline if possible. | |
1049 | .sp | |
1050 | \fBpanic\fR - Panic the system. This can be used to facilitate an automatic | |
1051 | fail-over to a properly configured fail-over partner. | |
1052 | .sp | |
1053 | Default value: \fBwait\fR. | |
b81a3ddc TC |
1054 | .RE |
1055 | ||
1056 | .sp | |
1057 | .ne 2 | |
1058 | .na | |
1059 | \fBzfs_deadman_checktime_ms\fR (int) | |
1060 | .ad | |
1061 | .RS 12n | |
8fb1ede1 BB |
1062 | Check time in milliseconds. This defines the frequency at which we check |
1063 | for hung I/O and potentially invoke the \fBzfs_deadman_failmode\fR behavior. | |
b81a3ddc | 1064 | .sp |
8fb1ede1 | 1065 | Default value: \fB60,000\fR. |
29714574 TF |
1066 | .RE |
1067 | ||
1068 | .sp | |
1069 | .ne 2 | |
1070 | .na | |
e8b96c60 | 1071 | \fBzfs_deadman_synctime_ms\fR (ulong) |
29714574 TF |
1072 | .ad |
1073 | .RS 12n | |
b81a3ddc | 1074 | Interval in milliseconds after which the deadman is triggered and also |
8fb1ede1 BB |
1075 | the interval after which a pool sync operation is considered to be "hung". |
1076 | Once this limit is exceeded the deadman will be invoked every | |
1077 | \fBzfs_deadman_checktime_ms\fR milliseconds until the pool sync completes. | |
1078 | .sp | |
1079 | Default value: \fB600,000\fR. | |
1080 | .RE | |
b81a3ddc | 1081 | |
29714574 | 1082 | .sp |
8fb1ede1 BB |
1083 | .ne 2 |
1084 | .na | |
1085 | \fBzfs_deadman_ziotime_ms\fR (ulong) | |
1086 | .ad | |
1087 | .RS 12n | |
1088 | Interval in milliseconds after which the deadman is triggered and an | |
1089 | individual IO operation is considered to be "hung". As long as the I/O | |
1090 | remains "hung" the deadman will be invoked every \fBzfs_deadman_checktime_ms\fR | |
1091 | milliseconds until the I/O completes. | |
1092 | .sp | |
1093 | Default value: \fB300,000\fR. | |
29714574 TF |
1094 | .RE |
1095 | ||
1096 | .sp | |
1097 | .ne 2 | |
1098 | .na | |
1099 | \fBzfs_dedup_prefetch\fR (int) | |
1100 | .ad | |
1101 | .RS 12n | |
1102 | Enable prefetching dedup-ed blks | |
1103 | .sp | |
0dfc7324 | 1104 | Use \fB1\fR for yes and \fB0\fR to disable (default). |
29714574 TF |
1105 | .RE |
1106 | ||
e8b96c60 MA |
1107 | .sp |
1108 | .ne 2 | |
1109 | .na | |
1110 | \fBzfs_delay_min_dirty_percent\fR (int) | |
1111 | .ad | |
1112 | .RS 12n | |
1113 | Start to delay each transaction once there is this amount of dirty data, | |
1114 | expressed as a percentage of \fBzfs_dirty_data_max\fR. | |
1115 | This value should be >= zfs_vdev_async_write_active_max_dirty_percent. | |
1116 | See the section "ZFS TRANSACTION DELAY". | |
1117 | .sp | |
be54a13c | 1118 | Default value: \fB60\fR%. |
e8b96c60 MA |
1119 | .RE |
1120 | ||
1121 | .sp | |
1122 | .ne 2 | |
1123 | .na | |
1124 | \fBzfs_delay_scale\fR (int) | |
1125 | .ad | |
1126 | .RS 12n | |
1127 | This controls how quickly the transaction delay approaches infinity. | |
1128 | Larger values cause longer delays for a given amount of dirty data. | |
1129 | .sp | |
1130 | For the smoothest delay, this value should be about 1 billion divided | |
1131 | by the maximum number of operations per second. This will smoothly | |
1132 | handle between 10x and 1/10th this number. | |
1133 | .sp | |
1134 | See the section "ZFS TRANSACTION DELAY". | |
1135 | .sp | |
1136 | Note: \fBzfs_delay_scale\fR * \fBzfs_dirty_data_max\fR must be < 2^64. | |
1137 | .sp | |
1138 | Default value: \fB500,000\fR. | |
1139 | .RE | |
1140 | ||
80d52c39 TH |
1141 | .sp |
1142 | .ne 2 | |
1143 | .na | |
1144 | \fBzfs_delays_per_second\fR (int) | |
1145 | .ad | |
1146 | .RS 12n | |
1147 | Rate limit IO delay events to this many per second. | |
1148 | .sp | |
1149 | Default value: 20 | |
1150 | .RE | |
1151 | ||
a966c564 K |
1152 | .sp |
1153 | .ne 2 | |
1154 | .na | |
1155 | \fBzfs_delete_blocks\fR (ulong) | |
1156 | .ad | |
1157 | .RS 12n | |
1158 | This is the used to define a large file for the purposes of delete. Files | |
1159 | containing more than \fBzfs_delete_blocks\fR will be deleted asynchronously | |
1160 | while smaller files are deleted synchronously. Decreasing this value will | |
1161 | reduce the time spent in an unlink(2) system call at the expense of a longer | |
1162 | delay before the freed space is available. | |
1163 | .sp | |
1164 | Default value: \fB20,480\fR. | |
1165 | .RE | |
1166 | ||
e8b96c60 MA |
1167 | .sp |
1168 | .ne 2 | |
1169 | .na | |
1170 | \fBzfs_dirty_data_max\fR (int) | |
1171 | .ad | |
1172 | .RS 12n | |
1173 | Determines the dirty space limit in bytes. Once this limit is exceeded, new | |
1174 | writes are halted until space frees up. This parameter takes precedence | |
1175 | over \fBzfs_dirty_data_max_percent\fR. | |
1176 | See the section "ZFS TRANSACTION DELAY". | |
1177 | .sp | |
be54a13c | 1178 | Default value: \fB10\fR% of physical RAM, capped at \fBzfs_dirty_data_max_max\fR. |
e8b96c60 MA |
1179 | .RE |
1180 | ||
1181 | .sp | |
1182 | .ne 2 | |
1183 | .na | |
1184 | \fBzfs_dirty_data_max_max\fR (int) | |
1185 | .ad | |
1186 | .RS 12n | |
1187 | Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed in bytes. | |
1188 | This limit is only enforced at module load time, and will be ignored if | |
1189 | \fBzfs_dirty_data_max\fR is later changed. This parameter takes | |
1190 | precedence over \fBzfs_dirty_data_max_max_percent\fR. See the section | |
1191 | "ZFS TRANSACTION DELAY". | |
1192 | .sp | |
be54a13c | 1193 | Default value: \fB25\fR% of physical RAM. |
e8b96c60 MA |
1194 | .RE |
1195 | ||
1196 | .sp | |
1197 | .ne 2 | |
1198 | .na | |
1199 | \fBzfs_dirty_data_max_max_percent\fR (int) | |
1200 | .ad | |
1201 | .RS 12n | |
1202 | Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed as a | |
1203 | percentage of physical RAM. This limit is only enforced at module load | |
1204 | time, and will be ignored if \fBzfs_dirty_data_max\fR is later changed. | |
1205 | The parameter \fBzfs_dirty_data_max_max\fR takes precedence over this | |
1206 | one. See the section "ZFS TRANSACTION DELAY". | |
1207 | .sp | |
be54a13c | 1208 | Default value: \fB25\fR%. |
e8b96c60 MA |
1209 | .RE |
1210 | ||
1211 | .sp | |
1212 | .ne 2 | |
1213 | .na | |
1214 | \fBzfs_dirty_data_max_percent\fR (int) | |
1215 | .ad | |
1216 | .RS 12n | |
1217 | Determines the dirty space limit, expressed as a percentage of all | |
1218 | memory. Once this limit is exceeded, new writes are halted until space frees | |
1219 | up. The parameter \fBzfs_dirty_data_max\fR takes precedence over this | |
1220 | one. See the section "ZFS TRANSACTION DELAY". | |
1221 | .sp | |
be54a13c | 1222 | Default value: \fB10\fR%, subject to \fBzfs_dirty_data_max_max\fR. |
e8b96c60 MA |
1223 | .RE |
1224 | ||
1225 | .sp | |
1226 | .ne 2 | |
1227 | .na | |
dfbe2675 | 1228 | \fBzfs_dirty_data_sync_percent\fR (int) |
e8b96c60 MA |
1229 | .ad |
1230 | .RS 12n | |
dfbe2675 MA |
1231 | Start syncing out a transaction group if there's at least this much dirty data |
1232 | as a percentage of \fBzfs_dirty_data_max\fR. This should be less than | |
1233 | \fBzfs_vdev_async_write_active_min_dirty_percent\fR. | |
e8b96c60 | 1234 | .sp |
dfbe2675 | 1235 | Default value: \fB20\fR% of \fBzfs_dirty_data_max\fR. |
e8b96c60 MA |
1236 | .RE |
1237 | ||
1eeb4562 JX |
1238 | .sp |
1239 | .ne 2 | |
1240 | .na | |
1241 | \fBzfs_fletcher_4_impl\fR (string) | |
1242 | .ad | |
1243 | .RS 12n | |
1244 | Select a fletcher 4 implementation. | |
1245 | .sp | |
35a76a03 | 1246 | Supported selectors are: \fBfastest\fR, \fBscalar\fR, \fBsse2\fR, \fBssse3\fR, |
24cdeaf1 | 1247 | \fBavx2\fR, \fBavx512f\fR, and \fBaarch64_neon\fR. |
70b258fc GN |
1248 | All of the selectors except \fBfastest\fR and \fBscalar\fR require instruction |
1249 | set extensions to be available and will only appear if ZFS detects that they are | |
1250 | present at runtime. If multiple implementations of fletcher 4 are available, | |
1251 | the \fBfastest\fR will be chosen using a micro benchmark. Selecting \fBscalar\fR | |
1252 | results in the original, CPU based calculation, being used. Selecting any option | |
1253 | other than \fBfastest\fR and \fBscalar\fR results in vector instructions from | |
1254 | the respective CPU instruction set being used. | |
1eeb4562 JX |
1255 | .sp |
1256 | Default value: \fBfastest\fR. | |
1257 | .RE | |
1258 | ||
ba5ad9a4 GW |
1259 | .sp |
1260 | .ne 2 | |
1261 | .na | |
1262 | \fBzfs_free_bpobj_enabled\fR (int) | |
1263 | .ad | |
1264 | .RS 12n | |
1265 | Enable/disable the processing of the free_bpobj object. | |
1266 | .sp | |
1267 | Default value: \fB1\fR. | |
1268 | .RE | |
1269 | ||
36283ca2 MG |
1270 | .sp |
1271 | .ne 2 | |
1272 | .na | |
a1d477c2 | 1273 | \fBzfs_async_block_max_blocks\fR (ulong) |
36283ca2 MG |
1274 | .ad |
1275 | .RS 12n | |
1276 | Maximum number of blocks freed in a single txg. | |
1277 | .sp | |
1278 | Default value: \fB100,000\fR. | |
1279 | .RE | |
1280 | ||
ca0845d5 PD |
1281 | .sp |
1282 | .ne 2 | |
1283 | .na | |
1284 | \fBzfs_override_estimate_recordsize\fR (ulong) | |
1285 | .ad | |
1286 | .RS 12n | |
1287 | Record size calculation override for zfs send estimates. | |
1288 | .sp | |
1289 | Default value: \fB0\fR. | |
1290 | .RE | |
1291 | ||
e8b96c60 MA |
1292 | .sp |
1293 | .ne 2 | |
1294 | .na | |
1295 | \fBzfs_vdev_async_read_max_active\fR (int) | |
1296 | .ad | |
1297 | .RS 12n | |
83426735 | 1298 | Maximum asynchronous read I/Os active to each device. |
e8b96c60 MA |
1299 | See the section "ZFS I/O SCHEDULER". |
1300 | .sp | |
1301 | Default value: \fB3\fR. | |
1302 | .RE | |
1303 | ||
1304 | .sp | |
1305 | .ne 2 | |
1306 | .na | |
1307 | \fBzfs_vdev_async_read_min_active\fR (int) | |
1308 | .ad | |
1309 | .RS 12n | |
1310 | Minimum asynchronous read I/Os active to each device. | |
1311 | See the section "ZFS I/O SCHEDULER". | |
1312 | .sp | |
1313 | Default value: \fB1\fR. | |
1314 | .RE | |
1315 | ||
1316 | .sp | |
1317 | .ne 2 | |
1318 | .na | |
1319 | \fBzfs_vdev_async_write_active_max_dirty_percent\fR (int) | |
1320 | .ad | |
1321 | .RS 12n | |
1322 | When the pool has more than | |
1323 | \fBzfs_vdev_async_write_active_max_dirty_percent\fR dirty data, use | |
1324 | \fBzfs_vdev_async_write_max_active\fR to limit active async writes. If | |
1325 | the dirty data is between min and max, the active I/O limit is linearly | |
1326 | interpolated. See the section "ZFS I/O SCHEDULER". | |
1327 | .sp | |
be54a13c | 1328 | Default value: \fB60\fR%. |
e8b96c60 MA |
1329 | .RE |
1330 | ||
1331 | .sp | |
1332 | .ne 2 | |
1333 | .na | |
1334 | \fBzfs_vdev_async_write_active_min_dirty_percent\fR (int) | |
1335 | .ad | |
1336 | .RS 12n | |
1337 | When the pool has less than | |
1338 | \fBzfs_vdev_async_write_active_min_dirty_percent\fR dirty data, use | |
1339 | \fBzfs_vdev_async_write_min_active\fR to limit active async writes. If | |
1340 | the dirty data is between min and max, the active I/O limit is linearly | |
1341 | interpolated. See the section "ZFS I/O SCHEDULER". | |
1342 | .sp | |
be54a13c | 1343 | Default value: \fB30\fR%. |
e8b96c60 MA |
1344 | .RE |
1345 | ||
1346 | .sp | |
1347 | .ne 2 | |
1348 | .na | |
1349 | \fBzfs_vdev_async_write_max_active\fR (int) | |
1350 | .ad | |
1351 | .RS 12n | |
83426735 | 1352 | Maximum asynchronous write I/Os active to each device. |
e8b96c60 MA |
1353 | See the section "ZFS I/O SCHEDULER". |
1354 | .sp | |
1355 | Default value: \fB10\fR. | |
1356 | .RE | |
1357 | ||
1358 | .sp | |
1359 | .ne 2 | |
1360 | .na | |
1361 | \fBzfs_vdev_async_write_min_active\fR (int) | |
1362 | .ad | |
1363 | .RS 12n | |
1364 | Minimum asynchronous write I/Os active to each device. | |
1365 | See the section "ZFS I/O SCHEDULER". | |
1366 | .sp | |
06226b59 D |
1367 | Lower values are associated with better latency on rotational media but poorer |
1368 | resilver performance. The default value of 2 was chosen as a compromise. A | |
1369 | value of 3 has been shown to improve resilver performance further at a cost of | |
1370 | further increasing latency. | |
1371 | .sp | |
1372 | Default value: \fB2\fR. | |
e8b96c60 MA |
1373 | .RE |
1374 | ||
1375 | .sp | |
1376 | .ne 2 | |
1377 | .na | |
1378 | \fBzfs_vdev_max_active\fR (int) | |
1379 | .ad | |
1380 | .RS 12n | |
1381 | The maximum number of I/Os active to each device. Ideally, this will be >= | |
1382 | the sum of each queue's max_active. It must be at least the sum of each | |
1383 | queue's min_active. See the section "ZFS I/O SCHEDULER". | |
1384 | .sp | |
1385 | Default value: \fB1,000\fR. | |
1386 | .RE | |
1387 | ||
1388 | .sp | |
1389 | .ne 2 | |
1390 | .na | |
1391 | \fBzfs_vdev_scrub_max_active\fR (int) | |
1392 | .ad | |
1393 | .RS 12n | |
83426735 | 1394 | Maximum scrub I/Os active to each device. |
e8b96c60 MA |
1395 | See the section "ZFS I/O SCHEDULER". |
1396 | .sp | |
1397 | Default value: \fB2\fR. | |
1398 | .RE | |
1399 | ||
1400 | .sp | |
1401 | .ne 2 | |
1402 | .na | |
1403 | \fBzfs_vdev_scrub_min_active\fR (int) | |
1404 | .ad | |
1405 | .RS 12n | |
1406 | Minimum scrub I/Os active to each device. | |
1407 | See the section "ZFS I/O SCHEDULER". | |
1408 | .sp | |
1409 | Default value: \fB1\fR. | |
1410 | .RE | |
1411 | ||
1412 | .sp | |
1413 | .ne 2 | |
1414 | .na | |
1415 | \fBzfs_vdev_sync_read_max_active\fR (int) | |
1416 | .ad | |
1417 | .RS 12n | |
83426735 | 1418 | Maximum synchronous read I/Os active to each device. |
e8b96c60 MA |
1419 | See the section "ZFS I/O SCHEDULER". |
1420 | .sp | |
1421 | Default value: \fB10\fR. | |
1422 | .RE | |
1423 | ||
1424 | .sp | |
1425 | .ne 2 | |
1426 | .na | |
1427 | \fBzfs_vdev_sync_read_min_active\fR (int) | |
1428 | .ad | |
1429 | .RS 12n | |
1430 | Minimum synchronous read I/Os active to each device. | |
1431 | See the section "ZFS I/O SCHEDULER". | |
1432 | .sp | |
1433 | Default value: \fB10\fR. | |
1434 | .RE | |
1435 | ||
1436 | .sp | |
1437 | .ne 2 | |
1438 | .na | |
1439 | \fBzfs_vdev_sync_write_max_active\fR (int) | |
1440 | .ad | |
1441 | .RS 12n | |
83426735 | 1442 | Maximum synchronous write I/Os active to each device. |
e8b96c60 MA |
1443 | See the section "ZFS I/O SCHEDULER". |
1444 | .sp | |
1445 | Default value: \fB10\fR. | |
1446 | .RE | |
1447 | ||
1448 | .sp | |
1449 | .ne 2 | |
1450 | .na | |
1451 | \fBzfs_vdev_sync_write_min_active\fR (int) | |
1452 | .ad | |
1453 | .RS 12n | |
1454 | Minimum synchronous write I/Os active to each device. | |
1455 | See the section "ZFS I/O SCHEDULER". | |
1456 | .sp | |
1457 | Default value: \fB10\fR. | |
1458 | .RE | |
1459 | ||
3dfb57a3 DB |
1460 | .sp |
1461 | .ne 2 | |
1462 | .na | |
1463 | \fBzfs_vdev_queue_depth_pct\fR (int) | |
1464 | .ad | |
1465 | .RS 12n | |
e815485f TC |
1466 | Maximum number of queued allocations per top-level vdev expressed as |
1467 | a percentage of \fBzfs_vdev_async_write_max_active\fR which allows the | |
1468 | system to detect devices that are more capable of handling allocations | |
1469 | and to allocate more blocks to those devices. It allows for dynamic | |
1470 | allocation distribution when devices are imbalanced as fuller devices | |
1471 | will tend to be slower than empty devices. | |
1472 | ||
1473 | See also \fBzio_dva_throttle_enabled\fR. | |
3dfb57a3 | 1474 | .sp |
be54a13c | 1475 | Default value: \fB1000\fR%. |
3dfb57a3 DB |
1476 | .RE |
1477 | ||
29714574 TF |
1478 | .sp |
1479 | .ne 2 | |
1480 | .na | |
1481 | \fBzfs_expire_snapshot\fR (int) | |
1482 | .ad | |
1483 | .RS 12n | |
1484 | Seconds to expire .zfs/snapshot | |
1485 | .sp | |
1486 | Default value: \fB300\fR. | |
1487 | .RE | |
1488 | ||
0500e835 BB |
1489 | .sp |
1490 | .ne 2 | |
1491 | .na | |
1492 | \fBzfs_admin_snapshot\fR (int) | |
1493 | .ad | |
1494 | .RS 12n | |
1495 | Allow the creation, removal, or renaming of entries in the .zfs/snapshot | |
1496 | directory to cause the creation, destruction, or renaming of snapshots. | |
1497 | When enabled this functionality works both locally and over NFS exports | |
1498 | which have the 'no_root_squash' option set. This functionality is disabled | |
1499 | by default. | |
1500 | .sp | |
1501 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1502 | .RE | |
1503 | ||
29714574 TF |
1504 | .sp |
1505 | .ne 2 | |
1506 | .na | |
1507 | \fBzfs_flags\fR (int) | |
1508 | .ad | |
1509 | .RS 12n | |
33b6dbbc NB |
1510 | Set additional debugging flags. The following flags may be bitwise-or'd |
1511 | together. | |
1512 | .sp | |
1513 | .TS | |
1514 | box; | |
1515 | rB lB | |
1516 | lB lB | |
1517 | r l. | |
1518 | Value Symbolic Name | |
1519 | Description | |
1520 | _ | |
1521 | 1 ZFS_DEBUG_DPRINTF | |
1522 | Enable dprintf entries in the debug log. | |
1523 | _ | |
1524 | 2 ZFS_DEBUG_DBUF_VERIFY * | |
1525 | Enable extra dbuf verifications. | |
1526 | _ | |
1527 | 4 ZFS_DEBUG_DNODE_VERIFY * | |
1528 | Enable extra dnode verifications. | |
1529 | _ | |
1530 | 8 ZFS_DEBUG_SNAPNAMES | |
1531 | Enable snapshot name verification. | |
1532 | _ | |
1533 | 16 ZFS_DEBUG_MODIFY | |
1534 | Check for illegally modified ARC buffers. | |
1535 | _ | |
33b6dbbc NB |
1536 | 64 ZFS_DEBUG_ZIO_FREE |
1537 | Enable verification of block frees. | |
1538 | _ | |
1539 | 128 ZFS_DEBUG_HISTOGRAM_VERIFY | |
1540 | Enable extra spacemap histogram verifications. | |
8740cf4a NB |
1541 | _ |
1542 | 256 ZFS_DEBUG_METASLAB_VERIFY | |
1543 | Verify space accounting on disk matches in-core range_trees. | |
1544 | _ | |
1545 | 512 ZFS_DEBUG_SET_ERROR | |
1546 | Enable SET_ERROR and dprintf entries in the debug log. | |
33b6dbbc NB |
1547 | .TE |
1548 | .sp | |
1549 | * Requires debug build. | |
29714574 | 1550 | .sp |
33b6dbbc | 1551 | Default value: \fB0\fR. |
29714574 TF |
1552 | .RE |
1553 | ||
fbeddd60 MA |
1554 | .sp |
1555 | .ne 2 | |
1556 | .na | |
1557 | \fBzfs_free_leak_on_eio\fR (int) | |
1558 | .ad | |
1559 | .RS 12n | |
1560 | If destroy encounters an EIO while reading metadata (e.g. indirect | |
1561 | blocks), space referenced by the missing metadata can not be freed. | |
1562 | Normally this causes the background destroy to become "stalled", as | |
1563 | it is unable to make forward progress. While in this stalled state, | |
1564 | all remaining space to free from the error-encountering filesystem is | |
1565 | "temporarily leaked". Set this flag to cause it to ignore the EIO, | |
1566 | permanently leak the space from indirect blocks that can not be read, | |
1567 | and continue to free everything else that it can. | |
1568 | ||
1569 | The default, "stalling" behavior is useful if the storage partially | |
1570 | fails (i.e. some but not all i/os fail), and then later recovers. In | |
1571 | this case, we will be able to continue pool operations while it is | |
1572 | partially failed, and when it recovers, we can continue to free the | |
1573 | space, with no leaks. However, note that this case is actually | |
1574 | fairly rare. | |
1575 | ||
1576 | Typically pools either (a) fail completely (but perhaps temporarily, | |
1577 | e.g. a top-level vdev going offline), or (b) have localized, | |
1578 | permanent errors (e.g. disk returns the wrong data due to bit flip or | |
1579 | firmware bug). In case (a), this setting does not matter because the | |
1580 | pool will be suspended and the sync thread will not be able to make | |
1581 | forward progress regardless. In case (b), because the error is | |
1582 | permanent, the best we can do is leak the minimum amount of space, | |
1583 | which is what setting this flag will do. Therefore, it is reasonable | |
1584 | for this flag to normally be set, but we chose the more conservative | |
1585 | approach of not setting it, so that there is no possibility of | |
1586 | leaking space in the "partial temporary" failure case. | |
1587 | .sp | |
1588 | Default value: \fB0\fR. | |
1589 | .RE | |
1590 | ||
29714574 TF |
1591 | .sp |
1592 | .ne 2 | |
1593 | .na | |
1594 | \fBzfs_free_min_time_ms\fR (int) | |
1595 | .ad | |
1596 | .RS 12n | |
6146e17e | 1597 | During a \fBzfs destroy\fR operation using \fBfeature@async_destroy\fR a minimum |
83426735 | 1598 | of this much time will be spent working on freeing blocks per txg. |
29714574 TF |
1599 | .sp |
1600 | Default value: \fB1,000\fR. | |
1601 | .RE | |
1602 | ||
1603 | .sp | |
1604 | .ne 2 | |
1605 | .na | |
1606 | \fBzfs_immediate_write_sz\fR (long) | |
1607 | .ad | |
1608 | .RS 12n | |
83426735 | 1609 | Largest data block to write to zil. Larger blocks will be treated as if the |
6146e17e | 1610 | dataset being written to had the property setting \fBlogbias=throughput\fR. |
29714574 TF |
1611 | .sp |
1612 | Default value: \fB32,768\fR. | |
1613 | .RE | |
1614 | ||
917f475f JG |
1615 | .sp |
1616 | .ne 2 | |
1617 | .na | |
1618 | \fBzfs_lua_max_instrlimit\fR (ulong) | |
1619 | .ad | |
1620 | .RS 12n | |
1621 | The maximum execution time limit that can be set for a ZFS channel program, | |
1622 | specified as a number of Lua instructions. | |
1623 | .sp | |
1624 | Default value: \fB100,000,000\fR. | |
1625 | .RE | |
1626 | ||
1627 | .sp | |
1628 | .ne 2 | |
1629 | .na | |
1630 | \fBzfs_lua_max_memlimit\fR (ulong) | |
1631 | .ad | |
1632 | .RS 12n | |
1633 | The maximum memory limit that can be set for a ZFS channel program, specified | |
1634 | in bytes. | |
1635 | .sp | |
1636 | Default value: \fB104,857,600\fR. | |
1637 | .RE | |
1638 | ||
a7ed98d8 SD |
1639 | .sp |
1640 | .ne 2 | |
1641 | .na | |
1642 | \fBzfs_max_dataset_nesting\fR (int) | |
1643 | .ad | |
1644 | .RS 12n | |
1645 | The maximum depth of nested datasets. This value can be tuned temporarily to | |
1646 | fix existing datasets that exceed the predefined limit. | |
1647 | .sp | |
1648 | Default value: \fB50\fR. | |
1649 | .RE | |
1650 | ||
f1512ee6 MA |
1651 | .sp |
1652 | .ne 2 | |
1653 | .na | |
1654 | \fBzfs_max_recordsize\fR (int) | |
1655 | .ad | |
1656 | .RS 12n | |
1657 | We currently support block sizes from 512 bytes to 16MB. The benefits of | |
1658 | larger blocks, and thus larger IO, need to be weighed against the cost of | |
1659 | COWing a giant block to modify one byte. Additionally, very large blocks | |
1660 | can have an impact on i/o latency, and also potentially on the memory | |
1661 | allocator. Therefore, we do not allow the recordsize to be set larger than | |
1662 | zfs_max_recordsize (default 1MB). Larger blocks can be created by changing | |
1663 | this tunable, and pools with larger blocks can always be imported and used, | |
1664 | regardless of this setting. | |
1665 | .sp | |
1666 | Default value: \fB1,048,576\fR. | |
1667 | .RE | |
1668 | ||
f3a7f661 GW |
1669 | .sp |
1670 | .ne 2 | |
1671 | .na | |
1672 | \fBzfs_metaslab_fragmentation_threshold\fR (int) | |
1673 | .ad | |
1674 | .RS 12n | |
1675 | Allow metaslabs to keep their active state as long as their fragmentation | |
1676 | percentage is less than or equal to this value. An active metaslab that | |
1677 | exceeds this threshold will no longer keep its active status allowing | |
1678 | better metaslabs to be selected. | |
1679 | .sp | |
1680 | Default value: \fB70\fR. | |
1681 | .RE | |
1682 | ||
1683 | .sp | |
1684 | .ne 2 | |
1685 | .na | |
1686 | \fBzfs_mg_fragmentation_threshold\fR (int) | |
1687 | .ad | |
1688 | .RS 12n | |
1689 | Metaslab groups are considered eligible for allocations if their | |
83426735 | 1690 | fragmentation metric (measured as a percentage) is less than or equal to |
f3a7f661 GW |
1691 | this value. If a metaslab group exceeds this threshold then it will be |
1692 | skipped unless all metaslab groups within the metaslab class have also | |
1693 | crossed this threshold. | |
1694 | .sp | |
1695 | Default value: \fB85\fR. | |
1696 | .RE | |
1697 | ||
f4a4046b TC |
1698 | .sp |
1699 | .ne 2 | |
1700 | .na | |
1701 | \fBzfs_mg_noalloc_threshold\fR (int) | |
1702 | .ad | |
1703 | .RS 12n | |
1704 | Defines a threshold at which metaslab groups should be eligible for | |
1705 | allocations. The value is expressed as a percentage of free space | |
1706 | beyond which a metaslab group is always eligible for allocations. | |
1707 | If a metaslab group's free space is less than or equal to the | |
6b4e21c6 | 1708 | threshold, the allocator will avoid allocating to that group |
f4a4046b TC |
1709 | unless all groups in the pool have reached the threshold. Once all |
1710 | groups have reached the threshold, all groups are allowed to accept | |
1711 | allocations. The default value of 0 disables the feature and causes | |
1712 | all metaslab groups to be eligible for allocations. | |
1713 | ||
b58237e7 | 1714 | This parameter allows one to deal with pools having heavily imbalanced |
f4a4046b TC |
1715 | vdevs such as would be the case when a new vdev has been added. |
1716 | Setting the threshold to a non-zero percentage will stop allocations | |
1717 | from being made to vdevs that aren't filled to the specified percentage | |
1718 | and allow lesser filled vdevs to acquire more allocations than they | |
1719 | otherwise would under the old \fBzfs_mg_alloc_failures\fR facility. | |
1720 | .sp | |
1721 | Default value: \fB0\fR. | |
1722 | .RE | |
1723 | ||
cc99f275 DB |
1724 | .sp |
1725 | .ne 2 | |
1726 | .na | |
1727 | \fBzfs_ddt_data_is_special\fR (int) | |
1728 | .ad | |
1729 | .RS 12n | |
1730 | If enabled, ZFS will place DDT data into the special allocation class. | |
1731 | .sp | |
1732 | Default value: \fB1\fR. | |
1733 | .RE | |
1734 | ||
1735 | .sp | |
1736 | .ne 2 | |
1737 | .na | |
1738 | \fBzfs_user_indirect_is_special\fR (int) | |
1739 | .ad | |
1740 | .RS 12n | |
1741 | If enabled, ZFS will place user data (both file and zvol) indirect blocks | |
1742 | into the special allocation class. | |
1743 | .sp | |
1744 | Default value: \fB1\fR. | |
1745 | .RE | |
1746 | ||
379ca9cf OF |
1747 | .sp |
1748 | .ne 2 | |
1749 | .na | |
1750 | \fBzfs_multihost_history\fR (int) | |
1751 | .ad | |
1752 | .RS 12n | |
1753 | Historical statistics for the last N multihost updates will be available in | |
1754 | \fB/proc/spl/kstat/zfs/<pool>/multihost\fR | |
1755 | .sp | |
1756 | Default value: \fB0\fR. | |
1757 | .RE | |
1758 | ||
1759 | .sp | |
1760 | .ne 2 | |
1761 | .na | |
1762 | \fBzfs_multihost_interval\fR (ulong) | |
1763 | .ad | |
1764 | .RS 12n | |
1765 | Used to control the frequency of multihost writes which are performed when the | |
1766 | \fBmultihost\fR pool property is on. This is one factor used to determine | |
1767 | the length of the activity check during import. | |
1768 | .sp | |
1769 | The multihost write period is \fBzfs_multihost_interval / leaf-vdevs\fR milliseconds. | |
1770 | This means that on average a multihost write will be issued for each leaf vdev every | |
1771 | \fBzfs_multihost_interval\fR milliseconds. In practice, the observed period can | |
1772 | vary with the I/O load and this observed value is the delay which is stored in | |
1773 | the uberblock. | |
1774 | .sp | |
1775 | On import the activity check waits a minimum amount of time determined by | |
1776 | \fBzfs_multihost_interval * zfs_multihost_import_intervals\fR. The activity | |
1777 | check time may be further extended if the value of mmp delay found in the best | |
1778 | uberblock indicates actual multihost updates happened at longer intervals than | |
1779 | \fBzfs_multihost_interval\fR. A minimum value of \fB100ms\fR is enforced. | |
1780 | .sp | |
1781 | Default value: \fB1000\fR. | |
1782 | .RE | |
1783 | ||
1784 | .sp | |
1785 | .ne 2 | |
1786 | .na | |
1787 | \fBzfs_multihost_import_intervals\fR (uint) | |
1788 | .ad | |
1789 | .RS 12n | |
1790 | Used to control the duration of the activity test on import. Smaller values of | |
1791 | \fBzfs_multihost_import_intervals\fR will reduce the import time but increase | |
1792 | the risk of failing to detect an active pool. The total activity check time is | |
1793 | never allowed to drop below one second. A value of 0 is ignored and treated as | |
1794 | if it was set to 1 | |
1795 | .sp | |
1796 | Default value: \fB10\fR. | |
1797 | .RE | |
1798 | ||
1799 | .sp | |
1800 | .ne 2 | |
1801 | .na | |
1802 | \fBzfs_multihost_fail_intervals\fR (uint) | |
1803 | .ad | |
1804 | .RS 12n | |
1805 | Controls the behavior of the pool when multihost write failures are detected. | |
1806 | .sp | |
1807 | When \fBzfs_multihost_fail_intervals = 0\fR then multihost write failures are ignored. | |
1808 | The failures will still be reported to the ZED which depending on its | |
1809 | configuration may take action such as suspending the pool or offlining a device. | |
1810 | .sp | |
1811 | When \fBzfs_multihost_fail_intervals > 0\fR then sequential multihost write failures | |
1812 | will cause the pool to be suspended. This occurs when | |
1813 | \fBzfs_multihost_fail_intervals * zfs_multihost_interval\fR milliseconds have | |
1814 | passed since the last successful multihost write. This guarantees the activity test | |
1815 | will see multihost writes if the pool is imported. | |
1816 | .sp | |
1817 | Default value: \fB5\fR. | |
1818 | .RE | |
1819 | ||
29714574 TF |
1820 | .sp |
1821 | .ne 2 | |
1822 | .na | |
1823 | \fBzfs_no_scrub_io\fR (int) | |
1824 | .ad | |
1825 | .RS 12n | |
83426735 D |
1826 | Set for no scrub I/O. This results in scrubs not actually scrubbing data and |
1827 | simply doing a metadata crawl of the pool instead. | |
29714574 TF |
1828 | .sp |
1829 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1830 | .RE | |
1831 | ||
1832 | .sp | |
1833 | .ne 2 | |
1834 | .na | |
1835 | \fBzfs_no_scrub_prefetch\fR (int) | |
1836 | .ad | |
1837 | .RS 12n | |
83426735 | 1838 | Set to disable block prefetching for scrubs. |
29714574 TF |
1839 | .sp |
1840 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1841 | .RE | |
1842 | ||
29714574 TF |
1843 | .sp |
1844 | .ne 2 | |
1845 | .na | |
1846 | \fBzfs_nocacheflush\fR (int) | |
1847 | .ad | |
1848 | .RS 12n | |
83426735 D |
1849 | Disable cache flush operations on disks when writing. Beware, this may cause |
1850 | corruption if disks re-order writes. | |
29714574 TF |
1851 | .sp |
1852 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1853 | .RE | |
1854 | ||
1855 | .sp | |
1856 | .ne 2 | |
1857 | .na | |
1858 | \fBzfs_nopwrite_enabled\fR (int) | |
1859 | .ad | |
1860 | .RS 12n | |
1861 | Enable NOP writes | |
1862 | .sp | |
1863 | Use \fB1\fR for yes (default) and \fB0\fR to disable. | |
1864 | .RE | |
1865 | ||
66aca247 DB |
1866 | .sp |
1867 | .ne 2 | |
1868 | .na | |
1869 | \fBzfs_dmu_offset_next_sync\fR (int) | |
1870 | .ad | |
1871 | .RS 12n | |
1872 | Enable forcing txg sync to find holes. When enabled forces ZFS to act | |
1873 | like prior versions when SEEK_HOLE or SEEK_DATA flags are used, which | |
1874 | when a dnode is dirty causes txg's to be synced so that this data can be | |
1875 | found. | |
1876 | .sp | |
1877 | Use \fB1\fR for yes and \fB0\fR to disable (default). | |
1878 | .RE | |
1879 | ||
29714574 TF |
1880 | .sp |
1881 | .ne 2 | |
1882 | .na | |
b738bc5a | 1883 | \fBzfs_pd_bytes_max\fR (int) |
29714574 TF |
1884 | .ad |
1885 | .RS 12n | |
83426735 | 1886 | The number of bytes which should be prefetched during a pool traversal |
6146e17e | 1887 | (eg: \fBzfs send\fR or other data crawling operations) |
29714574 | 1888 | .sp |
74aa2ba2 | 1889 | Default value: \fB52,428,800\fR. |
29714574 TF |
1890 | .RE |
1891 | ||
bef78122 DQ |
1892 | .sp |
1893 | .ne 2 | |
1894 | .na | |
1895 | \fBzfs_per_txg_dirty_frees_percent \fR (ulong) | |
1896 | .ad | |
1897 | .RS 12n | |
1898 | Tunable to control percentage of dirtied blocks from frees in one TXG. | |
1899 | After this threshold is crossed, additional dirty blocks from frees | |
1900 | wait until the next TXG. | |
1901 | A value of zero will disable this throttle. | |
1902 | .sp | |
1903 | Default value: \fB30\fR and \fB0\fR to disable. | |
1904 | .RE | |
1905 | ||
1906 | ||
1907 | ||
29714574 TF |
1908 | .sp |
1909 | .ne 2 | |
1910 | .na | |
1911 | \fBzfs_prefetch_disable\fR (int) | |
1912 | .ad | |
1913 | .RS 12n | |
7f60329a MA |
1914 | This tunable disables predictive prefetch. Note that it leaves "prescient" |
1915 | prefetch (e.g. prefetch for zfs send) intact. Unlike predictive prefetch, | |
1916 | prescient prefetch never issues i/os that end up not being needed, so it | |
1917 | can't hurt performance. | |
29714574 TF |
1918 | .sp |
1919 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1920 | .RE | |
1921 | ||
1922 | .sp | |
1923 | .ne 2 | |
1924 | .na | |
1925 | \fBzfs_read_chunk_size\fR (long) | |
1926 | .ad | |
1927 | .RS 12n | |
1928 | Bytes to read per chunk | |
1929 | .sp | |
1930 | Default value: \fB1,048,576\fR. | |
1931 | .RE | |
1932 | ||
1933 | .sp | |
1934 | .ne 2 | |
1935 | .na | |
1936 | \fBzfs_read_history\fR (int) | |
1937 | .ad | |
1938 | .RS 12n | |
379ca9cf OF |
1939 | Historical statistics for the last N reads will be available in |
1940 | \fB/proc/spl/kstat/zfs/<pool>/reads\fR | |
29714574 | 1941 | .sp |
83426735 | 1942 | Default value: \fB0\fR (no data is kept). |
29714574 TF |
1943 | .RE |
1944 | ||
1945 | .sp | |
1946 | .ne 2 | |
1947 | .na | |
1948 | \fBzfs_read_history_hits\fR (int) | |
1949 | .ad | |
1950 | .RS 12n | |
1951 | Include cache hits in read history | |
1952 | .sp | |
1953 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1954 | .RE | |
1955 | ||
9e052db4 MA |
1956 | .sp |
1957 | .ne 2 | |
1958 | .na | |
4589f3ae BB |
1959 | \fBzfs_reconstruct_indirect_combinations_max\fR (int) |
1960 | .ad | |
1961 | .RS 12na | |
1962 | If an indirect split block contains more than this many possible unique | |
1963 | combinations when being reconstructed, consider it too computationally | |
1964 | expensive to check them all. Instead, try at most | |
1965 | \fBzfs_reconstruct_indirect_combinations_max\fR randomly-selected | |
1966 | combinations each time the block is accessed. This allows all segment | |
1967 | copies to participate fairly in the reconstruction when all combinations | |
1968 | cannot be checked and prevents repeated use of one bad copy. | |
1969 | .sp | |
1258bd77 | 1970 | Default value: \fB256\fR. |
9e052db4 MA |
1971 | .RE |
1972 | ||
29714574 TF |
1973 | .sp |
1974 | .ne 2 | |
1975 | .na | |
1976 | \fBzfs_recover\fR (int) | |
1977 | .ad | |
1978 | .RS 12n | |
1979 | Set to attempt to recover from fatal errors. This should only be used as a | |
1980 | last resort, as it typically results in leaked space, or worse. | |
1981 | .sp | |
1982 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1983 | .RE | |
1984 | ||
1985 | .sp | |
1986 | .ne 2 | |
1987 | .na | |
d4a72f23 | 1988 | \fBzfs_resilver_min_time_ms\fR (int) |
29714574 TF |
1989 | .ad |
1990 | .RS 12n | |
d4a72f23 TC |
1991 | Resilvers are processed by the sync thread. While resilvering it will spend |
1992 | at least this much time working on a resilver between txg flushes. | |
29714574 | 1993 | .sp |
d4a72f23 | 1994 | Default value: \fB3,000\fR. |
29714574 TF |
1995 | .RE |
1996 | ||
02638a30 TC |
1997 | .sp |
1998 | .ne 2 | |
1999 | .na | |
2000 | \fBzfs_scan_ignore_errors\fR (int) | |
2001 | .ad | |
2002 | .RS 12n | |
2003 | If set to a nonzero value, remove the DTL (dirty time list) upon | |
2004 | completion of a pool scan (scrub) even if there were unrepairable | |
2005 | errors. It is intended to be used during pool repair or recovery to | |
2006 | stop resilvering when the pool is next imported. | |
2007 | .sp | |
2008 | Default value: \fB0\fR. | |
2009 | .RE | |
2010 | ||
29714574 TF |
2011 | .sp |
2012 | .ne 2 | |
2013 | .na | |
d4a72f23 | 2014 | \fBzfs_scrub_min_time_ms\fR (int) |
29714574 TF |
2015 | .ad |
2016 | .RS 12n | |
d4a72f23 TC |
2017 | Scrubs are processed by the sync thread. While scrubbing it will spend |
2018 | at least this much time working on a scrub between txg flushes. | |
29714574 | 2019 | .sp |
d4a72f23 | 2020 | Default value: \fB1,000\fR. |
29714574 TF |
2021 | .RE |
2022 | ||
2023 | .sp | |
2024 | .ne 2 | |
2025 | .na | |
d4a72f23 | 2026 | \fBzfs_scan_checkpoint_intval\fR (int) |
29714574 TF |
2027 | .ad |
2028 | .RS 12n | |
d4a72f23 TC |
2029 | To preserve progress across reboots the sequential scan algorithm periodically |
2030 | needs to stop metadata scanning and issue all the verifications I/Os to disk. | |
2031 | The frequency of this flushing is determined by the | |
a8577bdb | 2032 | \fBzfs_scan_checkpoint_intval\fR tunable. |
29714574 | 2033 | .sp |
d4a72f23 | 2034 | Default value: \fB7200\fR seconds (every 2 hours). |
29714574 TF |
2035 | .RE |
2036 | ||
2037 | .sp | |
2038 | .ne 2 | |
2039 | .na | |
d4a72f23 | 2040 | \fBzfs_scan_fill_weight\fR (int) |
29714574 TF |
2041 | .ad |
2042 | .RS 12n | |
d4a72f23 TC |
2043 | This tunable affects how scrub and resilver I/O segments are ordered. A higher |
2044 | number indicates that we care more about how filled in a segment is, while a | |
2045 | lower number indicates we care more about the size of the extent without | |
2046 | considering the gaps within a segment. This value is only tunable upon module | |
2047 | insertion. Changing the value afterwards will have no affect on scrub or | |
2048 | resilver performance. | |
29714574 | 2049 | .sp |
d4a72f23 | 2050 | Default value: \fB3\fR. |
29714574 TF |
2051 | .RE |
2052 | ||
2053 | .sp | |
2054 | .ne 2 | |
2055 | .na | |
d4a72f23 | 2056 | \fBzfs_scan_issue_strategy\fR (int) |
29714574 TF |
2057 | .ad |
2058 | .RS 12n | |
d4a72f23 TC |
2059 | Determines the order that data will be verified while scrubbing or resilvering. |
2060 | If set to \fB1\fR, data will be verified as sequentially as possible, given the | |
2061 | amount of memory reserved for scrubbing (see \fBzfs_scan_mem_lim_fact\fR). This | |
2062 | may improve scrub performance if the pool's data is very fragmented. If set to | |
2063 | \fB2\fR, the largest mostly-contiguous chunk of found data will be verified | |
2064 | first. By deferring scrubbing of small segments, we may later find adjacent data | |
2065 | to coalesce and increase the segment size. If set to \fB0\fR, zfs will use | |
2066 | strategy \fB1\fR during normal verification and strategy \fB2\fR while taking a | |
2067 | checkpoint. | |
29714574 | 2068 | .sp |
d4a72f23 TC |
2069 | Default value: \fB0\fR. |
2070 | .RE | |
2071 | ||
2072 | .sp | |
2073 | .ne 2 | |
2074 | .na | |
2075 | \fBzfs_scan_legacy\fR (int) | |
2076 | .ad | |
2077 | .RS 12n | |
2078 | A value of 0 indicates that scrubs and resilvers will gather metadata in | |
2079 | memory before issuing sequential I/O. A value of 1 indicates that the legacy | |
2080 | algorithm will be used where I/O is initiated as soon as it is discovered. | |
2081 | Changing this value to 0 will not affect scrubs or resilvers that are already | |
2082 | in progress. | |
2083 | .sp | |
2084 | Default value: \fB0\fR. | |
2085 | .RE | |
2086 | ||
2087 | .sp | |
2088 | .ne 2 | |
2089 | .na | |
2090 | \fBzfs_scan_max_ext_gap\fR (int) | |
2091 | .ad | |
2092 | .RS 12n | |
2093 | Indicates the largest gap in bytes between scrub / resilver I/Os that will still | |
2094 | be considered sequential for sorting purposes. Changing this value will not | |
2095 | affect scrubs or resilvers that are already in progress. | |
2096 | .sp | |
2097 | Default value: \fB2097152 (2 MB)\fR. | |
2098 | .RE | |
2099 | ||
2100 | .sp | |
2101 | .ne 2 | |
2102 | .na | |
2103 | \fBzfs_scan_mem_lim_fact\fR (int) | |
2104 | .ad | |
2105 | .RS 12n | |
2106 | Maximum fraction of RAM used for I/O sorting by sequential scan algorithm. | |
2107 | This tunable determines the hard limit for I/O sorting memory usage. | |
2108 | When the hard limit is reached we stop scanning metadata and start issuing | |
2109 | data verification I/O. This is done until we get below the soft limit. | |
2110 | .sp | |
2111 | Default value: \fB20\fR which is 5% of RAM (1/20). | |
2112 | .RE | |
2113 | ||
2114 | .sp | |
2115 | .ne 2 | |
2116 | .na | |
2117 | \fBzfs_scan_mem_lim_soft_fact\fR (int) | |
2118 | .ad | |
2119 | .RS 12n | |
2120 | The fraction of the hard limit used to determined the soft limit for I/O sorting | |
2121 | by the sequential scan algorithm. When we cross this limit from bellow no action | |
2122 | is taken. When we cross this limit from above it is because we are issuing | |
2123 | verification I/O. In this case (unless the metadata scan is done) we stop | |
2124 | issuing verification I/O and start scanning metadata again until we get to the | |
2125 | hard limit. | |
2126 | .sp | |
2127 | Default value: \fB20\fR which is 5% of the hard limit (1/20). | |
2128 | .RE | |
2129 | ||
2130 | .sp | |
2131 | .ne 2 | |
2132 | .na | |
2133 | \fBzfs_scan_vdev_limit\fR (int) | |
2134 | .ad | |
2135 | .RS 12n | |
2136 | Maximum amount of data that can be concurrently issued at once for scrubs and | |
2137 | resilvers per leaf device, given in bytes. | |
2138 | .sp | |
2139 | Default value: \fB41943040\fR. | |
29714574 TF |
2140 | .RE |
2141 | ||
fd8febbd TF |
2142 | .sp |
2143 | .ne 2 | |
2144 | .na | |
2145 | \fBzfs_send_corrupt_data\fR (int) | |
2146 | .ad | |
2147 | .RS 12n | |
83426735 | 2148 | Allow sending of corrupt data (ignore read/checksum errors when sending data) |
fd8febbd TF |
2149 | .sp |
2150 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2151 | .RE | |
2152 | ||
3b0d9928 BB |
2153 | .sp |
2154 | .ne 2 | |
2155 | .na | |
2156 | \fBzfs_send_queue_length\fR (int) | |
2157 | .ad | |
2158 | .RS 12n | |
2159 | The maximum number of bytes allowed in the \fBzfs send\fR queue. This value | |
2160 | must be at least twice the maximum block size in use. | |
2161 | .sp | |
2162 | Default value: \fB16,777,216\fR. | |
2163 | .RE | |
2164 | ||
2165 | .sp | |
2166 | .ne 2 | |
2167 | .na | |
2168 | \fBzfs_recv_queue_length\fR (int) | |
2169 | .ad | |
2170 | .RS 12n | |
2171 | .sp | |
2172 | The maximum number of bytes allowed in the \fBzfs receive\fR queue. This value | |
2173 | must be at least twice the maximum block size in use. | |
2174 | .sp | |
2175 | Default value: \fB16,777,216\fR. | |
2176 | .RE | |
2177 | ||
29714574 TF |
2178 | .sp |
2179 | .ne 2 | |
2180 | .na | |
2181 | \fBzfs_sync_pass_deferred_free\fR (int) | |
2182 | .ad | |
2183 | .RS 12n | |
83426735 | 2184 | Flushing of data to disk is done in passes. Defer frees starting in this pass |
29714574 TF |
2185 | .sp |
2186 | Default value: \fB2\fR. | |
2187 | .RE | |
2188 | ||
d2734cce SD |
2189 | .sp |
2190 | .ne 2 | |
2191 | .na | |
2192 | \fBzfs_spa_discard_memory_limit\fR (int) | |
2193 | .ad | |
2194 | .RS 12n | |
2195 | Maximum memory used for prefetching a checkpoint's space map on each | |
2196 | vdev while discarding the checkpoint. | |
2197 | .sp | |
2198 | Default value: \fB16,777,216\fR. | |
2199 | .RE | |
2200 | ||
29714574 TF |
2201 | .sp |
2202 | .ne 2 | |
2203 | .na | |
2204 | \fBzfs_sync_pass_dont_compress\fR (int) | |
2205 | .ad | |
2206 | .RS 12n | |
2207 | Don't compress starting in this pass | |
2208 | .sp | |
2209 | Default value: \fB5\fR. | |
2210 | .RE | |
2211 | ||
2212 | .sp | |
2213 | .ne 2 | |
2214 | .na | |
2215 | \fBzfs_sync_pass_rewrite\fR (int) | |
2216 | .ad | |
2217 | .RS 12n | |
83426735 | 2218 | Rewrite new block pointers starting in this pass |
29714574 TF |
2219 | .sp |
2220 | Default value: \fB2\fR. | |
2221 | .RE | |
2222 | ||
a032ac4b BB |
2223 | .sp |
2224 | .ne 2 | |
2225 | .na | |
2226 | \fBzfs_sync_taskq_batch_pct\fR (int) | |
2227 | .ad | |
2228 | .RS 12n | |
2229 | This controls the number of threads used by the dp_sync_taskq. The default | |
2230 | value of 75% will create a maximum of one thread per cpu. | |
2231 | .sp | |
be54a13c | 2232 | Default value: \fB75\fR%. |
a032ac4b BB |
2233 | .RE |
2234 | ||
29714574 TF |
2235 | .sp |
2236 | .ne 2 | |
2237 | .na | |
2238 | \fBzfs_txg_history\fR (int) | |
2239 | .ad | |
2240 | .RS 12n | |
379ca9cf OF |
2241 | Historical statistics for the last N txgs will be available in |
2242 | \fB/proc/spl/kstat/zfs/<pool>/txgs\fR | |
29714574 | 2243 | .sp |
ca85d690 | 2244 | Default value: \fB0\fR. |
29714574 TF |
2245 | .RE |
2246 | ||
29714574 TF |
2247 | .sp |
2248 | .ne 2 | |
2249 | .na | |
2250 | \fBzfs_txg_timeout\fR (int) | |
2251 | .ad | |
2252 | .RS 12n | |
83426735 | 2253 | Flush dirty data to disk at least every N seconds (maximum txg duration) |
29714574 TF |
2254 | .sp |
2255 | Default value: \fB5\fR. | |
2256 | .RE | |
2257 | ||
2258 | .sp | |
2259 | .ne 2 | |
2260 | .na | |
2261 | \fBzfs_vdev_aggregation_limit\fR (int) | |
2262 | .ad | |
2263 | .RS 12n | |
2264 | Max vdev I/O aggregation size | |
2265 | .sp | |
2266 | Default value: \fB131,072\fR. | |
2267 | .RE | |
2268 | ||
2269 | .sp | |
2270 | .ne 2 | |
2271 | .na | |
2272 | \fBzfs_vdev_cache_bshift\fR (int) | |
2273 | .ad | |
2274 | .RS 12n | |
2275 | Shift size to inflate reads too | |
2276 | .sp | |
83426735 | 2277 | Default value: \fB16\fR (effectively 65536). |
29714574 TF |
2278 | .RE |
2279 | ||
2280 | .sp | |
2281 | .ne 2 | |
2282 | .na | |
2283 | \fBzfs_vdev_cache_max\fR (int) | |
2284 | .ad | |
2285 | .RS 12n | |
ca85d690 | 2286 | Inflate reads smaller than this value to meet the \fBzfs_vdev_cache_bshift\fR |
2287 | size (default 64k). | |
83426735 D |
2288 | .sp |
2289 | Default value: \fB16384\fR. | |
29714574 TF |
2290 | .RE |
2291 | ||
2292 | .sp | |
2293 | .ne 2 | |
2294 | .na | |
2295 | \fBzfs_vdev_cache_size\fR (int) | |
2296 | .ad | |
2297 | .RS 12n | |
83426735 D |
2298 | Total size of the per-disk cache in bytes. |
2299 | .sp | |
2300 | Currently this feature is disabled as it has been found to not be helpful | |
2301 | for performance and in some cases harmful. | |
29714574 TF |
2302 | .sp |
2303 | Default value: \fB0\fR. | |
2304 | .RE | |
2305 | ||
29714574 TF |
2306 | .sp |
2307 | .ne 2 | |
2308 | .na | |
9f500936 | 2309 | \fBzfs_vdev_mirror_rotating_inc\fR (int) |
29714574 TF |
2310 | .ad |
2311 | .RS 12n | |
9f500936 | 2312 | A number by which the balancing algorithm increments the load calculation for |
2313 | the purpose of selecting the least busy mirror member when an I/O immediately | |
2314 | follows its predecessor on rotational vdevs for the purpose of making decisions | |
2315 | based on load. | |
29714574 | 2316 | .sp |
9f500936 | 2317 | Default value: \fB0\fR. |
2318 | .RE | |
2319 | ||
2320 | .sp | |
2321 | .ne 2 | |
2322 | .na | |
2323 | \fBzfs_vdev_mirror_rotating_seek_inc\fR (int) | |
2324 | .ad | |
2325 | .RS 12n | |
2326 | A number by which the balancing algorithm increments the load calculation for | |
2327 | the purpose of selecting the least busy mirror member when an I/O lacks | |
2328 | locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within | |
2329 | this that are not immediately following the previous I/O are incremented by | |
2330 | half. | |
2331 | .sp | |
2332 | Default value: \fB5\fR. | |
2333 | .RE | |
2334 | ||
2335 | .sp | |
2336 | .ne 2 | |
2337 | .na | |
2338 | \fBzfs_vdev_mirror_rotating_seek_offset\fR (int) | |
2339 | .ad | |
2340 | .RS 12n | |
2341 | The maximum distance for the last queued I/O in which the balancing algorithm | |
2342 | considers an I/O to have locality. | |
2343 | See the section "ZFS I/O SCHEDULER". | |
2344 | .sp | |
2345 | Default value: \fB1048576\fR. | |
2346 | .RE | |
2347 | ||
2348 | .sp | |
2349 | .ne 2 | |
2350 | .na | |
2351 | \fBzfs_vdev_mirror_non_rotating_inc\fR (int) | |
2352 | .ad | |
2353 | .RS 12n | |
2354 | A number by which the balancing algorithm increments the load calculation for | |
2355 | the purpose of selecting the least busy mirror member on non-rotational vdevs | |
2356 | when I/Os do not immediately follow one another. | |
2357 | .sp | |
2358 | Default value: \fB0\fR. | |
2359 | .RE | |
2360 | ||
2361 | .sp | |
2362 | .ne 2 | |
2363 | .na | |
2364 | \fBzfs_vdev_mirror_non_rotating_seek_inc\fR (int) | |
2365 | .ad | |
2366 | .RS 12n | |
2367 | A number by which the balancing algorithm increments the load calculation for | |
2368 | the purpose of selecting the least busy mirror member when an I/O lacks | |
2369 | locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within | |
2370 | this that are not immediately following the previous I/O are incremented by | |
2371 | half. | |
2372 | .sp | |
2373 | Default value: \fB1\fR. | |
29714574 TF |
2374 | .RE |
2375 | ||
29714574 TF |
2376 | .sp |
2377 | .ne 2 | |
2378 | .na | |
2379 | \fBzfs_vdev_read_gap_limit\fR (int) | |
2380 | .ad | |
2381 | .RS 12n | |
83426735 D |
2382 | Aggregate read I/O operations if the gap on-disk between them is within this |
2383 | threshold. | |
29714574 TF |
2384 | .sp |
2385 | Default value: \fB32,768\fR. | |
2386 | .RE | |
2387 | ||
2388 | .sp | |
2389 | .ne 2 | |
2390 | .na | |
2391 | \fBzfs_vdev_scheduler\fR (charp) | |
2392 | .ad | |
2393 | .RS 12n | |
ca85d690 | 2394 | Set the Linux I/O scheduler on whole disk vdevs to this scheduler. Valid options |
2395 | are noop, cfq, bfq & deadline | |
29714574 TF |
2396 | .sp |
2397 | Default value: \fBnoop\fR. | |
2398 | .RE | |
2399 | ||
29714574 TF |
2400 | .sp |
2401 | .ne 2 | |
2402 | .na | |
2403 | \fBzfs_vdev_write_gap_limit\fR (int) | |
2404 | .ad | |
2405 | .RS 12n | |
2406 | Aggregate write I/O over gap | |
2407 | .sp | |
2408 | Default value: \fB4,096\fR. | |
2409 | .RE | |
2410 | ||
ab9f4b0b GN |
2411 | .sp |
2412 | .ne 2 | |
2413 | .na | |
2414 | \fBzfs_vdev_raidz_impl\fR (string) | |
2415 | .ad | |
2416 | .RS 12n | |
c9187d86 | 2417 | Parameter for selecting raidz parity implementation to use. |
ab9f4b0b GN |
2418 | |
2419 | Options marked (always) below may be selected on module load as they are | |
2420 | supported on all systems. | |
2421 | The remaining options may only be set after the module is loaded, as they | |
2422 | are available only if the implementations are compiled in and supported | |
2423 | on the running system. | |
2424 | ||
2425 | Once the module is loaded, the content of | |
2426 | /sys/module/zfs/parameters/zfs_vdev_raidz_impl will show available options | |
2427 | with the currently selected one enclosed in []. | |
2428 | Possible options are: | |
2429 | fastest - (always) implementation selected using built-in benchmark | |
2430 | original - (always) original raidz implementation | |
2431 | scalar - (always) scalar raidz implementation | |
ae25d222 GN |
2432 | sse2 - implementation using SSE2 instruction set (64bit x86 only) |
2433 | ssse3 - implementation using SSSE3 instruction set (64bit x86 only) | |
ab9f4b0b | 2434 | avx2 - implementation using AVX2 instruction set (64bit x86 only) |
7f547f85 RD |
2435 | avx512f - implementation using AVX512F instruction set (64bit x86 only) |
2436 | avx512bw - implementation using AVX512F & AVX512BW instruction sets (64bit x86 only) | |
62a65a65 RD |
2437 | aarch64_neon - implementation using NEON (Aarch64/64 bit ARMv8 only) |
2438 | aarch64_neonx2 - implementation using NEON with more unrolling (Aarch64/64 bit ARMv8 only) | |
ab9f4b0b GN |
2439 | .sp |
2440 | Default value: \fBfastest\fR. | |
2441 | .RE | |
2442 | ||
29714574 TF |
2443 | .sp |
2444 | .ne 2 | |
2445 | .na | |
2446 | \fBzfs_zevent_cols\fR (int) | |
2447 | .ad | |
2448 | .RS 12n | |
83426735 | 2449 | When zevents are logged to the console use this as the word wrap width. |
29714574 TF |
2450 | .sp |
2451 | Default value: \fB80\fR. | |
2452 | .RE | |
2453 | ||
2454 | .sp | |
2455 | .ne 2 | |
2456 | .na | |
2457 | \fBzfs_zevent_console\fR (int) | |
2458 | .ad | |
2459 | .RS 12n | |
2460 | Log events to the console | |
2461 | .sp | |
2462 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2463 | .RE | |
2464 | ||
2465 | .sp | |
2466 | .ne 2 | |
2467 | .na | |
2468 | \fBzfs_zevent_len_max\fR (int) | |
2469 | .ad | |
2470 | .RS 12n | |
83426735 D |
2471 | Max event queue length. A value of 0 will result in a calculated value which |
2472 | increases with the number of CPUs in the system (minimum 64 events). Events | |
2473 | in the queue can be viewed with the \fBzpool events\fR command. | |
29714574 TF |
2474 | .sp |
2475 | Default value: \fB0\fR. | |
2476 | .RE | |
2477 | ||
a032ac4b BB |
2478 | .sp |
2479 | .ne 2 | |
2480 | .na | |
2481 | \fBzfs_zil_clean_taskq_maxalloc\fR (int) | |
2482 | .ad | |
2483 | .RS 12n | |
2484 | The maximum number of taskq entries that are allowed to be cached. When this | |
2fe61a7e | 2485 | limit is exceeded transaction records (itxs) will be cleaned synchronously. |
a032ac4b BB |
2486 | .sp |
2487 | Default value: \fB1048576\fR. | |
2488 | .RE | |
2489 | ||
2490 | .sp | |
2491 | .ne 2 | |
2492 | .na | |
2493 | \fBzfs_zil_clean_taskq_minalloc\fR (int) | |
2494 | .ad | |
2495 | .RS 12n | |
2496 | The number of taskq entries that are pre-populated when the taskq is first | |
2497 | created and are immediately available for use. | |
2498 | .sp | |
2499 | Default value: \fB1024\fR. | |
2500 | .RE | |
2501 | ||
2502 | .sp | |
2503 | .ne 2 | |
2504 | .na | |
2505 | \fBzfs_zil_clean_taskq_nthr_pct\fR (int) | |
2506 | .ad | |
2507 | .RS 12n | |
2508 | This controls the number of threads used by the dp_zil_clean_taskq. The default | |
2509 | value of 100% will create a maximum of one thread per cpu. | |
2510 | .sp | |
be54a13c | 2511 | Default value: \fB100\fR%. |
a032ac4b BB |
2512 | .RE |
2513 | ||
29714574 TF |
2514 | .sp |
2515 | .ne 2 | |
2516 | .na | |
2517 | \fBzil_replay_disable\fR (int) | |
2518 | .ad | |
2519 | .RS 12n | |
83426735 D |
2520 | Disable intent logging replay. Can be disabled for recovery from corrupted |
2521 | ZIL | |
29714574 TF |
2522 | .sp |
2523 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2524 | .RE | |
2525 | ||
2526 | .sp | |
2527 | .ne 2 | |
2528 | .na | |
1b7c1e5c | 2529 | \fBzil_slog_bulk\fR (ulong) |
29714574 TF |
2530 | .ad |
2531 | .RS 12n | |
1b7c1e5c GDN |
2532 | Limit SLOG write size per commit executed with synchronous priority. |
2533 | Any writes above that will be executed with lower (asynchronous) priority | |
2534 | to limit potential SLOG device abuse by single active ZIL writer. | |
29714574 | 2535 | .sp |
1b7c1e5c | 2536 | Default value: \fB786,432\fR. |
29714574 TF |
2537 | .RE |
2538 | ||
c3bd3fb4 TC |
2539 | .sp |
2540 | .ne 2 | |
2541 | .na | |
2542 | \fBzio_decompress_fail_fraction\fR (int) | |
2543 | .ad | |
2544 | .RS 12n | |
2545 | If non-zero, this value represents the denominator of the probability that zfs | |
2546 | should induce a decompression failure. For instance, for a 5% decompression | |
2547 | failure rate, this value should be set to 20. | |
2548 | .sp | |
2549 | Default value: \fB0\fR. | |
2550 | .RE | |
2551 | ||
29714574 TF |
2552 | .sp |
2553 | .ne 2 | |
2554 | .na | |
2555 | \fBzio_delay_max\fR (int) | |
2556 | .ad | |
2557 | .RS 12n | |
83426735 | 2558 | A zevent will be logged if a ZIO operation takes more than N milliseconds to |
ab9f4b0b | 2559 | complete. Note that this is only a logging facility, not a timeout on |
83426735 | 2560 | operations. |
29714574 TF |
2561 | .sp |
2562 | Default value: \fB30,000\fR. | |
2563 | .RE | |
2564 | ||
3dfb57a3 DB |
2565 | .sp |
2566 | .ne 2 | |
2567 | .na | |
2568 | \fBzio_dva_throttle_enabled\fR (int) | |
2569 | .ad | |
2570 | .RS 12n | |
2571 | Throttle block allocations in the ZIO pipeline. This allows for | |
2572 | dynamic allocation distribution when devices are imbalanced. | |
e815485f TC |
2573 | When enabled, the maximum number of pending allocations per top-level vdev |
2574 | is limited by \fBzfs_vdev_queue_depth_pct\fR. | |
3dfb57a3 | 2575 | .sp |
27f2b90d | 2576 | Default value: \fB1\fR. |
3dfb57a3 DB |
2577 | .RE |
2578 | ||
29714574 TF |
2579 | .sp |
2580 | .ne 2 | |
2581 | .na | |
2582 | \fBzio_requeue_io_start_cut_in_line\fR (int) | |
2583 | .ad | |
2584 | .RS 12n | |
2585 | Prioritize requeued I/O | |
2586 | .sp | |
2587 | Default value: \fB0\fR. | |
2588 | .RE | |
2589 | ||
dcb6bed1 D |
2590 | .sp |
2591 | .ne 2 | |
2592 | .na | |
2593 | \fBzio_taskq_batch_pct\fR (uint) | |
2594 | .ad | |
2595 | .RS 12n | |
2596 | Percentage of online CPUs (or CPU cores, etc) which will run a worker thread | |
2597 | for IO. These workers are responsible for IO work such as compression and | |
2598 | checksum calculations. Fractional number of CPUs will be rounded down. | |
2599 | .sp | |
2600 | The default value of 75 was chosen to avoid using all CPUs which can result in | |
2601 | latency issues and inconsistent application performance, especially when high | |
2602 | compression is enabled. | |
2603 | .sp | |
2604 | Default value: \fB75\fR. | |
2605 | .RE | |
2606 | ||
29714574 TF |
2607 | .sp |
2608 | .ne 2 | |
2609 | .na | |
2610 | \fBzvol_inhibit_dev\fR (uint) | |
2611 | .ad | |
2612 | .RS 12n | |
83426735 D |
2613 | Do not create zvol device nodes. This may slightly improve startup time on |
2614 | systems with a very large number of zvols. | |
29714574 TF |
2615 | .sp |
2616 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2617 | .RE | |
2618 | ||
2619 | .sp | |
2620 | .ne 2 | |
2621 | .na | |
2622 | \fBzvol_major\fR (uint) | |
2623 | .ad | |
2624 | .RS 12n | |
83426735 | 2625 | Major number for zvol block devices |
29714574 TF |
2626 | .sp |
2627 | Default value: \fB230\fR. | |
2628 | .RE | |
2629 | ||
2630 | .sp | |
2631 | .ne 2 | |
2632 | .na | |
2633 | \fBzvol_max_discard_blocks\fR (ulong) | |
2634 | .ad | |
2635 | .RS 12n | |
83426735 D |
2636 | Discard (aka TRIM) operations done on zvols will be done in batches of this |
2637 | many blocks, where block size is determined by the \fBvolblocksize\fR property | |
2638 | of a zvol. | |
29714574 TF |
2639 | .sp |
2640 | Default value: \fB16,384\fR. | |
2641 | .RE | |
2642 | ||
9965059a BB |
2643 | .sp |
2644 | .ne 2 | |
2645 | .na | |
2646 | \fBzvol_prefetch_bytes\fR (uint) | |
2647 | .ad | |
2648 | .RS 12n | |
2649 | When adding a zvol to the system prefetch \fBzvol_prefetch_bytes\fR | |
2650 | from the start and end of the volume. Prefetching these regions | |
2651 | of the volume is desirable because they are likely to be accessed | |
2652 | immediately by \fBblkid(8)\fR or by the kernel scanning for a partition | |
2653 | table. | |
2654 | .sp | |
2655 | Default value: \fB131,072\fR. | |
2656 | .RE | |
2657 | ||
692e55b8 CC |
2658 | .sp |
2659 | .ne 2 | |
2660 | .na | |
2661 | \fBzvol_request_sync\fR (uint) | |
2662 | .ad | |
2663 | .RS 12n | |
2664 | When processing I/O requests for a zvol submit them synchronously. This | |
2665 | effectively limits the queue depth to 1 for each I/O submitter. When set | |
2666 | to 0 requests are handled asynchronously by a thread pool. The number of | |
2667 | requests which can be handled concurrently is controller by \fBzvol_threads\fR. | |
2668 | .sp | |
8fa5250f | 2669 | Default value: \fB0\fR. |
692e55b8 CC |
2670 | .RE |
2671 | ||
2672 | .sp | |
2673 | .ne 2 | |
2674 | .na | |
2675 | \fBzvol_threads\fR (uint) | |
2676 | .ad | |
2677 | .RS 12n | |
2678 | Max number of threads which can handle zvol I/O requests concurrently. | |
2679 | .sp | |
2680 | Default value: \fB32\fR. | |
2681 | .RE | |
2682 | ||
cf8738d8 | 2683 | .sp |
2684 | .ne 2 | |
2685 | .na | |
2686 | \fBzvol_volmode\fR (uint) | |
2687 | .ad | |
2688 | .RS 12n | |
2689 | Defines zvol block devices behaviour when \fBvolmode\fR is set to \fBdefault\fR. | |
2690 | Valid values are \fB1\fR (full), \fB2\fR (dev) and \fB3\fR (none). | |
2691 | .sp | |
2692 | Default value: \fB1\fR. | |
2693 | .RE | |
2694 | ||
39ccc909 | 2695 | .sp |
2696 | .ne 2 | |
2697 | .na | |
2698 | \fBzfs_qat_disable\fR (int) | |
2699 | .ad | |
2700 | .RS 12n | |
cf637391 TC |
2701 | This tunable disables qat hardware acceleration for gzip compression and. |
2702 | AES-GCM encryption. It is available only if qat acceleration is compiled in | |
2703 | and the qat driver is present. | |
39ccc909 | 2704 | .sp |
2705 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2706 | .RE | |
2707 | ||
e8b96c60 MA |
2708 | .SH ZFS I/O SCHEDULER |
2709 | ZFS issues I/O operations to leaf vdevs to satisfy and complete I/Os. | |
2710 | The I/O scheduler determines when and in what order those operations are | |
2711 | issued. The I/O scheduler divides operations into five I/O classes | |
2712 | prioritized in the following order: sync read, sync write, async read, | |
2713 | async write, and scrub/resilver. Each queue defines the minimum and | |
2714 | maximum number of concurrent operations that may be issued to the | |
2715 | device. In addition, the device has an aggregate maximum, | |
2716 | \fBzfs_vdev_max_active\fR. Note that the sum of the per-queue minimums | |
2717 | must not exceed the aggregate maximum. If the sum of the per-queue | |
2718 | maximums exceeds the aggregate maximum, then the number of active I/Os | |
2719 | may reach \fBzfs_vdev_max_active\fR, in which case no further I/Os will | |
2720 | be issued regardless of whether all per-queue minimums have been met. | |
2721 | .sp | |
2722 | For many physical devices, throughput increases with the number of | |
2723 | concurrent operations, but latency typically suffers. Further, physical | |
2724 | devices typically have a limit at which more concurrent operations have no | |
2725 | effect on throughput or can actually cause it to decrease. | |
2726 | .sp | |
2727 | The scheduler selects the next operation to issue by first looking for an | |
2728 | I/O class whose minimum has not been satisfied. Once all are satisfied and | |
2729 | the aggregate maximum has not been hit, the scheduler looks for classes | |
2730 | whose maximum has not been satisfied. Iteration through the I/O classes is | |
2731 | done in the order specified above. No further operations are issued if the | |
2732 | aggregate maximum number of concurrent operations has been hit or if there | |
2733 | are no operations queued for an I/O class that has not hit its maximum. | |
2734 | Every time an I/O is queued or an operation completes, the I/O scheduler | |
2735 | looks for new operations to issue. | |
2736 | .sp | |
2737 | In general, smaller max_active's will lead to lower latency of synchronous | |
2738 | operations. Larger max_active's may lead to higher overall throughput, | |
2739 | depending on underlying storage. | |
2740 | .sp | |
2741 | The ratio of the queues' max_actives determines the balance of performance | |
2742 | between reads, writes, and scrubs. E.g., increasing | |
2743 | \fBzfs_vdev_scrub_max_active\fR will cause the scrub or resilver to complete | |
2744 | more quickly, but reads and writes to have higher latency and lower throughput. | |
2745 | .sp | |
2746 | All I/O classes have a fixed maximum number of outstanding operations | |
2747 | except for the async write class. Asynchronous writes represent the data | |
2748 | that is committed to stable storage during the syncing stage for | |
2749 | transaction groups. Transaction groups enter the syncing state | |
2750 | periodically so the number of queued async writes will quickly burst up | |
2751 | and then bleed down to zero. Rather than servicing them as quickly as | |
2752 | possible, the I/O scheduler changes the maximum number of active async | |
2753 | write I/Os according to the amount of dirty data in the pool. Since | |
2754 | both throughput and latency typically increase with the number of | |
2755 | concurrent operations issued to physical devices, reducing the | |
2756 | burstiness in the number of concurrent operations also stabilizes the | |
2757 | response time of operations from other -- and in particular synchronous | |
2758 | -- queues. In broad strokes, the I/O scheduler will issue more | |
2759 | concurrent operations from the async write queue as there's more dirty | |
2760 | data in the pool. | |
2761 | .sp | |
2762 | Async Writes | |
2763 | .sp | |
2764 | The number of concurrent operations issued for the async write I/O class | |
2765 | follows a piece-wise linear function defined by a few adjustable points. | |
2766 | .nf | |
2767 | ||
2768 | | o---------| <-- zfs_vdev_async_write_max_active | |
2769 | ^ | /^ | | |
2770 | | | / | | | |
2771 | active | / | | | |
2772 | I/O | / | | | |
2773 | count | / | | | |
2774 | | / | | | |
2775 | |-------o | | <-- zfs_vdev_async_write_min_active | |
2776 | 0|_______^______|_________| | |
2777 | 0% | | 100% of zfs_dirty_data_max | |
2778 | | | | |
2779 | | `-- zfs_vdev_async_write_active_max_dirty_percent | |
2780 | `--------- zfs_vdev_async_write_active_min_dirty_percent | |
2781 | ||
2782 | .fi | |
2783 | Until the amount of dirty data exceeds a minimum percentage of the dirty | |
2784 | data allowed in the pool, the I/O scheduler will limit the number of | |
2785 | concurrent operations to the minimum. As that threshold is crossed, the | |
2786 | number of concurrent operations issued increases linearly to the maximum at | |
2787 | the specified maximum percentage of the dirty data allowed in the pool. | |
2788 | .sp | |
2789 | Ideally, the amount of dirty data on a busy pool will stay in the sloped | |
2790 | part of the function between \fBzfs_vdev_async_write_active_min_dirty_percent\fR | |
2791 | and \fBzfs_vdev_async_write_active_max_dirty_percent\fR. If it exceeds the | |
2792 | maximum percentage, this indicates that the rate of incoming data is | |
2793 | greater than the rate that the backend storage can handle. In this case, we | |
2794 | must further throttle incoming writes, as described in the next section. | |
2795 | ||
2796 | .SH ZFS TRANSACTION DELAY | |
2797 | We delay transactions when we've determined that the backend storage | |
2798 | isn't able to accommodate the rate of incoming writes. | |
2799 | .sp | |
2800 | If there is already a transaction waiting, we delay relative to when | |
2801 | that transaction will finish waiting. This way the calculated delay time | |
2802 | is independent of the number of threads concurrently executing | |
2803 | transactions. | |
2804 | .sp | |
2805 | If we are the only waiter, wait relative to when the transaction | |
2806 | started, rather than the current time. This credits the transaction for | |
2807 | "time already served", e.g. reading indirect blocks. | |
2808 | .sp | |
2809 | The minimum time for a transaction to take is calculated as: | |
2810 | .nf | |
2811 | min_time = zfs_delay_scale * (dirty - min) / (max - dirty) | |
2812 | min_time is then capped at 100 milliseconds. | |
2813 | .fi | |
2814 | .sp | |
2815 | The delay has two degrees of freedom that can be adjusted via tunables. The | |
2816 | percentage of dirty data at which we start to delay is defined by | |
2817 | \fBzfs_delay_min_dirty_percent\fR. This should typically be at or above | |
2818 | \fBzfs_vdev_async_write_active_max_dirty_percent\fR so that we only start to | |
2819 | delay after writing at full speed has failed to keep up with the incoming write | |
2820 | rate. The scale of the curve is defined by \fBzfs_delay_scale\fR. Roughly speaking, | |
2821 | this variable determines the amount of delay at the midpoint of the curve. | |
2822 | .sp | |
2823 | .nf | |
2824 | delay | |
2825 | 10ms +-------------------------------------------------------------*+ | |
2826 | | *| | |
2827 | 9ms + *+ | |
2828 | | *| | |
2829 | 8ms + *+ | |
2830 | | * | | |
2831 | 7ms + * + | |
2832 | | * | | |
2833 | 6ms + * + | |
2834 | | * | | |
2835 | 5ms + * + | |
2836 | | * | | |
2837 | 4ms + * + | |
2838 | | * | | |
2839 | 3ms + * + | |
2840 | | * | | |
2841 | 2ms + (midpoint) * + | |
2842 | | | ** | | |
2843 | 1ms + v *** + | |
2844 | | zfs_delay_scale ----------> ******** | | |
2845 | 0 +-------------------------------------*********----------------+ | |
2846 | 0% <- zfs_dirty_data_max -> 100% | |
2847 | .fi | |
2848 | .sp | |
2849 | Note that since the delay is added to the outstanding time remaining on the | |
2850 | most recent transaction, the delay is effectively the inverse of IOPS. | |
2851 | Here the midpoint of 500us translates to 2000 IOPS. The shape of the curve | |
2852 | was chosen such that small changes in the amount of accumulated dirty data | |
2853 | in the first 3/4 of the curve yield relatively small differences in the | |
2854 | amount of delay. | |
2855 | .sp | |
2856 | The effects can be easier to understand when the amount of delay is | |
2857 | represented on a log scale: | |
2858 | .sp | |
2859 | .nf | |
2860 | delay | |
2861 | 100ms +-------------------------------------------------------------++ | |
2862 | + + | |
2863 | | | | |
2864 | + *+ | |
2865 | 10ms + *+ | |
2866 | + ** + | |
2867 | | (midpoint) ** | | |
2868 | + | ** + | |
2869 | 1ms + v **** + | |
2870 | + zfs_delay_scale ----------> ***** + | |
2871 | | **** | | |
2872 | + **** + | |
2873 | 100us + ** + | |
2874 | + * + | |
2875 | | * | | |
2876 | + * + | |
2877 | 10us + * + | |
2878 | + + | |
2879 | | | | |
2880 | + + | |
2881 | +--------------------------------------------------------------+ | |
2882 | 0% <- zfs_dirty_data_max -> 100% | |
2883 | .fi | |
2884 | .sp | |
2885 | Note here that only as the amount of dirty data approaches its limit does | |
2886 | the delay start to increase rapidly. The goal of a properly tuned system | |
2887 | should be to keep the amount of dirty data out of that range by first | |
2888 | ensuring that the appropriate limits are set for the I/O scheduler to reach | |
2889 | optimal throughput on the backend storage, and then by changing the value | |
2890 | of \fBzfs_delay_scale\fR to increase the steepness of the curve. |