]>
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 | ||
44 | .sp | |
45 | .ne 2 | |
46 | .na | |
47 | \fBdbuf_cache_hiwater_pct\fR (uint) | |
48 | .ad | |
49 | .RS 12n | |
50 | The percentage over \fBdbuf_cache_max_bytes\fR when dbufs must be evicted | |
51 | directly. | |
52 | .sp | |
53 | Default value: \fB10\fR%. | |
54 | .RE | |
55 | ||
56 | .sp | |
57 | .ne 2 | |
58 | .na | |
59 | \fBdbuf_cache_lowater_pct\fR (uint) | |
60 | .ad | |
61 | .RS 12n | |
62 | The percentage below \fBdbuf_cache_max_bytes\fR when the evict thread stops | |
63 | evicting dbufs. | |
64 | .sp | |
65 | Default value: \fB10\fR%. | |
66 | .RE | |
67 | ||
68 | .sp | |
69 | .ne 2 | |
70 | .na | |
71 | \fBdbuf_cache_shift\fR (int) | |
72 | .ad | |
73 | .RS 12n | |
74 | Set the size of the dbuf cache, \fBdbuf_cache_max_bytes\fR, to a log2 fraction | |
75 | of the target arc size. | |
76 | .sp | |
77 | Default value: \fB5\fR. | |
78 | .RE | |
79 | ||
6d836e6f RE |
80 | .sp |
81 | .ne 2 | |
82 | .na | |
83 | \fBignore_hole_birth\fR (int) | |
84 | .ad | |
85 | .RS 12n | |
86 | When set, the hole_birth optimization will not be used, and all holes will | |
87 | always be sent on zfs send. Useful if you suspect your datasets are affected | |
88 | by a bug in hole_birth. | |
89 | .sp | |
9ea9e0b9 | 90 | Use \fB1\fR for on (default) and \fB0\fR for off. |
6d836e6f RE |
91 | .RE |
92 | ||
29714574 TF |
93 | .sp |
94 | .ne 2 | |
95 | .na | |
96 | \fBl2arc_feed_again\fR (int) | |
97 | .ad | |
98 | .RS 12n | |
83426735 D |
99 | Turbo L2ARC warm-up. When the L2ARC is cold the fill interval will be set as |
100 | fast as possible. | |
29714574 TF |
101 | .sp |
102 | Use \fB1\fR for yes (default) and \fB0\fR to disable. | |
103 | .RE | |
104 | ||
105 | .sp | |
106 | .ne 2 | |
107 | .na | |
108 | \fBl2arc_feed_min_ms\fR (ulong) | |
109 | .ad | |
110 | .RS 12n | |
83426735 D |
111 | Min feed interval in milliseconds. Requires \fBl2arc_feed_again=1\fR and only |
112 | applicable in related situations. | |
29714574 TF |
113 | .sp |
114 | Default value: \fB200\fR. | |
115 | .RE | |
116 | ||
117 | .sp | |
118 | .ne 2 | |
119 | .na | |
120 | \fBl2arc_feed_secs\fR (ulong) | |
121 | .ad | |
122 | .RS 12n | |
123 | Seconds between L2ARC writing | |
124 | .sp | |
125 | Default value: \fB1\fR. | |
126 | .RE | |
127 | ||
128 | .sp | |
129 | .ne 2 | |
130 | .na | |
131 | \fBl2arc_headroom\fR (ulong) | |
132 | .ad | |
133 | .RS 12n | |
83426735 D |
134 | How far through the ARC lists to search for L2ARC cacheable content, expressed |
135 | as a multiplier of \fBl2arc_write_max\fR | |
29714574 TF |
136 | .sp |
137 | Default value: \fB2\fR. | |
138 | .RE | |
139 | ||
140 | .sp | |
141 | .ne 2 | |
142 | .na | |
143 | \fBl2arc_headroom_boost\fR (ulong) | |
144 | .ad | |
145 | .RS 12n | |
83426735 D |
146 | Scales \fBl2arc_headroom\fR by this percentage when L2ARC contents are being |
147 | successfully compressed before writing. A value of 100 disables this feature. | |
29714574 | 148 | .sp |
be54a13c | 149 | Default value: \fB200\fR%. |
29714574 TF |
150 | .RE |
151 | ||
29714574 TF |
152 | .sp |
153 | .ne 2 | |
154 | .na | |
155 | \fBl2arc_noprefetch\fR (int) | |
156 | .ad | |
157 | .RS 12n | |
83426735 D |
158 | Do not write buffers to L2ARC if they were prefetched but not used by |
159 | applications | |
29714574 TF |
160 | .sp |
161 | Use \fB1\fR for yes (default) and \fB0\fR to disable. | |
162 | .RE | |
163 | ||
164 | .sp | |
165 | .ne 2 | |
166 | .na | |
167 | \fBl2arc_norw\fR (int) | |
168 | .ad | |
169 | .RS 12n | |
170 | No reads during writes | |
171 | .sp | |
172 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
173 | .RE | |
174 | ||
175 | .sp | |
176 | .ne 2 | |
177 | .na | |
178 | \fBl2arc_write_boost\fR (ulong) | |
179 | .ad | |
180 | .RS 12n | |
603a1784 | 181 | Cold L2ARC devices will have \fBl2arc_write_max\fR increased by this amount |
83426735 | 182 | while they remain cold. |
29714574 TF |
183 | .sp |
184 | Default value: \fB8,388,608\fR. | |
185 | .RE | |
186 | ||
187 | .sp | |
188 | .ne 2 | |
189 | .na | |
190 | \fBl2arc_write_max\fR (ulong) | |
191 | .ad | |
192 | .RS 12n | |
193 | Max write bytes per interval | |
194 | .sp | |
195 | Default value: \fB8,388,608\fR. | |
196 | .RE | |
197 | ||
99b14de4 ED |
198 | .sp |
199 | .ne 2 | |
200 | .na | |
201 | \fBmetaslab_aliquot\fR (ulong) | |
202 | .ad | |
203 | .RS 12n | |
204 | Metaslab granularity, in bytes. This is roughly similar to what would be | |
205 | referred to as the "stripe size" in traditional RAID arrays. In normal | |
206 | operation, ZFS will try to write this amount of data to a top-level vdev | |
207 | before moving on to the next one. | |
208 | .sp | |
209 | Default value: \fB524,288\fR. | |
210 | .RE | |
211 | ||
f3a7f661 GW |
212 | .sp |
213 | .ne 2 | |
214 | .na | |
215 | \fBmetaslab_bias_enabled\fR (int) | |
216 | .ad | |
217 | .RS 12n | |
218 | Enable metaslab group biasing based on its vdev's over- or under-utilization | |
219 | relative to the pool. | |
220 | .sp | |
221 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
222 | .RE | |
223 | ||
d830d479 MA |
224 | .sp |
225 | .ne 2 | |
226 | .na | |
227 | \fBmetaslab_force_ganging\fR (ulong) | |
228 | .ad | |
229 | .RS 12n | |
230 | Make some blocks above a certain size be gang blocks. This option is used | |
231 | by the test suite to facilitate testing. | |
232 | .sp | |
233 | Default value: \fB16,777,217\fR. | |
234 | .RE | |
235 | ||
4e21fd06 DB |
236 | .sp |
237 | .ne 2 | |
238 | .na | |
239 | \fBzfs_metaslab_segment_weight_enabled\fR (int) | |
240 | .ad | |
241 | .RS 12n | |
242 | Enable/disable segment-based metaslab selection. | |
243 | .sp | |
244 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
245 | .RE | |
246 | ||
247 | .sp | |
248 | .ne 2 | |
249 | .na | |
250 | \fBzfs_metaslab_switch_threshold\fR (int) | |
251 | .ad | |
252 | .RS 12n | |
253 | When using segment-based metaslab selection, continue allocating | |
321204be | 254 | from the active metaslab until \fBzfs_metaslab_switch_threshold\fR |
4e21fd06 DB |
255 | worth of buckets have been exhausted. |
256 | .sp | |
257 | Default value: \fB2\fR. | |
258 | .RE | |
259 | ||
29714574 TF |
260 | .sp |
261 | .ne 2 | |
262 | .na | |
aa7d06a9 | 263 | \fBmetaslab_debug_load\fR (int) |
29714574 TF |
264 | .ad |
265 | .RS 12n | |
aa7d06a9 GW |
266 | Load all metaslabs during pool import. |
267 | .sp | |
268 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
269 | .RE | |
270 | ||
271 | .sp | |
272 | .ne 2 | |
273 | .na | |
274 | \fBmetaslab_debug_unload\fR (int) | |
275 | .ad | |
276 | .RS 12n | |
277 | Prevent metaslabs from being unloaded. | |
29714574 TF |
278 | .sp |
279 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
280 | .RE | |
281 | ||
f3a7f661 GW |
282 | .sp |
283 | .ne 2 | |
284 | .na | |
285 | \fBmetaslab_fragmentation_factor_enabled\fR (int) | |
286 | .ad | |
287 | .RS 12n | |
288 | Enable use of the fragmentation metric in computing metaslab weights. | |
289 | .sp | |
290 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
291 | .RE | |
292 | ||
b8bcca18 MA |
293 | .sp |
294 | .ne 2 | |
295 | .na | |
296 | \fBmetaslabs_per_vdev\fR (int) | |
297 | .ad | |
298 | .RS 12n | |
299 | When a vdev is added, it will be divided into approximately (but no more than) this number of metaslabs. | |
300 | .sp | |
301 | Default value: \fB200\fR. | |
302 | .RE | |
303 | ||
f3a7f661 GW |
304 | .sp |
305 | .ne 2 | |
306 | .na | |
307 | \fBmetaslab_preload_enabled\fR (int) | |
308 | .ad | |
309 | .RS 12n | |
310 | Enable metaslab group preloading. | |
311 | .sp | |
312 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
313 | .RE | |
314 | ||
315 | .sp | |
316 | .ne 2 | |
317 | .na | |
318 | \fBmetaslab_lba_weighting_enabled\fR (int) | |
319 | .ad | |
320 | .RS 12n | |
321 | Give more weight to metaslabs with lower LBAs, assuming they have | |
322 | greater bandwidth as is typically the case on a modern constant | |
323 | angular velocity disk drive. | |
324 | .sp | |
325 | Use \fB1\fR for yes (default) and \fB0\fR for no. | |
326 | .RE | |
327 | ||
29714574 TF |
328 | .sp |
329 | .ne 2 | |
330 | .na | |
331 | \fBspa_config_path\fR (charp) | |
332 | .ad | |
333 | .RS 12n | |
334 | SPA config file | |
335 | .sp | |
336 | Default value: \fB/etc/zfs/zpool.cache\fR. | |
337 | .RE | |
338 | ||
e8b96c60 MA |
339 | .sp |
340 | .ne 2 | |
341 | .na | |
342 | \fBspa_asize_inflation\fR (int) | |
343 | .ad | |
344 | .RS 12n | |
345 | Multiplication factor used to estimate actual disk consumption from the | |
346 | size of data being written. The default value is a worst case estimate, | |
347 | but lower values may be valid for a given pool depending on its | |
348 | configuration. Pool administrators who understand the factors involved | |
349 | may wish to specify a more realistic inflation factor, particularly if | |
350 | they operate close to quota or capacity limits. | |
351 | .sp | |
83426735 | 352 | Default value: \fB24\fR. |
e8b96c60 MA |
353 | .RE |
354 | ||
6cb8e530 PZ |
355 | .sp |
356 | .ne 2 | |
357 | .na | |
358 | \fBspa_load_print_vdev_tree\fR (int) | |
359 | .ad | |
360 | .RS 12n | |
361 | Whether to print the vdev tree in the debugging message buffer during pool import. | |
362 | Use 0 to disable and 1 to enable. | |
363 | .sp | |
364 | Default value: \fB0\fR. | |
365 | .RE | |
366 | ||
dea377c0 MA |
367 | .sp |
368 | .ne 2 | |
369 | .na | |
370 | \fBspa_load_verify_data\fR (int) | |
371 | .ad | |
372 | .RS 12n | |
373 | Whether to traverse data blocks during an "extreme rewind" (\fB-X\fR) | |
374 | import. Use 0 to disable and 1 to enable. | |
375 | ||
376 | An extreme rewind import normally performs a full traversal of all | |
377 | blocks in the pool for verification. If this parameter is set to 0, | |
378 | the traversal skips non-metadata blocks. It can be toggled once the | |
379 | import has started to stop or start the traversal of non-metadata blocks. | |
380 | .sp | |
83426735 | 381 | Default value: \fB1\fR. |
dea377c0 MA |
382 | .RE |
383 | ||
384 | .sp | |
385 | .ne 2 | |
386 | .na | |
387 | \fBspa_load_verify_metadata\fR (int) | |
388 | .ad | |
389 | .RS 12n | |
390 | Whether to traverse blocks during an "extreme rewind" (\fB-X\fR) | |
391 | pool import. Use 0 to disable and 1 to enable. | |
392 | ||
393 | An extreme rewind import normally performs a full traversal of all | |
1c012083 | 394 | blocks in the pool for verification. If this parameter is set to 0, |
dea377c0 MA |
395 | the traversal is not performed. It can be toggled once the import has |
396 | started to stop or start the traversal. | |
397 | .sp | |
83426735 | 398 | Default value: \fB1\fR. |
dea377c0 MA |
399 | .RE |
400 | ||
401 | .sp | |
402 | .ne 2 | |
403 | .na | |
404 | \fBspa_load_verify_maxinflight\fR (int) | |
405 | .ad | |
406 | .RS 12n | |
407 | Maximum concurrent I/Os during the traversal performed during an "extreme | |
408 | rewind" (\fB-X\fR) pool import. | |
409 | .sp | |
83426735 | 410 | Default value: \fB10000\fR. |
dea377c0 MA |
411 | .RE |
412 | ||
6cde6435 BB |
413 | .sp |
414 | .ne 2 | |
415 | .na | |
416 | \fBspa_slop_shift\fR (int) | |
417 | .ad | |
418 | .RS 12n | |
419 | Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space | |
420 | in the pool to be consumed. This ensures that we don't run the pool | |
421 | completely out of space, due to unaccounted changes (e.g. to the MOS). | |
422 | It also limits the worst-case time to allocate space. If we have | |
423 | less than this amount of free space, most ZPL operations (e.g. write, | |
424 | create) will return ENOSPC. | |
425 | .sp | |
83426735 | 426 | Default value: \fB5\fR. |
6cde6435 BB |
427 | .RE |
428 | ||
0dc2f70c MA |
429 | .sp |
430 | .ne 2 | |
431 | .na | |
432 | \fBvdev_removal_max_span\fR (int) | |
433 | .ad | |
434 | .RS 12n | |
435 | During top-level vdev removal, chunks of data are copied from the vdev | |
436 | which may include free space in order to trade bandwidth for IOPS. | |
437 | This parameter determines the maximum span of free space (in bytes) | |
438 | which will be included as "unnecessary" data in a chunk of copied data. | |
439 | ||
440 | The default value here was chosen to align with | |
441 | \fBzfs_vdev_read_gap_limit\fR, which is a similar concept when doing | |
442 | regular reads (but there's no reason it has to be the same). | |
443 | .sp | |
444 | Default value: \fB32,768\fR. | |
445 | .RE | |
446 | ||
29714574 TF |
447 | .sp |
448 | .ne 2 | |
449 | .na | |
450 | \fBzfetch_array_rd_sz\fR (ulong) | |
451 | .ad | |
452 | .RS 12n | |
27b293be | 453 | If prefetching is enabled, disable prefetching for reads larger than this size. |
29714574 TF |
454 | .sp |
455 | Default value: \fB1,048,576\fR. | |
456 | .RE | |
457 | ||
458 | .sp | |
459 | .ne 2 | |
460 | .na | |
7f60329a | 461 | \fBzfetch_max_distance\fR (uint) |
29714574 TF |
462 | .ad |
463 | .RS 12n | |
7f60329a | 464 | Max bytes to prefetch per stream (default 8MB). |
29714574 | 465 | .sp |
7f60329a | 466 | Default value: \fB8,388,608\fR. |
29714574 TF |
467 | .RE |
468 | ||
469 | .sp | |
470 | .ne 2 | |
471 | .na | |
472 | \fBzfetch_max_streams\fR (uint) | |
473 | .ad | |
474 | .RS 12n | |
27b293be | 475 | Max number of streams per zfetch (prefetch streams per file). |
29714574 TF |
476 | .sp |
477 | Default value: \fB8\fR. | |
478 | .RE | |
479 | ||
480 | .sp | |
481 | .ne 2 | |
482 | .na | |
483 | \fBzfetch_min_sec_reap\fR (uint) | |
484 | .ad | |
485 | .RS 12n | |
27b293be | 486 | Min time before an active prefetch stream can be reclaimed |
29714574 TF |
487 | .sp |
488 | Default value: \fB2\fR. | |
489 | .RE | |
490 | ||
25458cbe TC |
491 | .sp |
492 | .ne 2 | |
493 | .na | |
494 | \fBzfs_arc_dnode_limit\fR (ulong) | |
495 | .ad | |
496 | .RS 12n | |
497 | When the number of bytes consumed by dnodes in the ARC exceeds this number of | |
9907cc1c | 498 | bytes, try to unpin some of it in response to demand for non-metadata. This |
627791f3 | 499 | value acts as a ceiling to the amount of dnode metadata, and defaults to 0 which |
9907cc1c G |
500 | indicates that a percent which is based on \fBzfs_arc_dnode_limit_percent\fR of |
501 | the ARC meta buffers that may be used for dnodes. | |
25458cbe TC |
502 | |
503 | See also \fBzfs_arc_meta_prune\fR which serves a similar purpose but is used | |
504 | when the amount of metadata in the ARC exceeds \fBzfs_arc_meta_limit\fR rather | |
505 | than in response to overall demand for non-metadata. | |
506 | ||
507 | .sp | |
9907cc1c G |
508 | Default value: \fB0\fR. |
509 | .RE | |
510 | ||
511 | .sp | |
512 | .ne 2 | |
513 | .na | |
514 | \fBzfs_arc_dnode_limit_percent\fR (ulong) | |
515 | .ad | |
516 | .RS 12n | |
517 | Percentage that can be consumed by dnodes of ARC meta buffers. | |
518 | .sp | |
519 | See also \fBzfs_arc_dnode_limit\fR which serves a similar purpose but has a | |
520 | higher priority if set to nonzero value. | |
521 | .sp | |
be54a13c | 522 | Default value: \fB10\fR%. |
25458cbe TC |
523 | .RE |
524 | ||
525 | .sp | |
526 | .ne 2 | |
527 | .na | |
528 | \fBzfs_arc_dnode_reduce_percent\fR (ulong) | |
529 | .ad | |
530 | .RS 12n | |
531 | Percentage of ARC dnodes to try to scan in response to demand for non-metadata | |
6146e17e | 532 | when the number of bytes consumed by dnodes exceeds \fBzfs_arc_dnode_limit\fR. |
25458cbe TC |
533 | |
534 | .sp | |
be54a13c | 535 | Default value: \fB10\fR% of the number of dnodes in the ARC. |
25458cbe TC |
536 | .RE |
537 | ||
49ddb315 MA |
538 | .sp |
539 | .ne 2 | |
540 | .na | |
541 | \fBzfs_arc_average_blocksize\fR (int) | |
542 | .ad | |
543 | .RS 12n | |
544 | The ARC's buffer hash table is sized based on the assumption of an average | |
545 | block size of \fBzfs_arc_average_blocksize\fR (default 8K). This works out | |
546 | to roughly 1MB of hash table per 1GB of physical memory with 8-byte pointers. | |
547 | For configurations with a known larger average block size this value can be | |
548 | increased to reduce the memory footprint. | |
549 | ||
550 | .sp | |
551 | Default value: \fB8192\fR. | |
552 | .RE | |
553 | ||
ca0bf58d PS |
554 | .sp |
555 | .ne 2 | |
556 | .na | |
557 | \fBzfs_arc_evict_batch_limit\fR (int) | |
558 | .ad | |
559 | .RS 12n | |
8f343973 | 560 | Number ARC headers to evict per sub-list before proceeding to another sub-list. |
ca0bf58d PS |
561 | This batch-style operation prevents entire sub-lists from being evicted at once |
562 | but comes at a cost of additional unlocking and locking. | |
563 | .sp | |
564 | Default value: \fB10\fR. | |
565 | .RE | |
566 | ||
29714574 TF |
567 | .sp |
568 | .ne 2 | |
569 | .na | |
570 | \fBzfs_arc_grow_retry\fR (int) | |
571 | .ad | |
572 | .RS 12n | |
ca85d690 | 573 | If set to a non zero value, it will replace the arc_grow_retry value with this value. |
d4a72f23 | 574 | The arc_grow_retry value (default 5) is the number of seconds the ARC will wait before |
ca85d690 | 575 | trying to resume growth after a memory pressure event. |
29714574 | 576 | .sp |
ca85d690 | 577 | Default value: \fB0\fR. |
29714574 TF |
578 | .RE |
579 | ||
580 | .sp | |
581 | .ne 2 | |
582 | .na | |
7e8bddd0 | 583 | \fBzfs_arc_lotsfree_percent\fR (int) |
29714574 TF |
584 | .ad |
585 | .RS 12n | |
7e8bddd0 BB |
586 | Throttle I/O when free system memory drops below this percentage of total |
587 | system memory. Setting this value to 0 will disable the throttle. | |
29714574 | 588 | .sp |
be54a13c | 589 | Default value: \fB10\fR%. |
29714574 TF |
590 | .RE |
591 | ||
592 | .sp | |
593 | .ne 2 | |
594 | .na | |
7e8bddd0 | 595 | \fBzfs_arc_max\fR (ulong) |
29714574 TF |
596 | .ad |
597 | .RS 12n | |
83426735 D |
598 | Max arc size of ARC in bytes. If set to 0 then it will consume 1/2 of system |
599 | RAM. This value must be at least 67108864 (64 megabytes). | |
600 | .sp | |
601 | This value can be changed dynamically with some caveats. It cannot be set back | |
602 | to 0 while running and reducing it below the current ARC size will not cause | |
603 | the ARC to shrink without memory pressure to induce shrinking. | |
29714574 | 604 | .sp |
7e8bddd0 | 605 | Default value: \fB0\fR. |
29714574 TF |
606 | .RE |
607 | ||
ca85d690 | 608 | .sp |
609 | .ne 2 | |
610 | .na | |
611 | \fBzfs_arc_meta_adjust_restarts\fR (ulong) | |
612 | .ad | |
613 | .RS 12n | |
614 | The number of restart passes to make while scanning the ARC attempting | |
615 | the free buffers in order to stay below the \fBzfs_arc_meta_limit\fR. | |
616 | This value should not need to be tuned but is available to facilitate | |
617 | performance analysis. | |
618 | .sp | |
619 | Default value: \fB4096\fR. | |
620 | .RE | |
621 | ||
29714574 TF |
622 | .sp |
623 | .ne 2 | |
624 | .na | |
625 | \fBzfs_arc_meta_limit\fR (ulong) | |
626 | .ad | |
627 | .RS 12n | |
2cbb06b5 BB |
628 | The maximum allowed size in bytes that meta data buffers are allowed to |
629 | consume in the ARC. When this limit is reached meta data buffers will | |
630 | be reclaimed even if the overall arc_c_max has not been reached. This | |
9907cc1c G |
631 | value defaults to 0 which indicates that a percent which is based on |
632 | \fBzfs_arc_meta_limit_percent\fR of the ARC may be used for meta data. | |
29714574 | 633 | .sp |
83426735 | 634 | This value my be changed dynamically except that it cannot be set back to 0 |
9907cc1c | 635 | for a specific percent of the ARC; it must be set to an explicit value. |
83426735 | 636 | .sp |
29714574 TF |
637 | Default value: \fB0\fR. |
638 | .RE | |
639 | ||
9907cc1c G |
640 | .sp |
641 | .ne 2 | |
642 | .na | |
643 | \fBzfs_arc_meta_limit_percent\fR (ulong) | |
644 | .ad | |
645 | .RS 12n | |
646 | Percentage of ARC buffers that can be used for meta data. | |
647 | ||
648 | See also \fBzfs_arc_meta_limit\fR which serves a similar purpose but has a | |
649 | higher priority if set to nonzero value. | |
650 | ||
651 | .sp | |
be54a13c | 652 | Default value: \fB75\fR%. |
9907cc1c G |
653 | .RE |
654 | ||
ca0bf58d PS |
655 | .sp |
656 | .ne 2 | |
657 | .na | |
658 | \fBzfs_arc_meta_min\fR (ulong) | |
659 | .ad | |
660 | .RS 12n | |
661 | The minimum allowed size in bytes that meta data buffers may consume in | |
662 | the ARC. This value defaults to 0 which disables a floor on the amount | |
663 | of the ARC devoted meta data. | |
664 | .sp | |
665 | Default value: \fB0\fR. | |
666 | .RE | |
667 | ||
29714574 TF |
668 | .sp |
669 | .ne 2 | |
670 | .na | |
671 | \fBzfs_arc_meta_prune\fR (int) | |
672 | .ad | |
673 | .RS 12n | |
2cbb06b5 BB |
674 | The number of dentries and inodes to be scanned looking for entries |
675 | which can be dropped. This may be required when the ARC reaches the | |
676 | \fBzfs_arc_meta_limit\fR because dentries and inodes can pin buffers | |
677 | in the ARC. Increasing this value will cause to dentry and inode caches | |
678 | to be pruned more aggressively. Setting this value to 0 will disable | |
679 | pruning the inode and dentry caches. | |
29714574 | 680 | .sp |
2cbb06b5 | 681 | Default value: \fB10,000\fR. |
29714574 TF |
682 | .RE |
683 | ||
bc888666 BB |
684 | .sp |
685 | .ne 2 | |
686 | .na | |
ca85d690 | 687 | \fBzfs_arc_meta_strategy\fR (int) |
bc888666 BB |
688 | .ad |
689 | .RS 12n | |
ca85d690 | 690 | Define the strategy for ARC meta data buffer eviction (meta reclaim strategy). |
691 | A value of 0 (META_ONLY) will evict only the ARC meta data buffers. | |
d4a72f23 | 692 | A value of 1 (BALANCED) indicates that additional data buffers may be evicted if |
ca85d690 | 693 | that is required to in order to evict the required number of meta data buffers. |
bc888666 | 694 | .sp |
ca85d690 | 695 | Default value: \fB1\fR. |
bc888666 BB |
696 | .RE |
697 | ||
29714574 TF |
698 | .sp |
699 | .ne 2 | |
700 | .na | |
701 | \fBzfs_arc_min\fR (ulong) | |
702 | .ad | |
703 | .RS 12n | |
ca85d690 | 704 | Min arc size of ARC in bytes. If set to 0 then arc_c_min will default to |
705 | consuming the larger of 32M or 1/32 of total system memory. | |
29714574 | 706 | .sp |
ca85d690 | 707 | Default value: \fB0\fR. |
29714574 TF |
708 | .RE |
709 | ||
710 | .sp | |
711 | .ne 2 | |
712 | .na | |
d4a72f23 | 713 | \fBzfs_arc_min_prefetch_ms\fR (int) |
29714574 TF |
714 | .ad |
715 | .RS 12n | |
d4a72f23 | 716 | Minimum time prefetched blocks are locked in the ARC, specified in ms. |
2b84817f | 717 | A value of \fB0\fR will default to 1000 ms. |
d4a72f23 TC |
718 | .sp |
719 | Default value: \fB0\fR. | |
720 | .RE | |
721 | ||
722 | .sp | |
723 | .ne 2 | |
724 | .na | |
725 | \fBzfs_arc_min_prescient_prefetch_ms\fR (int) | |
726 | .ad | |
727 | .RS 12n | |
728 | Minimum time "prescient prefetched" blocks are locked in the ARC, specified | |
729 | in ms. These blocks are meant to be prefetched fairly aggresively ahead of | |
2b84817f | 730 | the code that may use them. A value of \fB0\fR will default to 6000 ms. |
29714574 | 731 | .sp |
83426735 | 732 | Default value: \fB0\fR. |
29714574 TF |
733 | .RE |
734 | ||
6cb8e530 PZ |
735 | .sp |
736 | .ne 2 | |
737 | .na | |
738 | \fBzfs_max_missing_tvds\fR (int) | |
739 | .ad | |
740 | .RS 12n | |
741 | Number of missing top-level vdevs which will be allowed during | |
742 | pool import (only in read-only mode). | |
743 | .sp | |
744 | Default value: \fB0\fR | |
745 | .RE | |
746 | ||
ca0bf58d PS |
747 | .sp |
748 | .ne 2 | |
749 | .na | |
c30e58c4 | 750 | \fBzfs_multilist_num_sublists\fR (int) |
ca0bf58d PS |
751 | .ad |
752 | .RS 12n | |
753 | To allow more fine-grained locking, each ARC state contains a series | |
754 | of lists for both data and meta data objects. Locking is performed at | |
755 | the level of these "sub-lists". This parameters controls the number of | |
c30e58c4 MA |
756 | sub-lists per ARC state, and also applies to other uses of the |
757 | multilist data structure. | |
ca0bf58d | 758 | .sp |
c30e58c4 | 759 | Default value: \fB4\fR or the number of online CPUs, whichever is greater |
ca0bf58d PS |
760 | .RE |
761 | ||
762 | .sp | |
763 | .ne 2 | |
764 | .na | |
765 | \fBzfs_arc_overflow_shift\fR (int) | |
766 | .ad | |
767 | .RS 12n | |
768 | The ARC size is considered to be overflowing if it exceeds the current | |
769 | ARC target size (arc_c) by a threshold determined by this parameter. | |
770 | The threshold is calculated as a fraction of arc_c using the formula | |
771 | "arc_c >> \fBzfs_arc_overflow_shift\fR". | |
772 | ||
773 | The default value of 8 causes the ARC to be considered to be overflowing | |
774 | if it exceeds the target size by 1/256th (0.3%) of the target size. | |
775 | ||
776 | When the ARC is overflowing, new buffer allocations are stalled until | |
777 | the reclaim thread catches up and the overflow condition no longer exists. | |
778 | .sp | |
779 | Default value: \fB8\fR. | |
780 | .RE | |
781 | ||
728d6ae9 BB |
782 | .sp |
783 | .ne 2 | |
784 | .na | |
785 | ||
786 | \fBzfs_arc_p_min_shift\fR (int) | |
787 | .ad | |
788 | .RS 12n | |
ca85d690 | 789 | If set to a non zero value, this will update arc_p_min_shift (default 4) |
790 | with the new value. | |
d4a72f23 | 791 | arc_p_min_shift is used to shift of arc_c for calculating both min and max |
ca85d690 | 792 | max arc_p |
728d6ae9 | 793 | .sp |
ca85d690 | 794 | Default value: \fB0\fR. |
728d6ae9 BB |
795 | .RE |
796 | ||
62422785 PS |
797 | .sp |
798 | .ne 2 | |
799 | .na | |
800 | \fBzfs_arc_p_dampener_disable\fR (int) | |
801 | .ad | |
802 | .RS 12n | |
803 | Disable arc_p adapt dampener | |
804 | .sp | |
805 | Use \fB1\fR for yes (default) and \fB0\fR to disable. | |
806 | .RE | |
807 | ||
29714574 TF |
808 | .sp |
809 | .ne 2 | |
810 | .na | |
811 | \fBzfs_arc_shrink_shift\fR (int) | |
812 | .ad | |
813 | .RS 12n | |
ca85d690 | 814 | If set to a non zero value, this will update arc_shrink_shift (default 7) |
815 | with the new value. | |
29714574 | 816 | .sp |
ca85d690 | 817 | Default value: \fB0\fR. |
29714574 TF |
818 | .RE |
819 | ||
03b60eee DB |
820 | .sp |
821 | .ne 2 | |
822 | .na | |
823 | \fBzfs_arc_pc_percent\fR (uint) | |
824 | .ad | |
825 | .RS 12n | |
826 | Percent of pagecache to reclaim arc to | |
827 | ||
828 | This tunable allows ZFS arc to play more nicely with the kernel's LRU | |
829 | pagecache. It can guarantee that the arc size won't collapse under scanning | |
830 | pressure on the pagecache, yet still allows arc to be reclaimed down to | |
831 | zfs_arc_min if necessary. This value is specified as percent of pagecache | |
832 | size (as measured by NR_FILE_PAGES) where that percent may exceed 100. This | |
833 | only operates during memory pressure/reclaim. | |
834 | .sp | |
be54a13c | 835 | Default value: \fB0\fR% (disabled). |
03b60eee DB |
836 | .RE |
837 | ||
11f552fa BB |
838 | .sp |
839 | .ne 2 | |
840 | .na | |
841 | \fBzfs_arc_sys_free\fR (ulong) | |
842 | .ad | |
843 | .RS 12n | |
844 | The target number of bytes the ARC should leave as free memory on the system. | |
845 | Defaults to the larger of 1/64 of physical memory or 512K. Setting this | |
846 | option to a non-zero value will override the default. | |
847 | .sp | |
848 | Default value: \fB0\fR. | |
849 | .RE | |
850 | ||
29714574 TF |
851 | .sp |
852 | .ne 2 | |
853 | .na | |
854 | \fBzfs_autoimport_disable\fR (int) | |
855 | .ad | |
856 | .RS 12n | |
27b293be | 857 | Disable pool import at module load by ignoring the cache file (typically \fB/etc/zfs/zpool.cache\fR). |
29714574 | 858 | .sp |
70081096 | 859 | Use \fB1\fR for yes (default) and \fB0\fR for no. |
29714574 TF |
860 | .RE |
861 | ||
80d52c39 TH |
862 | .sp |
863 | .ne 2 | |
864 | .na | |
865 | \fBzfs_checksums_per_second\fR (int) | |
866 | .ad | |
867 | .RS 12n | |
868 | Rate limit checksum events to this many per second. Note that this should | |
869 | not be set below the zed thresholds (currently 10 checksums over 10 sec) | |
870 | or else zed may not trigger any action. | |
871 | .sp | |
872 | Default value: 20 | |
873 | .RE | |
874 | ||
2fe61a7e PS |
875 | .sp |
876 | .ne 2 | |
877 | .na | |
878 | \fBzfs_commit_timeout_pct\fR (int) | |
879 | .ad | |
880 | .RS 12n | |
881 | This controls the amount of time that a ZIL block (lwb) will remain "open" | |
882 | when it isn't "full", and it has a thread waiting for it to be committed to | |
883 | stable storage. The timeout is scaled based on a percentage of the last lwb | |
884 | latency to avoid significantly impacting the latency of each individual | |
885 | transaction record (itx). | |
886 | .sp | |
be54a13c | 887 | Default value: \fB5\fR%. |
2fe61a7e PS |
888 | .RE |
889 | ||
0dc2f70c MA |
890 | .sp |
891 | .ne 2 | |
892 | .na | |
893 | \fBzfs_condense_indirect_vdevs_enable\fR (int) | |
894 | .ad | |
895 | .RS 12n | |
896 | Enable condensing indirect vdev mappings. When set to a non-zero value, | |
897 | attempt to condense indirect vdev mappings if the mapping uses more than | |
898 | \fBzfs_condense_min_mapping_bytes\fR bytes of memory and if the obsolete | |
899 | space map object uses more than \fBzfs_condense_max_obsolete_bytes\fR | |
900 | bytes on-disk. The condensing process is an attempt to save memory by | |
901 | removing obsolete mappings. | |
902 | .sp | |
903 | Default value: \fB1\fR. | |
904 | .RE | |
905 | ||
906 | .sp | |
907 | .ne 2 | |
908 | .na | |
909 | \fBzfs_condense_max_obsolete_bytes\fR (ulong) | |
910 | .ad | |
911 | .RS 12n | |
912 | Only attempt to condense indirect vdev mappings if the on-disk size | |
913 | of the obsolete space map object is greater than this number of bytes | |
914 | (see \fBfBzfs_condense_indirect_vdevs_enable\fR). | |
915 | .sp | |
916 | Default value: \fB1,073,741,824\fR. | |
917 | .RE | |
918 | ||
919 | .sp | |
920 | .ne 2 | |
921 | .na | |
922 | \fBzfs_condense_min_mapping_bytes\fR (ulong) | |
923 | .ad | |
924 | .RS 12n | |
925 | Minimum size vdev mapping to attempt to condense (see | |
926 | \fBzfs_condense_indirect_vdevs_enable\fR). | |
927 | .sp | |
928 | Default value: \fB131,072\fR. | |
929 | .RE | |
930 | ||
3b36f831 BB |
931 | .sp |
932 | .ne 2 | |
933 | .na | |
934 | \fBzfs_dbgmsg_enable\fR (int) | |
935 | .ad | |
936 | .RS 12n | |
937 | Internally ZFS keeps a small log to facilitate debugging. By default the log | |
938 | is disabled, to enable it set this option to 1. The contents of the log can | |
939 | be accessed by reading the /proc/spl/kstat/zfs/dbgmsg file. Writing 0 to | |
940 | this proc file clears the log. | |
941 | .sp | |
942 | Default value: \fB0\fR. | |
943 | .RE | |
944 | ||
945 | .sp | |
946 | .ne 2 | |
947 | .na | |
948 | \fBzfs_dbgmsg_maxsize\fR (int) | |
949 | .ad | |
950 | .RS 12n | |
951 | The maximum size in bytes of the internal ZFS debug log. | |
952 | .sp | |
953 | Default value: \fB4M\fR. | |
954 | .RE | |
955 | ||
29714574 TF |
956 | .sp |
957 | .ne 2 | |
958 | .na | |
959 | \fBzfs_dbuf_state_index\fR (int) | |
960 | .ad | |
961 | .RS 12n | |
83426735 D |
962 | This feature is currently unused. It is normally used for controlling what |
963 | reporting is available under /proc/spl/kstat/zfs. | |
29714574 TF |
964 | .sp |
965 | Default value: \fB0\fR. | |
966 | .RE | |
967 | ||
968 | .sp | |
969 | .ne 2 | |
970 | .na | |
971 | \fBzfs_deadman_enabled\fR (int) | |
972 | .ad | |
973 | .RS 12n | |
b81a3ddc | 974 | When a pool sync operation takes longer than \fBzfs_deadman_synctime_ms\fR |
8fb1ede1 BB |
975 | milliseconds, or when an individual I/O takes longer than |
976 | \fBzfs_deadman_ziotime_ms\fR milliseconds, then the operation is considered to | |
977 | be "hung". If \fBzfs_deadman_enabled\fR is set then the deadman behavior is | |
978 | invoked as described by the \fBzfs_deadman_failmode\fR module option. | |
979 | By default the deadman is enabled and configured to \fBwait\fR which results | |
980 | in "hung" I/Os only being logged. The deadman is automatically disabled | |
981 | when a pool gets suspended. | |
29714574 | 982 | .sp |
8fb1ede1 BB |
983 | Default value: \fB1\fR. |
984 | .RE | |
985 | ||
986 | .sp | |
987 | .ne 2 | |
988 | .na | |
989 | \fBzfs_deadman_failmode\fR (charp) | |
990 | .ad | |
991 | .RS 12n | |
992 | Controls the failure behavior when the deadman detects a "hung" I/O. Valid | |
993 | values are \fBwait\fR, \fBcontinue\fR, and \fBpanic\fR. | |
994 | .sp | |
995 | \fBwait\fR - Wait for a "hung" I/O to complete. For each "hung" I/O a | |
996 | "deadman" event will be posted describing that I/O. | |
997 | .sp | |
998 | \fBcontinue\fR - Attempt to recover from a "hung" I/O by re-dispatching it | |
999 | to the I/O pipeline if possible. | |
1000 | .sp | |
1001 | \fBpanic\fR - Panic the system. This can be used to facilitate an automatic | |
1002 | fail-over to a properly configured fail-over partner. | |
1003 | .sp | |
1004 | Default value: \fBwait\fR. | |
b81a3ddc TC |
1005 | .RE |
1006 | ||
1007 | .sp | |
1008 | .ne 2 | |
1009 | .na | |
1010 | \fBzfs_deadman_checktime_ms\fR (int) | |
1011 | .ad | |
1012 | .RS 12n | |
8fb1ede1 BB |
1013 | Check time in milliseconds. This defines the frequency at which we check |
1014 | for hung I/O and potentially invoke the \fBzfs_deadman_failmode\fR behavior. | |
b81a3ddc | 1015 | .sp |
8fb1ede1 | 1016 | Default value: \fB60,000\fR. |
29714574 TF |
1017 | .RE |
1018 | ||
1019 | .sp | |
1020 | .ne 2 | |
1021 | .na | |
e8b96c60 | 1022 | \fBzfs_deadman_synctime_ms\fR (ulong) |
29714574 TF |
1023 | .ad |
1024 | .RS 12n | |
b81a3ddc | 1025 | Interval in milliseconds after which the deadman is triggered and also |
8fb1ede1 BB |
1026 | the interval after which a pool sync operation is considered to be "hung". |
1027 | Once this limit is exceeded the deadman will be invoked every | |
1028 | \fBzfs_deadman_checktime_ms\fR milliseconds until the pool sync completes. | |
1029 | .sp | |
1030 | Default value: \fB600,000\fR. | |
1031 | .RE | |
b81a3ddc | 1032 | |
29714574 | 1033 | .sp |
8fb1ede1 BB |
1034 | .ne 2 |
1035 | .na | |
1036 | \fBzfs_deadman_ziotime_ms\fR (ulong) | |
1037 | .ad | |
1038 | .RS 12n | |
1039 | Interval in milliseconds after which the deadman is triggered and an | |
1040 | individual IO operation is considered to be "hung". As long as the I/O | |
1041 | remains "hung" the deadman will be invoked every \fBzfs_deadman_checktime_ms\fR | |
1042 | milliseconds until the I/O completes. | |
1043 | .sp | |
1044 | Default value: \fB300,000\fR. | |
29714574 TF |
1045 | .RE |
1046 | ||
1047 | .sp | |
1048 | .ne 2 | |
1049 | .na | |
1050 | \fBzfs_dedup_prefetch\fR (int) | |
1051 | .ad | |
1052 | .RS 12n | |
1053 | Enable prefetching dedup-ed blks | |
1054 | .sp | |
0dfc7324 | 1055 | Use \fB1\fR for yes and \fB0\fR to disable (default). |
29714574 TF |
1056 | .RE |
1057 | ||
e8b96c60 MA |
1058 | .sp |
1059 | .ne 2 | |
1060 | .na | |
1061 | \fBzfs_delay_min_dirty_percent\fR (int) | |
1062 | .ad | |
1063 | .RS 12n | |
1064 | Start to delay each transaction once there is this amount of dirty data, | |
1065 | expressed as a percentage of \fBzfs_dirty_data_max\fR. | |
1066 | This value should be >= zfs_vdev_async_write_active_max_dirty_percent. | |
1067 | See the section "ZFS TRANSACTION DELAY". | |
1068 | .sp | |
be54a13c | 1069 | Default value: \fB60\fR%. |
e8b96c60 MA |
1070 | .RE |
1071 | ||
1072 | .sp | |
1073 | .ne 2 | |
1074 | .na | |
1075 | \fBzfs_delay_scale\fR (int) | |
1076 | .ad | |
1077 | .RS 12n | |
1078 | This controls how quickly the transaction delay approaches infinity. | |
1079 | Larger values cause longer delays for a given amount of dirty data. | |
1080 | .sp | |
1081 | For the smoothest delay, this value should be about 1 billion divided | |
1082 | by the maximum number of operations per second. This will smoothly | |
1083 | handle between 10x and 1/10th this number. | |
1084 | .sp | |
1085 | See the section "ZFS TRANSACTION DELAY". | |
1086 | .sp | |
1087 | Note: \fBzfs_delay_scale\fR * \fBzfs_dirty_data_max\fR must be < 2^64. | |
1088 | .sp | |
1089 | Default value: \fB500,000\fR. | |
1090 | .RE | |
1091 | ||
80d52c39 TH |
1092 | .sp |
1093 | .ne 2 | |
1094 | .na | |
1095 | \fBzfs_delays_per_second\fR (int) | |
1096 | .ad | |
1097 | .RS 12n | |
1098 | Rate limit IO delay events to this many per second. | |
1099 | .sp | |
1100 | Default value: 20 | |
1101 | .RE | |
1102 | ||
a966c564 K |
1103 | .sp |
1104 | .ne 2 | |
1105 | .na | |
1106 | \fBzfs_delete_blocks\fR (ulong) | |
1107 | .ad | |
1108 | .RS 12n | |
1109 | This is the used to define a large file for the purposes of delete. Files | |
1110 | containing more than \fBzfs_delete_blocks\fR will be deleted asynchronously | |
1111 | while smaller files are deleted synchronously. Decreasing this value will | |
1112 | reduce the time spent in an unlink(2) system call at the expense of a longer | |
1113 | delay before the freed space is available. | |
1114 | .sp | |
1115 | Default value: \fB20,480\fR. | |
1116 | .RE | |
1117 | ||
e8b96c60 MA |
1118 | .sp |
1119 | .ne 2 | |
1120 | .na | |
1121 | \fBzfs_dirty_data_max\fR (int) | |
1122 | .ad | |
1123 | .RS 12n | |
1124 | Determines the dirty space limit in bytes. Once this limit is exceeded, new | |
1125 | writes are halted until space frees up. This parameter takes precedence | |
1126 | over \fBzfs_dirty_data_max_percent\fR. | |
1127 | See the section "ZFS TRANSACTION DELAY". | |
1128 | .sp | |
be54a13c | 1129 | Default value: \fB10\fR% of physical RAM, capped at \fBzfs_dirty_data_max_max\fR. |
e8b96c60 MA |
1130 | .RE |
1131 | ||
1132 | .sp | |
1133 | .ne 2 | |
1134 | .na | |
1135 | \fBzfs_dirty_data_max_max\fR (int) | |
1136 | .ad | |
1137 | .RS 12n | |
1138 | Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed in bytes. | |
1139 | This limit is only enforced at module load time, and will be ignored if | |
1140 | \fBzfs_dirty_data_max\fR is later changed. This parameter takes | |
1141 | precedence over \fBzfs_dirty_data_max_max_percent\fR. See the section | |
1142 | "ZFS TRANSACTION DELAY". | |
1143 | .sp | |
be54a13c | 1144 | Default value: \fB25\fR% of physical RAM. |
e8b96c60 MA |
1145 | .RE |
1146 | ||
1147 | .sp | |
1148 | .ne 2 | |
1149 | .na | |
1150 | \fBzfs_dirty_data_max_max_percent\fR (int) | |
1151 | .ad | |
1152 | .RS 12n | |
1153 | Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed as a | |
1154 | percentage of physical RAM. This limit is only enforced at module load | |
1155 | time, and will be ignored if \fBzfs_dirty_data_max\fR is later changed. | |
1156 | The parameter \fBzfs_dirty_data_max_max\fR takes precedence over this | |
1157 | one. See the section "ZFS TRANSACTION DELAY". | |
1158 | .sp | |
be54a13c | 1159 | Default value: \fB25\fR%. |
e8b96c60 MA |
1160 | .RE |
1161 | ||
1162 | .sp | |
1163 | .ne 2 | |
1164 | .na | |
1165 | \fBzfs_dirty_data_max_percent\fR (int) | |
1166 | .ad | |
1167 | .RS 12n | |
1168 | Determines the dirty space limit, expressed as a percentage of all | |
1169 | memory. Once this limit is exceeded, new writes are halted until space frees | |
1170 | up. The parameter \fBzfs_dirty_data_max\fR takes precedence over this | |
1171 | one. See the section "ZFS TRANSACTION DELAY". | |
1172 | .sp | |
be54a13c | 1173 | Default value: \fB10\fR%, subject to \fBzfs_dirty_data_max_max\fR. |
e8b96c60 MA |
1174 | .RE |
1175 | ||
1176 | .sp | |
1177 | .ne 2 | |
1178 | .na | |
1179 | \fBzfs_dirty_data_sync\fR (int) | |
1180 | .ad | |
1181 | .RS 12n | |
1182 | Start syncing out a transaction group if there is at least this much dirty data. | |
1183 | .sp | |
1184 | Default value: \fB67,108,864\fR. | |
1185 | .RE | |
1186 | ||
1eeb4562 JX |
1187 | .sp |
1188 | .ne 2 | |
1189 | .na | |
1190 | \fBzfs_fletcher_4_impl\fR (string) | |
1191 | .ad | |
1192 | .RS 12n | |
1193 | Select a fletcher 4 implementation. | |
1194 | .sp | |
35a76a03 | 1195 | Supported selectors are: \fBfastest\fR, \fBscalar\fR, \fBsse2\fR, \fBssse3\fR, |
24cdeaf1 | 1196 | \fBavx2\fR, \fBavx512f\fR, and \fBaarch64_neon\fR. |
70b258fc GN |
1197 | All of the selectors except \fBfastest\fR and \fBscalar\fR require instruction |
1198 | set extensions to be available and will only appear if ZFS detects that they are | |
1199 | present at runtime. If multiple implementations of fletcher 4 are available, | |
1200 | the \fBfastest\fR will be chosen using a micro benchmark. Selecting \fBscalar\fR | |
1201 | results in the original, CPU based calculation, being used. Selecting any option | |
1202 | other than \fBfastest\fR and \fBscalar\fR results in vector instructions from | |
1203 | the respective CPU instruction set being used. | |
1eeb4562 JX |
1204 | .sp |
1205 | Default value: \fBfastest\fR. | |
1206 | .RE | |
1207 | ||
ba5ad9a4 GW |
1208 | .sp |
1209 | .ne 2 | |
1210 | .na | |
1211 | \fBzfs_free_bpobj_enabled\fR (int) | |
1212 | .ad | |
1213 | .RS 12n | |
1214 | Enable/disable the processing of the free_bpobj object. | |
1215 | .sp | |
1216 | Default value: \fB1\fR. | |
1217 | .RE | |
1218 | ||
36283ca2 MG |
1219 | .sp |
1220 | .ne 2 | |
1221 | .na | |
a1d477c2 | 1222 | \fBzfs_async_block_max_blocks\fR (ulong) |
36283ca2 MG |
1223 | .ad |
1224 | .RS 12n | |
1225 | Maximum number of blocks freed in a single txg. | |
1226 | .sp | |
1227 | Default value: \fB100,000\fR. | |
1228 | .RE | |
1229 | ||
ca0845d5 PD |
1230 | .sp |
1231 | .ne 2 | |
1232 | .na | |
1233 | \fBzfs_override_estimate_recordsize\fR (ulong) | |
1234 | .ad | |
1235 | .RS 12n | |
1236 | Record size calculation override for zfs send estimates. | |
1237 | .sp | |
1238 | Default value: \fB0\fR. | |
1239 | .RE | |
1240 | ||
e8b96c60 MA |
1241 | .sp |
1242 | .ne 2 | |
1243 | .na | |
1244 | \fBzfs_vdev_async_read_max_active\fR (int) | |
1245 | .ad | |
1246 | .RS 12n | |
83426735 | 1247 | Maximum asynchronous read I/Os active to each device. |
e8b96c60 MA |
1248 | See the section "ZFS I/O SCHEDULER". |
1249 | .sp | |
1250 | Default value: \fB3\fR. | |
1251 | .RE | |
1252 | ||
1253 | .sp | |
1254 | .ne 2 | |
1255 | .na | |
1256 | \fBzfs_vdev_async_read_min_active\fR (int) | |
1257 | .ad | |
1258 | .RS 12n | |
1259 | Minimum asynchronous read I/Os active to each device. | |
1260 | See the section "ZFS I/O SCHEDULER". | |
1261 | .sp | |
1262 | Default value: \fB1\fR. | |
1263 | .RE | |
1264 | ||
1265 | .sp | |
1266 | .ne 2 | |
1267 | .na | |
1268 | \fBzfs_vdev_async_write_active_max_dirty_percent\fR (int) | |
1269 | .ad | |
1270 | .RS 12n | |
1271 | When the pool has more than | |
1272 | \fBzfs_vdev_async_write_active_max_dirty_percent\fR dirty data, use | |
1273 | \fBzfs_vdev_async_write_max_active\fR to limit active async writes. If | |
1274 | the dirty data is between min and max, the active I/O limit is linearly | |
1275 | interpolated. See the section "ZFS I/O SCHEDULER". | |
1276 | .sp | |
be54a13c | 1277 | Default value: \fB60\fR%. |
e8b96c60 MA |
1278 | .RE |
1279 | ||
1280 | .sp | |
1281 | .ne 2 | |
1282 | .na | |
1283 | \fBzfs_vdev_async_write_active_min_dirty_percent\fR (int) | |
1284 | .ad | |
1285 | .RS 12n | |
1286 | When the pool has less than | |
1287 | \fBzfs_vdev_async_write_active_min_dirty_percent\fR dirty data, use | |
1288 | \fBzfs_vdev_async_write_min_active\fR to limit active async writes. If | |
1289 | the dirty data is between min and max, the active I/O limit is linearly | |
1290 | interpolated. See the section "ZFS I/O SCHEDULER". | |
1291 | .sp | |
be54a13c | 1292 | Default value: \fB30\fR%. |
e8b96c60 MA |
1293 | .RE |
1294 | ||
1295 | .sp | |
1296 | .ne 2 | |
1297 | .na | |
1298 | \fBzfs_vdev_async_write_max_active\fR (int) | |
1299 | .ad | |
1300 | .RS 12n | |
83426735 | 1301 | Maximum asynchronous write I/Os active to each device. |
e8b96c60 MA |
1302 | See the section "ZFS I/O SCHEDULER". |
1303 | .sp | |
1304 | Default value: \fB10\fR. | |
1305 | .RE | |
1306 | ||
1307 | .sp | |
1308 | .ne 2 | |
1309 | .na | |
1310 | \fBzfs_vdev_async_write_min_active\fR (int) | |
1311 | .ad | |
1312 | .RS 12n | |
1313 | Minimum asynchronous write I/Os active to each device. | |
1314 | See the section "ZFS I/O SCHEDULER". | |
1315 | .sp | |
06226b59 D |
1316 | Lower values are associated with better latency on rotational media but poorer |
1317 | resilver performance. The default value of 2 was chosen as a compromise. A | |
1318 | value of 3 has been shown to improve resilver performance further at a cost of | |
1319 | further increasing latency. | |
1320 | .sp | |
1321 | Default value: \fB2\fR. | |
e8b96c60 MA |
1322 | .RE |
1323 | ||
1324 | .sp | |
1325 | .ne 2 | |
1326 | .na | |
1327 | \fBzfs_vdev_max_active\fR (int) | |
1328 | .ad | |
1329 | .RS 12n | |
1330 | The maximum number of I/Os active to each device. Ideally, this will be >= | |
1331 | the sum of each queue's max_active. It must be at least the sum of each | |
1332 | queue's min_active. See the section "ZFS I/O SCHEDULER". | |
1333 | .sp | |
1334 | Default value: \fB1,000\fR. | |
1335 | .RE | |
1336 | ||
1337 | .sp | |
1338 | .ne 2 | |
1339 | .na | |
1340 | \fBzfs_vdev_scrub_max_active\fR (int) | |
1341 | .ad | |
1342 | .RS 12n | |
83426735 | 1343 | Maximum scrub I/Os active to each device. |
e8b96c60 MA |
1344 | See the section "ZFS I/O SCHEDULER". |
1345 | .sp | |
1346 | Default value: \fB2\fR. | |
1347 | .RE | |
1348 | ||
1349 | .sp | |
1350 | .ne 2 | |
1351 | .na | |
1352 | \fBzfs_vdev_scrub_min_active\fR (int) | |
1353 | .ad | |
1354 | .RS 12n | |
1355 | Minimum scrub I/Os active to each device. | |
1356 | See the section "ZFS I/O SCHEDULER". | |
1357 | .sp | |
1358 | Default value: \fB1\fR. | |
1359 | .RE | |
1360 | ||
1361 | .sp | |
1362 | .ne 2 | |
1363 | .na | |
1364 | \fBzfs_vdev_sync_read_max_active\fR (int) | |
1365 | .ad | |
1366 | .RS 12n | |
83426735 | 1367 | Maximum synchronous read I/Os active to each device. |
e8b96c60 MA |
1368 | See the section "ZFS I/O SCHEDULER". |
1369 | .sp | |
1370 | Default value: \fB10\fR. | |
1371 | .RE | |
1372 | ||
1373 | .sp | |
1374 | .ne 2 | |
1375 | .na | |
1376 | \fBzfs_vdev_sync_read_min_active\fR (int) | |
1377 | .ad | |
1378 | .RS 12n | |
1379 | Minimum synchronous read I/Os active to each device. | |
1380 | See the section "ZFS I/O SCHEDULER". | |
1381 | .sp | |
1382 | Default value: \fB10\fR. | |
1383 | .RE | |
1384 | ||
1385 | .sp | |
1386 | .ne 2 | |
1387 | .na | |
1388 | \fBzfs_vdev_sync_write_max_active\fR (int) | |
1389 | .ad | |
1390 | .RS 12n | |
83426735 | 1391 | Maximum synchronous write I/Os active to each device. |
e8b96c60 MA |
1392 | See the section "ZFS I/O SCHEDULER". |
1393 | .sp | |
1394 | Default value: \fB10\fR. | |
1395 | .RE | |
1396 | ||
1397 | .sp | |
1398 | .ne 2 | |
1399 | .na | |
1400 | \fBzfs_vdev_sync_write_min_active\fR (int) | |
1401 | .ad | |
1402 | .RS 12n | |
1403 | Minimum synchronous write I/Os active to each device. | |
1404 | See the section "ZFS I/O SCHEDULER". | |
1405 | .sp | |
1406 | Default value: \fB10\fR. | |
1407 | .RE | |
1408 | ||
3dfb57a3 DB |
1409 | .sp |
1410 | .ne 2 | |
1411 | .na | |
1412 | \fBzfs_vdev_queue_depth_pct\fR (int) | |
1413 | .ad | |
1414 | .RS 12n | |
e815485f TC |
1415 | Maximum number of queued allocations per top-level vdev expressed as |
1416 | a percentage of \fBzfs_vdev_async_write_max_active\fR which allows the | |
1417 | system to detect devices that are more capable of handling allocations | |
1418 | and to allocate more blocks to those devices. It allows for dynamic | |
1419 | allocation distribution when devices are imbalanced as fuller devices | |
1420 | will tend to be slower than empty devices. | |
1421 | ||
1422 | See also \fBzio_dva_throttle_enabled\fR. | |
3dfb57a3 | 1423 | .sp |
be54a13c | 1424 | Default value: \fB1000\fR%. |
3dfb57a3 DB |
1425 | .RE |
1426 | ||
29714574 TF |
1427 | .sp |
1428 | .ne 2 | |
1429 | .na | |
1430 | \fBzfs_expire_snapshot\fR (int) | |
1431 | .ad | |
1432 | .RS 12n | |
1433 | Seconds to expire .zfs/snapshot | |
1434 | .sp | |
1435 | Default value: \fB300\fR. | |
1436 | .RE | |
1437 | ||
0500e835 BB |
1438 | .sp |
1439 | .ne 2 | |
1440 | .na | |
1441 | \fBzfs_admin_snapshot\fR (int) | |
1442 | .ad | |
1443 | .RS 12n | |
1444 | Allow the creation, removal, or renaming of entries in the .zfs/snapshot | |
1445 | directory to cause the creation, destruction, or renaming of snapshots. | |
1446 | When enabled this functionality works both locally and over NFS exports | |
1447 | which have the 'no_root_squash' option set. This functionality is disabled | |
1448 | by default. | |
1449 | .sp | |
1450 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1451 | .RE | |
1452 | ||
29714574 TF |
1453 | .sp |
1454 | .ne 2 | |
1455 | .na | |
1456 | \fBzfs_flags\fR (int) | |
1457 | .ad | |
1458 | .RS 12n | |
33b6dbbc NB |
1459 | Set additional debugging flags. The following flags may be bitwise-or'd |
1460 | together. | |
1461 | .sp | |
1462 | .TS | |
1463 | box; | |
1464 | rB lB | |
1465 | lB lB | |
1466 | r l. | |
1467 | Value Symbolic Name | |
1468 | Description | |
1469 | _ | |
1470 | 1 ZFS_DEBUG_DPRINTF | |
1471 | Enable dprintf entries in the debug log. | |
1472 | _ | |
1473 | 2 ZFS_DEBUG_DBUF_VERIFY * | |
1474 | Enable extra dbuf verifications. | |
1475 | _ | |
1476 | 4 ZFS_DEBUG_DNODE_VERIFY * | |
1477 | Enable extra dnode verifications. | |
1478 | _ | |
1479 | 8 ZFS_DEBUG_SNAPNAMES | |
1480 | Enable snapshot name verification. | |
1481 | _ | |
1482 | 16 ZFS_DEBUG_MODIFY | |
1483 | Check for illegally modified ARC buffers. | |
1484 | _ | |
33b6dbbc NB |
1485 | 64 ZFS_DEBUG_ZIO_FREE |
1486 | Enable verification of block frees. | |
1487 | _ | |
1488 | 128 ZFS_DEBUG_HISTOGRAM_VERIFY | |
1489 | Enable extra spacemap histogram verifications. | |
8740cf4a NB |
1490 | _ |
1491 | 256 ZFS_DEBUG_METASLAB_VERIFY | |
1492 | Verify space accounting on disk matches in-core range_trees. | |
1493 | _ | |
1494 | 512 ZFS_DEBUG_SET_ERROR | |
1495 | Enable SET_ERROR and dprintf entries in the debug log. | |
33b6dbbc NB |
1496 | .TE |
1497 | .sp | |
1498 | * Requires debug build. | |
29714574 | 1499 | .sp |
33b6dbbc | 1500 | Default value: \fB0\fR. |
29714574 TF |
1501 | .RE |
1502 | ||
fbeddd60 MA |
1503 | .sp |
1504 | .ne 2 | |
1505 | .na | |
1506 | \fBzfs_free_leak_on_eio\fR (int) | |
1507 | .ad | |
1508 | .RS 12n | |
1509 | If destroy encounters an EIO while reading metadata (e.g. indirect | |
1510 | blocks), space referenced by the missing metadata can not be freed. | |
1511 | Normally this causes the background destroy to become "stalled", as | |
1512 | it is unable to make forward progress. While in this stalled state, | |
1513 | all remaining space to free from the error-encountering filesystem is | |
1514 | "temporarily leaked". Set this flag to cause it to ignore the EIO, | |
1515 | permanently leak the space from indirect blocks that can not be read, | |
1516 | and continue to free everything else that it can. | |
1517 | ||
1518 | The default, "stalling" behavior is useful if the storage partially | |
1519 | fails (i.e. some but not all i/os fail), and then later recovers. In | |
1520 | this case, we will be able to continue pool operations while it is | |
1521 | partially failed, and when it recovers, we can continue to free the | |
1522 | space, with no leaks. However, note that this case is actually | |
1523 | fairly rare. | |
1524 | ||
1525 | Typically pools either (a) fail completely (but perhaps temporarily, | |
1526 | e.g. a top-level vdev going offline), or (b) have localized, | |
1527 | permanent errors (e.g. disk returns the wrong data due to bit flip or | |
1528 | firmware bug). In case (a), this setting does not matter because the | |
1529 | pool will be suspended and the sync thread will not be able to make | |
1530 | forward progress regardless. In case (b), because the error is | |
1531 | permanent, the best we can do is leak the minimum amount of space, | |
1532 | which is what setting this flag will do. Therefore, it is reasonable | |
1533 | for this flag to normally be set, but we chose the more conservative | |
1534 | approach of not setting it, so that there is no possibility of | |
1535 | leaking space in the "partial temporary" failure case. | |
1536 | .sp | |
1537 | Default value: \fB0\fR. | |
1538 | .RE | |
1539 | ||
29714574 TF |
1540 | .sp |
1541 | .ne 2 | |
1542 | .na | |
1543 | \fBzfs_free_min_time_ms\fR (int) | |
1544 | .ad | |
1545 | .RS 12n | |
6146e17e | 1546 | During a \fBzfs destroy\fR operation using \fBfeature@async_destroy\fR a minimum |
83426735 | 1547 | of this much time will be spent working on freeing blocks per txg. |
29714574 TF |
1548 | .sp |
1549 | Default value: \fB1,000\fR. | |
1550 | .RE | |
1551 | ||
1552 | .sp | |
1553 | .ne 2 | |
1554 | .na | |
1555 | \fBzfs_immediate_write_sz\fR (long) | |
1556 | .ad | |
1557 | .RS 12n | |
83426735 | 1558 | Largest data block to write to zil. Larger blocks will be treated as if the |
6146e17e | 1559 | dataset being written to had the property setting \fBlogbias=throughput\fR. |
29714574 TF |
1560 | .sp |
1561 | Default value: \fB32,768\fR. | |
1562 | .RE | |
1563 | ||
917f475f JG |
1564 | .sp |
1565 | .ne 2 | |
1566 | .na | |
1567 | \fBzfs_lua_max_instrlimit\fR (ulong) | |
1568 | .ad | |
1569 | .RS 12n | |
1570 | The maximum execution time limit that can be set for a ZFS channel program, | |
1571 | specified as a number of Lua instructions. | |
1572 | .sp | |
1573 | Default value: \fB100,000,000\fR. | |
1574 | .RE | |
1575 | ||
1576 | .sp | |
1577 | .ne 2 | |
1578 | .na | |
1579 | \fBzfs_lua_max_memlimit\fR (ulong) | |
1580 | .ad | |
1581 | .RS 12n | |
1582 | The maximum memory limit that can be set for a ZFS channel program, specified | |
1583 | in bytes. | |
1584 | .sp | |
1585 | Default value: \fB104,857,600\fR. | |
1586 | .RE | |
1587 | ||
f1512ee6 MA |
1588 | .sp |
1589 | .ne 2 | |
1590 | .na | |
1591 | \fBzfs_max_recordsize\fR (int) | |
1592 | .ad | |
1593 | .RS 12n | |
1594 | We currently support block sizes from 512 bytes to 16MB. The benefits of | |
1595 | larger blocks, and thus larger IO, need to be weighed against the cost of | |
1596 | COWing a giant block to modify one byte. Additionally, very large blocks | |
1597 | can have an impact on i/o latency, and also potentially on the memory | |
1598 | allocator. Therefore, we do not allow the recordsize to be set larger than | |
1599 | zfs_max_recordsize (default 1MB). Larger blocks can be created by changing | |
1600 | this tunable, and pools with larger blocks can always be imported and used, | |
1601 | regardless of this setting. | |
1602 | .sp | |
1603 | Default value: \fB1,048,576\fR. | |
1604 | .RE | |
1605 | ||
f3a7f661 GW |
1606 | .sp |
1607 | .ne 2 | |
1608 | .na | |
1609 | \fBzfs_metaslab_fragmentation_threshold\fR (int) | |
1610 | .ad | |
1611 | .RS 12n | |
1612 | Allow metaslabs to keep their active state as long as their fragmentation | |
1613 | percentage is less than or equal to this value. An active metaslab that | |
1614 | exceeds this threshold will no longer keep its active status allowing | |
1615 | better metaslabs to be selected. | |
1616 | .sp | |
1617 | Default value: \fB70\fR. | |
1618 | .RE | |
1619 | ||
1620 | .sp | |
1621 | .ne 2 | |
1622 | .na | |
1623 | \fBzfs_mg_fragmentation_threshold\fR (int) | |
1624 | .ad | |
1625 | .RS 12n | |
1626 | Metaslab groups are considered eligible for allocations if their | |
83426735 | 1627 | fragmentation metric (measured as a percentage) is less than or equal to |
f3a7f661 GW |
1628 | this value. If a metaslab group exceeds this threshold then it will be |
1629 | skipped unless all metaslab groups within the metaslab class have also | |
1630 | crossed this threshold. | |
1631 | .sp | |
1632 | Default value: \fB85\fR. | |
1633 | .RE | |
1634 | ||
f4a4046b TC |
1635 | .sp |
1636 | .ne 2 | |
1637 | .na | |
1638 | \fBzfs_mg_noalloc_threshold\fR (int) | |
1639 | .ad | |
1640 | .RS 12n | |
1641 | Defines a threshold at which metaslab groups should be eligible for | |
1642 | allocations. The value is expressed as a percentage of free space | |
1643 | beyond which a metaslab group is always eligible for allocations. | |
1644 | If a metaslab group's free space is less than or equal to the | |
6b4e21c6 | 1645 | threshold, the allocator will avoid allocating to that group |
f4a4046b TC |
1646 | unless all groups in the pool have reached the threshold. Once all |
1647 | groups have reached the threshold, all groups are allowed to accept | |
1648 | allocations. The default value of 0 disables the feature and causes | |
1649 | all metaslab groups to be eligible for allocations. | |
1650 | ||
b58237e7 | 1651 | This parameter allows one to deal with pools having heavily imbalanced |
f4a4046b TC |
1652 | vdevs such as would be the case when a new vdev has been added. |
1653 | Setting the threshold to a non-zero percentage will stop allocations | |
1654 | from being made to vdevs that aren't filled to the specified percentage | |
1655 | and allow lesser filled vdevs to acquire more allocations than they | |
1656 | otherwise would under the old \fBzfs_mg_alloc_failures\fR facility. | |
1657 | .sp | |
1658 | Default value: \fB0\fR. | |
1659 | .RE | |
1660 | ||
379ca9cf OF |
1661 | .sp |
1662 | .ne 2 | |
1663 | .na | |
1664 | \fBzfs_multihost_history\fR (int) | |
1665 | .ad | |
1666 | .RS 12n | |
1667 | Historical statistics for the last N multihost updates will be available in | |
1668 | \fB/proc/spl/kstat/zfs/<pool>/multihost\fR | |
1669 | .sp | |
1670 | Default value: \fB0\fR. | |
1671 | .RE | |
1672 | ||
1673 | .sp | |
1674 | .ne 2 | |
1675 | .na | |
1676 | \fBzfs_multihost_interval\fR (ulong) | |
1677 | .ad | |
1678 | .RS 12n | |
1679 | Used to control the frequency of multihost writes which are performed when the | |
1680 | \fBmultihost\fR pool property is on. This is one factor used to determine | |
1681 | the length of the activity check during import. | |
1682 | .sp | |
1683 | The multihost write period is \fBzfs_multihost_interval / leaf-vdevs\fR milliseconds. | |
1684 | This means that on average a multihost write will be issued for each leaf vdev every | |
1685 | \fBzfs_multihost_interval\fR milliseconds. In practice, the observed period can | |
1686 | vary with the I/O load and this observed value is the delay which is stored in | |
1687 | the uberblock. | |
1688 | .sp | |
1689 | On import the activity check waits a minimum amount of time determined by | |
1690 | \fBzfs_multihost_interval * zfs_multihost_import_intervals\fR. The activity | |
1691 | check time may be further extended if the value of mmp delay found in the best | |
1692 | uberblock indicates actual multihost updates happened at longer intervals than | |
1693 | \fBzfs_multihost_interval\fR. A minimum value of \fB100ms\fR is enforced. | |
1694 | .sp | |
1695 | Default value: \fB1000\fR. | |
1696 | .RE | |
1697 | ||
1698 | .sp | |
1699 | .ne 2 | |
1700 | .na | |
1701 | \fBzfs_multihost_import_intervals\fR (uint) | |
1702 | .ad | |
1703 | .RS 12n | |
1704 | Used to control the duration of the activity test on import. Smaller values of | |
1705 | \fBzfs_multihost_import_intervals\fR will reduce the import time but increase | |
1706 | the risk of failing to detect an active pool. The total activity check time is | |
1707 | never allowed to drop below one second. A value of 0 is ignored and treated as | |
1708 | if it was set to 1 | |
1709 | .sp | |
1710 | Default value: \fB10\fR. | |
1711 | .RE | |
1712 | ||
1713 | .sp | |
1714 | .ne 2 | |
1715 | .na | |
1716 | \fBzfs_multihost_fail_intervals\fR (uint) | |
1717 | .ad | |
1718 | .RS 12n | |
1719 | Controls the behavior of the pool when multihost write failures are detected. | |
1720 | .sp | |
1721 | When \fBzfs_multihost_fail_intervals = 0\fR then multihost write failures are ignored. | |
1722 | The failures will still be reported to the ZED which depending on its | |
1723 | configuration may take action such as suspending the pool or offlining a device. | |
1724 | .sp | |
1725 | When \fBzfs_multihost_fail_intervals > 0\fR then sequential multihost write failures | |
1726 | will cause the pool to be suspended. This occurs when | |
1727 | \fBzfs_multihost_fail_intervals * zfs_multihost_interval\fR milliseconds have | |
1728 | passed since the last successful multihost write. This guarantees the activity test | |
1729 | will see multihost writes if the pool is imported. | |
1730 | .sp | |
1731 | Default value: \fB5\fR. | |
1732 | .RE | |
1733 | ||
29714574 TF |
1734 | .sp |
1735 | .ne 2 | |
1736 | .na | |
1737 | \fBzfs_no_scrub_io\fR (int) | |
1738 | .ad | |
1739 | .RS 12n | |
83426735 D |
1740 | Set for no scrub I/O. This results in scrubs not actually scrubbing data and |
1741 | simply doing a metadata crawl of the pool instead. | |
29714574 TF |
1742 | .sp |
1743 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1744 | .RE | |
1745 | ||
1746 | .sp | |
1747 | .ne 2 | |
1748 | .na | |
1749 | \fBzfs_no_scrub_prefetch\fR (int) | |
1750 | .ad | |
1751 | .RS 12n | |
83426735 | 1752 | Set to disable block prefetching for scrubs. |
29714574 TF |
1753 | .sp |
1754 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1755 | .RE | |
1756 | ||
29714574 TF |
1757 | .sp |
1758 | .ne 2 | |
1759 | .na | |
1760 | \fBzfs_nocacheflush\fR (int) | |
1761 | .ad | |
1762 | .RS 12n | |
83426735 D |
1763 | Disable cache flush operations on disks when writing. Beware, this may cause |
1764 | corruption if disks re-order writes. | |
29714574 TF |
1765 | .sp |
1766 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1767 | .RE | |
1768 | ||
1769 | .sp | |
1770 | .ne 2 | |
1771 | .na | |
1772 | \fBzfs_nopwrite_enabled\fR (int) | |
1773 | .ad | |
1774 | .RS 12n | |
1775 | Enable NOP writes | |
1776 | .sp | |
1777 | Use \fB1\fR for yes (default) and \fB0\fR to disable. | |
1778 | .RE | |
1779 | ||
66aca247 DB |
1780 | .sp |
1781 | .ne 2 | |
1782 | .na | |
1783 | \fBzfs_dmu_offset_next_sync\fR (int) | |
1784 | .ad | |
1785 | .RS 12n | |
1786 | Enable forcing txg sync to find holes. When enabled forces ZFS to act | |
1787 | like prior versions when SEEK_HOLE or SEEK_DATA flags are used, which | |
1788 | when a dnode is dirty causes txg's to be synced so that this data can be | |
1789 | found. | |
1790 | .sp | |
1791 | Use \fB1\fR for yes and \fB0\fR to disable (default). | |
1792 | .RE | |
1793 | ||
29714574 TF |
1794 | .sp |
1795 | .ne 2 | |
1796 | .na | |
b738bc5a | 1797 | \fBzfs_pd_bytes_max\fR (int) |
29714574 TF |
1798 | .ad |
1799 | .RS 12n | |
83426735 | 1800 | The number of bytes which should be prefetched during a pool traversal |
6146e17e | 1801 | (eg: \fBzfs send\fR or other data crawling operations) |
29714574 | 1802 | .sp |
74aa2ba2 | 1803 | Default value: \fB52,428,800\fR. |
29714574 TF |
1804 | .RE |
1805 | ||
bef78122 DQ |
1806 | .sp |
1807 | .ne 2 | |
1808 | .na | |
1809 | \fBzfs_per_txg_dirty_frees_percent \fR (ulong) | |
1810 | .ad | |
1811 | .RS 12n | |
1812 | Tunable to control percentage of dirtied blocks from frees in one TXG. | |
1813 | After this threshold is crossed, additional dirty blocks from frees | |
1814 | wait until the next TXG. | |
1815 | A value of zero will disable this throttle. | |
1816 | .sp | |
1817 | Default value: \fB30\fR and \fB0\fR to disable. | |
1818 | .RE | |
1819 | ||
1820 | ||
1821 | ||
29714574 TF |
1822 | .sp |
1823 | .ne 2 | |
1824 | .na | |
1825 | \fBzfs_prefetch_disable\fR (int) | |
1826 | .ad | |
1827 | .RS 12n | |
7f60329a MA |
1828 | This tunable disables predictive prefetch. Note that it leaves "prescient" |
1829 | prefetch (e.g. prefetch for zfs send) intact. Unlike predictive prefetch, | |
1830 | prescient prefetch never issues i/os that end up not being needed, so it | |
1831 | can't hurt performance. | |
29714574 TF |
1832 | .sp |
1833 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1834 | .RE | |
1835 | ||
1836 | .sp | |
1837 | .ne 2 | |
1838 | .na | |
1839 | \fBzfs_read_chunk_size\fR (long) | |
1840 | .ad | |
1841 | .RS 12n | |
1842 | Bytes to read per chunk | |
1843 | .sp | |
1844 | Default value: \fB1,048,576\fR. | |
1845 | .RE | |
1846 | ||
1847 | .sp | |
1848 | .ne 2 | |
1849 | .na | |
1850 | \fBzfs_read_history\fR (int) | |
1851 | .ad | |
1852 | .RS 12n | |
379ca9cf OF |
1853 | Historical statistics for the last N reads will be available in |
1854 | \fB/proc/spl/kstat/zfs/<pool>/reads\fR | |
29714574 | 1855 | .sp |
83426735 | 1856 | Default value: \fB0\fR (no data is kept). |
29714574 TF |
1857 | .RE |
1858 | ||
1859 | .sp | |
1860 | .ne 2 | |
1861 | .na | |
1862 | \fBzfs_read_history_hits\fR (int) | |
1863 | .ad | |
1864 | .RS 12n | |
1865 | Include cache hits in read history | |
1866 | .sp | |
1867 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1868 | .RE | |
1869 | ||
9e052db4 MA |
1870 | .sp |
1871 | .ne 2 | |
1872 | .na | |
4589f3ae BB |
1873 | \fBzfs_reconstruct_indirect_combinations_max\fR (int) |
1874 | .ad | |
1875 | .RS 12na | |
1876 | If an indirect split block contains more than this many possible unique | |
1877 | combinations when being reconstructed, consider it too computationally | |
1878 | expensive to check them all. Instead, try at most | |
1879 | \fBzfs_reconstruct_indirect_combinations_max\fR randomly-selected | |
1880 | combinations each time the block is accessed. This allows all segment | |
1881 | copies to participate fairly in the reconstruction when all combinations | |
1882 | cannot be checked and prevents repeated use of one bad copy. | |
1883 | .sp | |
1884 | Default value: \fB100\fR. | |
9e052db4 MA |
1885 | .RE |
1886 | ||
29714574 TF |
1887 | .sp |
1888 | .ne 2 | |
1889 | .na | |
1890 | \fBzfs_recover\fR (int) | |
1891 | .ad | |
1892 | .RS 12n | |
1893 | Set to attempt to recover from fatal errors. This should only be used as a | |
1894 | last resort, as it typically results in leaked space, or worse. | |
1895 | .sp | |
1896 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
1897 | .RE | |
1898 | ||
1899 | .sp | |
1900 | .ne 2 | |
1901 | .na | |
d4a72f23 | 1902 | \fBzfs_resilver_min_time_ms\fR (int) |
29714574 TF |
1903 | .ad |
1904 | .RS 12n | |
d4a72f23 TC |
1905 | Resilvers are processed by the sync thread. While resilvering it will spend |
1906 | at least this much time working on a resilver between txg flushes. | |
29714574 | 1907 | .sp |
d4a72f23 | 1908 | Default value: \fB3,000\fR. |
29714574 TF |
1909 | .RE |
1910 | ||
02638a30 TC |
1911 | .sp |
1912 | .ne 2 | |
1913 | .na | |
1914 | \fBzfs_scan_ignore_errors\fR (int) | |
1915 | .ad | |
1916 | .RS 12n | |
1917 | If set to a nonzero value, remove the DTL (dirty time list) upon | |
1918 | completion of a pool scan (scrub) even if there were unrepairable | |
1919 | errors. It is intended to be used during pool repair or recovery to | |
1920 | stop resilvering when the pool is next imported. | |
1921 | .sp | |
1922 | Default value: \fB0\fR. | |
1923 | .RE | |
1924 | ||
29714574 TF |
1925 | .sp |
1926 | .ne 2 | |
1927 | .na | |
d4a72f23 | 1928 | \fBzfs_scrub_min_time_ms\fR (int) |
29714574 TF |
1929 | .ad |
1930 | .RS 12n | |
d4a72f23 TC |
1931 | Scrubs are processed by the sync thread. While scrubbing it will spend |
1932 | at least this much time working on a scrub between txg flushes. | |
29714574 | 1933 | .sp |
d4a72f23 | 1934 | Default value: \fB1,000\fR. |
29714574 TF |
1935 | .RE |
1936 | ||
1937 | .sp | |
1938 | .ne 2 | |
1939 | .na | |
d4a72f23 | 1940 | \fBzfs_scan_checkpoint_intval\fR (int) |
29714574 TF |
1941 | .ad |
1942 | .RS 12n | |
d4a72f23 TC |
1943 | To preserve progress across reboots the sequential scan algorithm periodically |
1944 | needs to stop metadata scanning and issue all the verifications I/Os to disk. | |
1945 | The frequency of this flushing is determined by the | |
1946 | \fBfBzfs_scan_checkpoint_intval\fR tunable. | |
29714574 | 1947 | .sp |
d4a72f23 | 1948 | Default value: \fB7200\fR seconds (every 2 hours). |
29714574 TF |
1949 | .RE |
1950 | ||
1951 | .sp | |
1952 | .ne 2 | |
1953 | .na | |
d4a72f23 | 1954 | \fBzfs_scan_fill_weight\fR (int) |
29714574 TF |
1955 | .ad |
1956 | .RS 12n | |
d4a72f23 TC |
1957 | This tunable affects how scrub and resilver I/O segments are ordered. A higher |
1958 | number indicates that we care more about how filled in a segment is, while a | |
1959 | lower number indicates we care more about the size of the extent without | |
1960 | considering the gaps within a segment. This value is only tunable upon module | |
1961 | insertion. Changing the value afterwards will have no affect on scrub or | |
1962 | resilver performance. | |
29714574 | 1963 | .sp |
d4a72f23 | 1964 | Default value: \fB3\fR. |
29714574 TF |
1965 | .RE |
1966 | ||
1967 | .sp | |
1968 | .ne 2 | |
1969 | .na | |
d4a72f23 | 1970 | \fBzfs_scan_issue_strategy\fR (int) |
29714574 TF |
1971 | .ad |
1972 | .RS 12n | |
d4a72f23 TC |
1973 | Determines the order that data will be verified while scrubbing or resilvering. |
1974 | If set to \fB1\fR, data will be verified as sequentially as possible, given the | |
1975 | amount of memory reserved for scrubbing (see \fBzfs_scan_mem_lim_fact\fR). This | |
1976 | may improve scrub performance if the pool's data is very fragmented. If set to | |
1977 | \fB2\fR, the largest mostly-contiguous chunk of found data will be verified | |
1978 | first. By deferring scrubbing of small segments, we may later find adjacent data | |
1979 | to coalesce and increase the segment size. If set to \fB0\fR, zfs will use | |
1980 | strategy \fB1\fR during normal verification and strategy \fB2\fR while taking a | |
1981 | checkpoint. | |
29714574 | 1982 | .sp |
d4a72f23 TC |
1983 | Default value: \fB0\fR. |
1984 | .RE | |
1985 | ||
1986 | .sp | |
1987 | .ne 2 | |
1988 | .na | |
1989 | \fBzfs_scan_legacy\fR (int) | |
1990 | .ad | |
1991 | .RS 12n | |
1992 | A value of 0 indicates that scrubs and resilvers will gather metadata in | |
1993 | memory before issuing sequential I/O. A value of 1 indicates that the legacy | |
1994 | algorithm will be used where I/O is initiated as soon as it is discovered. | |
1995 | Changing this value to 0 will not affect scrubs or resilvers that are already | |
1996 | in progress. | |
1997 | .sp | |
1998 | Default value: \fB0\fR. | |
1999 | .RE | |
2000 | ||
2001 | .sp | |
2002 | .ne 2 | |
2003 | .na | |
2004 | \fBzfs_scan_max_ext_gap\fR (int) | |
2005 | .ad | |
2006 | .RS 12n | |
2007 | Indicates the largest gap in bytes between scrub / resilver I/Os that will still | |
2008 | be considered sequential for sorting purposes. Changing this value will not | |
2009 | affect scrubs or resilvers that are already in progress. | |
2010 | .sp | |
2011 | Default value: \fB2097152 (2 MB)\fR. | |
2012 | .RE | |
2013 | ||
2014 | .sp | |
2015 | .ne 2 | |
2016 | .na | |
2017 | \fBzfs_scan_mem_lim_fact\fR (int) | |
2018 | .ad | |
2019 | .RS 12n | |
2020 | Maximum fraction of RAM used for I/O sorting by sequential scan algorithm. | |
2021 | This tunable determines the hard limit for I/O sorting memory usage. | |
2022 | When the hard limit is reached we stop scanning metadata and start issuing | |
2023 | data verification I/O. This is done until we get below the soft limit. | |
2024 | .sp | |
2025 | Default value: \fB20\fR which is 5% of RAM (1/20). | |
2026 | .RE | |
2027 | ||
2028 | .sp | |
2029 | .ne 2 | |
2030 | .na | |
2031 | \fBzfs_scan_mem_lim_soft_fact\fR (int) | |
2032 | .ad | |
2033 | .RS 12n | |
2034 | The fraction of the hard limit used to determined the soft limit for I/O sorting | |
2035 | by the sequential scan algorithm. When we cross this limit from bellow no action | |
2036 | is taken. When we cross this limit from above it is because we are issuing | |
2037 | verification I/O. In this case (unless the metadata scan is done) we stop | |
2038 | issuing verification I/O and start scanning metadata again until we get to the | |
2039 | hard limit. | |
2040 | .sp | |
2041 | Default value: \fB20\fR which is 5% of the hard limit (1/20). | |
2042 | .RE | |
2043 | ||
2044 | .sp | |
2045 | .ne 2 | |
2046 | .na | |
2047 | \fBzfs_scan_vdev_limit\fR (int) | |
2048 | .ad | |
2049 | .RS 12n | |
2050 | Maximum amount of data that can be concurrently issued at once for scrubs and | |
2051 | resilvers per leaf device, given in bytes. | |
2052 | .sp | |
2053 | Default value: \fB41943040\fR. | |
29714574 TF |
2054 | .RE |
2055 | ||
fd8febbd TF |
2056 | .sp |
2057 | .ne 2 | |
2058 | .na | |
2059 | \fBzfs_send_corrupt_data\fR (int) | |
2060 | .ad | |
2061 | .RS 12n | |
83426735 | 2062 | Allow sending of corrupt data (ignore read/checksum errors when sending data) |
fd8febbd TF |
2063 | .sp |
2064 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2065 | .RE | |
2066 | ||
3b0d9928 BB |
2067 | .sp |
2068 | .ne 2 | |
2069 | .na | |
2070 | \fBzfs_send_queue_length\fR (int) | |
2071 | .ad | |
2072 | .RS 12n | |
2073 | The maximum number of bytes allowed in the \fBzfs send\fR queue. This value | |
2074 | must be at least twice the maximum block size in use. | |
2075 | .sp | |
2076 | Default value: \fB16,777,216\fR. | |
2077 | .RE | |
2078 | ||
2079 | .sp | |
2080 | .ne 2 | |
2081 | .na | |
2082 | \fBzfs_recv_queue_length\fR (int) | |
2083 | .ad | |
2084 | .RS 12n | |
2085 | .sp | |
2086 | The maximum number of bytes allowed in the \fBzfs receive\fR queue. This value | |
2087 | must be at least twice the maximum block size in use. | |
2088 | .sp | |
2089 | Default value: \fB16,777,216\fR. | |
2090 | .RE | |
2091 | ||
29714574 TF |
2092 | .sp |
2093 | .ne 2 | |
2094 | .na | |
2095 | \fBzfs_sync_pass_deferred_free\fR (int) | |
2096 | .ad | |
2097 | .RS 12n | |
83426735 | 2098 | Flushing of data to disk is done in passes. Defer frees starting in this pass |
29714574 TF |
2099 | .sp |
2100 | Default value: \fB2\fR. | |
2101 | .RE | |
2102 | ||
2103 | .sp | |
2104 | .ne 2 | |
2105 | .na | |
2106 | \fBzfs_sync_pass_dont_compress\fR (int) | |
2107 | .ad | |
2108 | .RS 12n | |
2109 | Don't compress starting in this pass | |
2110 | .sp | |
2111 | Default value: \fB5\fR. | |
2112 | .RE | |
2113 | ||
2114 | .sp | |
2115 | .ne 2 | |
2116 | .na | |
2117 | \fBzfs_sync_pass_rewrite\fR (int) | |
2118 | .ad | |
2119 | .RS 12n | |
83426735 | 2120 | Rewrite new block pointers starting in this pass |
29714574 TF |
2121 | .sp |
2122 | Default value: \fB2\fR. | |
2123 | .RE | |
2124 | ||
a032ac4b BB |
2125 | .sp |
2126 | .ne 2 | |
2127 | .na | |
2128 | \fBzfs_sync_taskq_batch_pct\fR (int) | |
2129 | .ad | |
2130 | .RS 12n | |
2131 | This controls the number of threads used by the dp_sync_taskq. The default | |
2132 | value of 75% will create a maximum of one thread per cpu. | |
2133 | .sp | |
be54a13c | 2134 | Default value: \fB75\fR%. |
a032ac4b BB |
2135 | .RE |
2136 | ||
29714574 TF |
2137 | .sp |
2138 | .ne 2 | |
2139 | .na | |
2140 | \fBzfs_txg_history\fR (int) | |
2141 | .ad | |
2142 | .RS 12n | |
379ca9cf OF |
2143 | Historical statistics for the last N txgs will be available in |
2144 | \fB/proc/spl/kstat/zfs/<pool>/txgs\fR | |
29714574 | 2145 | .sp |
ca85d690 | 2146 | Default value: \fB0\fR. |
29714574 TF |
2147 | .RE |
2148 | ||
29714574 TF |
2149 | .sp |
2150 | .ne 2 | |
2151 | .na | |
2152 | \fBzfs_txg_timeout\fR (int) | |
2153 | .ad | |
2154 | .RS 12n | |
83426735 | 2155 | Flush dirty data to disk at least every N seconds (maximum txg duration) |
29714574 TF |
2156 | .sp |
2157 | Default value: \fB5\fR. | |
2158 | .RE | |
2159 | ||
2160 | .sp | |
2161 | .ne 2 | |
2162 | .na | |
2163 | \fBzfs_vdev_aggregation_limit\fR (int) | |
2164 | .ad | |
2165 | .RS 12n | |
2166 | Max vdev I/O aggregation size | |
2167 | .sp | |
2168 | Default value: \fB131,072\fR. | |
2169 | .RE | |
2170 | ||
2171 | .sp | |
2172 | .ne 2 | |
2173 | .na | |
2174 | \fBzfs_vdev_cache_bshift\fR (int) | |
2175 | .ad | |
2176 | .RS 12n | |
2177 | Shift size to inflate reads too | |
2178 | .sp | |
83426735 | 2179 | Default value: \fB16\fR (effectively 65536). |
29714574 TF |
2180 | .RE |
2181 | ||
2182 | .sp | |
2183 | .ne 2 | |
2184 | .na | |
2185 | \fBzfs_vdev_cache_max\fR (int) | |
2186 | .ad | |
2187 | .RS 12n | |
ca85d690 | 2188 | Inflate reads smaller than this value to meet the \fBzfs_vdev_cache_bshift\fR |
2189 | size (default 64k). | |
83426735 D |
2190 | .sp |
2191 | Default value: \fB16384\fR. | |
29714574 TF |
2192 | .RE |
2193 | ||
2194 | .sp | |
2195 | .ne 2 | |
2196 | .na | |
2197 | \fBzfs_vdev_cache_size\fR (int) | |
2198 | .ad | |
2199 | .RS 12n | |
83426735 D |
2200 | Total size of the per-disk cache in bytes. |
2201 | .sp | |
2202 | Currently this feature is disabled as it has been found to not be helpful | |
2203 | for performance and in some cases harmful. | |
29714574 TF |
2204 | .sp |
2205 | Default value: \fB0\fR. | |
2206 | .RE | |
2207 | ||
29714574 TF |
2208 | .sp |
2209 | .ne 2 | |
2210 | .na | |
9f500936 | 2211 | \fBzfs_vdev_mirror_rotating_inc\fR (int) |
29714574 TF |
2212 | .ad |
2213 | .RS 12n | |
9f500936 | 2214 | A number by which the balancing algorithm increments the load calculation for |
2215 | the purpose of selecting the least busy mirror member when an I/O immediately | |
2216 | follows its predecessor on rotational vdevs for the purpose of making decisions | |
2217 | based on load. | |
29714574 | 2218 | .sp |
9f500936 | 2219 | Default value: \fB0\fR. |
2220 | .RE | |
2221 | ||
2222 | .sp | |
2223 | .ne 2 | |
2224 | .na | |
2225 | \fBzfs_vdev_mirror_rotating_seek_inc\fR (int) | |
2226 | .ad | |
2227 | .RS 12n | |
2228 | A number by which the balancing algorithm increments the load calculation for | |
2229 | the purpose of selecting the least busy mirror member when an I/O lacks | |
2230 | locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within | |
2231 | this that are not immediately following the previous I/O are incremented by | |
2232 | half. | |
2233 | .sp | |
2234 | Default value: \fB5\fR. | |
2235 | .RE | |
2236 | ||
2237 | .sp | |
2238 | .ne 2 | |
2239 | .na | |
2240 | \fBzfs_vdev_mirror_rotating_seek_offset\fR (int) | |
2241 | .ad | |
2242 | .RS 12n | |
2243 | The maximum distance for the last queued I/O in which the balancing algorithm | |
2244 | considers an I/O to have locality. | |
2245 | See the section "ZFS I/O SCHEDULER". | |
2246 | .sp | |
2247 | Default value: \fB1048576\fR. | |
2248 | .RE | |
2249 | ||
2250 | .sp | |
2251 | .ne 2 | |
2252 | .na | |
2253 | \fBzfs_vdev_mirror_non_rotating_inc\fR (int) | |
2254 | .ad | |
2255 | .RS 12n | |
2256 | A number by which the balancing algorithm increments the load calculation for | |
2257 | the purpose of selecting the least busy mirror member on non-rotational vdevs | |
2258 | when I/Os do not immediately follow one another. | |
2259 | .sp | |
2260 | Default value: \fB0\fR. | |
2261 | .RE | |
2262 | ||
2263 | .sp | |
2264 | .ne 2 | |
2265 | .na | |
2266 | \fBzfs_vdev_mirror_non_rotating_seek_inc\fR (int) | |
2267 | .ad | |
2268 | .RS 12n | |
2269 | A number by which the balancing algorithm increments the load calculation for | |
2270 | the purpose of selecting the least busy mirror member when an I/O lacks | |
2271 | locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within | |
2272 | this that are not immediately following the previous I/O are incremented by | |
2273 | half. | |
2274 | .sp | |
2275 | Default value: \fB1\fR. | |
29714574 TF |
2276 | .RE |
2277 | ||
29714574 TF |
2278 | .sp |
2279 | .ne 2 | |
2280 | .na | |
2281 | \fBzfs_vdev_read_gap_limit\fR (int) | |
2282 | .ad | |
2283 | .RS 12n | |
83426735 D |
2284 | Aggregate read I/O operations if the gap on-disk between them is within this |
2285 | threshold. | |
29714574 TF |
2286 | .sp |
2287 | Default value: \fB32,768\fR. | |
2288 | .RE | |
2289 | ||
2290 | .sp | |
2291 | .ne 2 | |
2292 | .na | |
2293 | \fBzfs_vdev_scheduler\fR (charp) | |
2294 | .ad | |
2295 | .RS 12n | |
ca85d690 | 2296 | Set the Linux I/O scheduler on whole disk vdevs to this scheduler. Valid options |
2297 | are noop, cfq, bfq & deadline | |
29714574 TF |
2298 | .sp |
2299 | Default value: \fBnoop\fR. | |
2300 | .RE | |
2301 | ||
29714574 TF |
2302 | .sp |
2303 | .ne 2 | |
2304 | .na | |
2305 | \fBzfs_vdev_write_gap_limit\fR (int) | |
2306 | .ad | |
2307 | .RS 12n | |
2308 | Aggregate write I/O over gap | |
2309 | .sp | |
2310 | Default value: \fB4,096\fR. | |
2311 | .RE | |
2312 | ||
ab9f4b0b GN |
2313 | .sp |
2314 | .ne 2 | |
2315 | .na | |
2316 | \fBzfs_vdev_raidz_impl\fR (string) | |
2317 | .ad | |
2318 | .RS 12n | |
c9187d86 | 2319 | Parameter for selecting raidz parity implementation to use. |
ab9f4b0b GN |
2320 | |
2321 | Options marked (always) below may be selected on module load as they are | |
2322 | supported on all systems. | |
2323 | The remaining options may only be set after the module is loaded, as they | |
2324 | are available only if the implementations are compiled in and supported | |
2325 | on the running system. | |
2326 | ||
2327 | Once the module is loaded, the content of | |
2328 | /sys/module/zfs/parameters/zfs_vdev_raidz_impl will show available options | |
2329 | with the currently selected one enclosed in []. | |
2330 | Possible options are: | |
2331 | fastest - (always) implementation selected using built-in benchmark | |
2332 | original - (always) original raidz implementation | |
2333 | scalar - (always) scalar raidz implementation | |
ae25d222 GN |
2334 | sse2 - implementation using SSE2 instruction set (64bit x86 only) |
2335 | ssse3 - implementation using SSSE3 instruction set (64bit x86 only) | |
ab9f4b0b | 2336 | avx2 - implementation using AVX2 instruction set (64bit x86 only) |
7f547f85 RD |
2337 | avx512f - implementation using AVX512F instruction set (64bit x86 only) |
2338 | avx512bw - implementation using AVX512F & AVX512BW instruction sets (64bit x86 only) | |
62a65a65 RD |
2339 | aarch64_neon - implementation using NEON (Aarch64/64 bit ARMv8 only) |
2340 | aarch64_neonx2 - implementation using NEON with more unrolling (Aarch64/64 bit ARMv8 only) | |
ab9f4b0b GN |
2341 | .sp |
2342 | Default value: \fBfastest\fR. | |
2343 | .RE | |
2344 | ||
29714574 TF |
2345 | .sp |
2346 | .ne 2 | |
2347 | .na | |
2348 | \fBzfs_zevent_cols\fR (int) | |
2349 | .ad | |
2350 | .RS 12n | |
83426735 | 2351 | When zevents are logged to the console use this as the word wrap width. |
29714574 TF |
2352 | .sp |
2353 | Default value: \fB80\fR. | |
2354 | .RE | |
2355 | ||
2356 | .sp | |
2357 | .ne 2 | |
2358 | .na | |
2359 | \fBzfs_zevent_console\fR (int) | |
2360 | .ad | |
2361 | .RS 12n | |
2362 | Log events to the console | |
2363 | .sp | |
2364 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2365 | .RE | |
2366 | ||
2367 | .sp | |
2368 | .ne 2 | |
2369 | .na | |
2370 | \fBzfs_zevent_len_max\fR (int) | |
2371 | .ad | |
2372 | .RS 12n | |
83426735 D |
2373 | Max event queue length. A value of 0 will result in a calculated value which |
2374 | increases with the number of CPUs in the system (minimum 64 events). Events | |
2375 | in the queue can be viewed with the \fBzpool events\fR command. | |
29714574 TF |
2376 | .sp |
2377 | Default value: \fB0\fR. | |
2378 | .RE | |
2379 | ||
a032ac4b BB |
2380 | .sp |
2381 | .ne 2 | |
2382 | .na | |
2383 | \fBzfs_zil_clean_taskq_maxalloc\fR (int) | |
2384 | .ad | |
2385 | .RS 12n | |
2386 | The maximum number of taskq entries that are allowed to be cached. When this | |
2fe61a7e | 2387 | limit is exceeded transaction records (itxs) will be cleaned synchronously. |
a032ac4b BB |
2388 | .sp |
2389 | Default value: \fB1048576\fR. | |
2390 | .RE | |
2391 | ||
2392 | .sp | |
2393 | .ne 2 | |
2394 | .na | |
2395 | \fBzfs_zil_clean_taskq_minalloc\fR (int) | |
2396 | .ad | |
2397 | .RS 12n | |
2398 | The number of taskq entries that are pre-populated when the taskq is first | |
2399 | created and are immediately available for use. | |
2400 | .sp | |
2401 | Default value: \fB1024\fR. | |
2402 | .RE | |
2403 | ||
2404 | .sp | |
2405 | .ne 2 | |
2406 | .na | |
2407 | \fBzfs_zil_clean_taskq_nthr_pct\fR (int) | |
2408 | .ad | |
2409 | .RS 12n | |
2410 | This controls the number of threads used by the dp_zil_clean_taskq. The default | |
2411 | value of 100% will create a maximum of one thread per cpu. | |
2412 | .sp | |
be54a13c | 2413 | Default value: \fB100\fR%. |
a032ac4b BB |
2414 | .RE |
2415 | ||
29714574 TF |
2416 | .sp |
2417 | .ne 2 | |
2418 | .na | |
2419 | \fBzil_replay_disable\fR (int) | |
2420 | .ad | |
2421 | .RS 12n | |
83426735 D |
2422 | Disable intent logging replay. Can be disabled for recovery from corrupted |
2423 | ZIL | |
29714574 TF |
2424 | .sp |
2425 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2426 | .RE | |
2427 | ||
2428 | .sp | |
2429 | .ne 2 | |
2430 | .na | |
1b7c1e5c | 2431 | \fBzil_slog_bulk\fR (ulong) |
29714574 TF |
2432 | .ad |
2433 | .RS 12n | |
1b7c1e5c GDN |
2434 | Limit SLOG write size per commit executed with synchronous priority. |
2435 | Any writes above that will be executed with lower (asynchronous) priority | |
2436 | to limit potential SLOG device abuse by single active ZIL writer. | |
29714574 | 2437 | .sp |
1b7c1e5c | 2438 | Default value: \fB786,432\fR. |
29714574 TF |
2439 | .RE |
2440 | ||
29714574 TF |
2441 | .sp |
2442 | .ne 2 | |
2443 | .na | |
2444 | \fBzio_delay_max\fR (int) | |
2445 | .ad | |
2446 | .RS 12n | |
83426735 | 2447 | A zevent will be logged if a ZIO operation takes more than N milliseconds to |
ab9f4b0b | 2448 | complete. Note that this is only a logging facility, not a timeout on |
83426735 | 2449 | operations. |
29714574 TF |
2450 | .sp |
2451 | Default value: \fB30,000\fR. | |
2452 | .RE | |
2453 | ||
3dfb57a3 DB |
2454 | .sp |
2455 | .ne 2 | |
2456 | .na | |
2457 | \fBzio_dva_throttle_enabled\fR (int) | |
2458 | .ad | |
2459 | .RS 12n | |
2460 | Throttle block allocations in the ZIO pipeline. This allows for | |
2461 | dynamic allocation distribution when devices are imbalanced. | |
e815485f TC |
2462 | When enabled, the maximum number of pending allocations per top-level vdev |
2463 | is limited by \fBzfs_vdev_queue_depth_pct\fR. | |
3dfb57a3 | 2464 | .sp |
27f2b90d | 2465 | Default value: \fB1\fR. |
3dfb57a3 DB |
2466 | .RE |
2467 | ||
29714574 TF |
2468 | .sp |
2469 | .ne 2 | |
2470 | .na | |
2471 | \fBzio_requeue_io_start_cut_in_line\fR (int) | |
2472 | .ad | |
2473 | .RS 12n | |
2474 | Prioritize requeued I/O | |
2475 | .sp | |
2476 | Default value: \fB0\fR. | |
2477 | .RE | |
2478 | ||
dcb6bed1 D |
2479 | .sp |
2480 | .ne 2 | |
2481 | .na | |
2482 | \fBzio_taskq_batch_pct\fR (uint) | |
2483 | .ad | |
2484 | .RS 12n | |
2485 | Percentage of online CPUs (or CPU cores, etc) which will run a worker thread | |
2486 | for IO. These workers are responsible for IO work such as compression and | |
2487 | checksum calculations. Fractional number of CPUs will be rounded down. | |
2488 | .sp | |
2489 | The default value of 75 was chosen to avoid using all CPUs which can result in | |
2490 | latency issues and inconsistent application performance, especially when high | |
2491 | compression is enabled. | |
2492 | .sp | |
2493 | Default value: \fB75\fR. | |
2494 | .RE | |
2495 | ||
29714574 TF |
2496 | .sp |
2497 | .ne 2 | |
2498 | .na | |
2499 | \fBzvol_inhibit_dev\fR (uint) | |
2500 | .ad | |
2501 | .RS 12n | |
83426735 D |
2502 | Do not create zvol device nodes. This may slightly improve startup time on |
2503 | systems with a very large number of zvols. | |
29714574 TF |
2504 | .sp |
2505 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2506 | .RE | |
2507 | ||
2508 | .sp | |
2509 | .ne 2 | |
2510 | .na | |
2511 | \fBzvol_major\fR (uint) | |
2512 | .ad | |
2513 | .RS 12n | |
83426735 | 2514 | Major number for zvol block devices |
29714574 TF |
2515 | .sp |
2516 | Default value: \fB230\fR. | |
2517 | .RE | |
2518 | ||
2519 | .sp | |
2520 | .ne 2 | |
2521 | .na | |
2522 | \fBzvol_max_discard_blocks\fR (ulong) | |
2523 | .ad | |
2524 | .RS 12n | |
83426735 D |
2525 | Discard (aka TRIM) operations done on zvols will be done in batches of this |
2526 | many blocks, where block size is determined by the \fBvolblocksize\fR property | |
2527 | of a zvol. | |
29714574 TF |
2528 | .sp |
2529 | Default value: \fB16,384\fR. | |
2530 | .RE | |
2531 | ||
9965059a BB |
2532 | .sp |
2533 | .ne 2 | |
2534 | .na | |
2535 | \fBzvol_prefetch_bytes\fR (uint) | |
2536 | .ad | |
2537 | .RS 12n | |
2538 | When adding a zvol to the system prefetch \fBzvol_prefetch_bytes\fR | |
2539 | from the start and end of the volume. Prefetching these regions | |
2540 | of the volume is desirable because they are likely to be accessed | |
2541 | immediately by \fBblkid(8)\fR or by the kernel scanning for a partition | |
2542 | table. | |
2543 | .sp | |
2544 | Default value: \fB131,072\fR. | |
2545 | .RE | |
2546 | ||
692e55b8 CC |
2547 | .sp |
2548 | .ne 2 | |
2549 | .na | |
2550 | \fBzvol_request_sync\fR (uint) | |
2551 | .ad | |
2552 | .RS 12n | |
2553 | When processing I/O requests for a zvol submit them synchronously. This | |
2554 | effectively limits the queue depth to 1 for each I/O submitter. When set | |
2555 | to 0 requests are handled asynchronously by a thread pool. The number of | |
2556 | requests which can be handled concurrently is controller by \fBzvol_threads\fR. | |
2557 | .sp | |
8fa5250f | 2558 | Default value: \fB0\fR. |
692e55b8 CC |
2559 | .RE |
2560 | ||
2561 | .sp | |
2562 | .ne 2 | |
2563 | .na | |
2564 | \fBzvol_threads\fR (uint) | |
2565 | .ad | |
2566 | .RS 12n | |
2567 | Max number of threads which can handle zvol I/O requests concurrently. | |
2568 | .sp | |
2569 | Default value: \fB32\fR. | |
2570 | .RE | |
2571 | ||
cf8738d8 | 2572 | .sp |
2573 | .ne 2 | |
2574 | .na | |
2575 | \fBzvol_volmode\fR (uint) | |
2576 | .ad | |
2577 | .RS 12n | |
2578 | Defines zvol block devices behaviour when \fBvolmode\fR is set to \fBdefault\fR. | |
2579 | Valid values are \fB1\fR (full), \fB2\fR (dev) and \fB3\fR (none). | |
2580 | .sp | |
2581 | Default value: \fB1\fR. | |
2582 | .RE | |
2583 | ||
39ccc909 | 2584 | .sp |
2585 | .ne 2 | |
2586 | .na | |
2587 | \fBzfs_qat_disable\fR (int) | |
2588 | .ad | |
2589 | .RS 12n | |
cf637391 TC |
2590 | This tunable disables qat hardware acceleration for gzip compression and. |
2591 | AES-GCM encryption. It is available only if qat acceleration is compiled in | |
2592 | and the qat driver is present. | |
39ccc909 | 2593 | .sp |
2594 | Use \fB1\fR for yes and \fB0\fR for no (default). | |
2595 | .RE | |
2596 | ||
e8b96c60 MA |
2597 | .SH ZFS I/O SCHEDULER |
2598 | ZFS issues I/O operations to leaf vdevs to satisfy and complete I/Os. | |
2599 | The I/O scheduler determines when and in what order those operations are | |
2600 | issued. The I/O scheduler divides operations into five I/O classes | |
2601 | prioritized in the following order: sync read, sync write, async read, | |
2602 | async write, and scrub/resilver. Each queue defines the minimum and | |
2603 | maximum number of concurrent operations that may be issued to the | |
2604 | device. In addition, the device has an aggregate maximum, | |
2605 | \fBzfs_vdev_max_active\fR. Note that the sum of the per-queue minimums | |
2606 | must not exceed the aggregate maximum. If the sum of the per-queue | |
2607 | maximums exceeds the aggregate maximum, then the number of active I/Os | |
2608 | may reach \fBzfs_vdev_max_active\fR, in which case no further I/Os will | |
2609 | be issued regardless of whether all per-queue minimums have been met. | |
2610 | .sp | |
2611 | For many physical devices, throughput increases with the number of | |
2612 | concurrent operations, but latency typically suffers. Further, physical | |
2613 | devices typically have a limit at which more concurrent operations have no | |
2614 | effect on throughput or can actually cause it to decrease. | |
2615 | .sp | |
2616 | The scheduler selects the next operation to issue by first looking for an | |
2617 | I/O class whose minimum has not been satisfied. Once all are satisfied and | |
2618 | the aggregate maximum has not been hit, the scheduler looks for classes | |
2619 | whose maximum has not been satisfied. Iteration through the I/O classes is | |
2620 | done in the order specified above. No further operations are issued if the | |
2621 | aggregate maximum number of concurrent operations has been hit or if there | |
2622 | are no operations queued for an I/O class that has not hit its maximum. | |
2623 | Every time an I/O is queued or an operation completes, the I/O scheduler | |
2624 | looks for new operations to issue. | |
2625 | .sp | |
2626 | In general, smaller max_active's will lead to lower latency of synchronous | |
2627 | operations. Larger max_active's may lead to higher overall throughput, | |
2628 | depending on underlying storage. | |
2629 | .sp | |
2630 | The ratio of the queues' max_actives determines the balance of performance | |
2631 | between reads, writes, and scrubs. E.g., increasing | |
2632 | \fBzfs_vdev_scrub_max_active\fR will cause the scrub or resilver to complete | |
2633 | more quickly, but reads and writes to have higher latency and lower throughput. | |
2634 | .sp | |
2635 | All I/O classes have a fixed maximum number of outstanding operations | |
2636 | except for the async write class. Asynchronous writes represent the data | |
2637 | that is committed to stable storage during the syncing stage for | |
2638 | transaction groups. Transaction groups enter the syncing state | |
2639 | periodically so the number of queued async writes will quickly burst up | |
2640 | and then bleed down to zero. Rather than servicing them as quickly as | |
2641 | possible, the I/O scheduler changes the maximum number of active async | |
2642 | write I/Os according to the amount of dirty data in the pool. Since | |
2643 | both throughput and latency typically increase with the number of | |
2644 | concurrent operations issued to physical devices, reducing the | |
2645 | burstiness in the number of concurrent operations also stabilizes the | |
2646 | response time of operations from other -- and in particular synchronous | |
2647 | -- queues. In broad strokes, the I/O scheduler will issue more | |
2648 | concurrent operations from the async write queue as there's more dirty | |
2649 | data in the pool. | |
2650 | .sp | |
2651 | Async Writes | |
2652 | .sp | |
2653 | The number of concurrent operations issued for the async write I/O class | |
2654 | follows a piece-wise linear function defined by a few adjustable points. | |
2655 | .nf | |
2656 | ||
2657 | | o---------| <-- zfs_vdev_async_write_max_active | |
2658 | ^ | /^ | | |
2659 | | | / | | | |
2660 | active | / | | | |
2661 | I/O | / | | | |
2662 | count | / | | | |
2663 | | / | | | |
2664 | |-------o | | <-- zfs_vdev_async_write_min_active | |
2665 | 0|_______^______|_________| | |
2666 | 0% | | 100% of zfs_dirty_data_max | |
2667 | | | | |
2668 | | `-- zfs_vdev_async_write_active_max_dirty_percent | |
2669 | `--------- zfs_vdev_async_write_active_min_dirty_percent | |
2670 | ||
2671 | .fi | |
2672 | Until the amount of dirty data exceeds a minimum percentage of the dirty | |
2673 | data allowed in the pool, the I/O scheduler will limit the number of | |
2674 | concurrent operations to the minimum. As that threshold is crossed, the | |
2675 | number of concurrent operations issued increases linearly to the maximum at | |
2676 | the specified maximum percentage of the dirty data allowed in the pool. | |
2677 | .sp | |
2678 | Ideally, the amount of dirty data on a busy pool will stay in the sloped | |
2679 | part of the function between \fBzfs_vdev_async_write_active_min_dirty_percent\fR | |
2680 | and \fBzfs_vdev_async_write_active_max_dirty_percent\fR. If it exceeds the | |
2681 | maximum percentage, this indicates that the rate of incoming data is | |
2682 | greater than the rate that the backend storage can handle. In this case, we | |
2683 | must further throttle incoming writes, as described in the next section. | |
2684 | ||
2685 | .SH ZFS TRANSACTION DELAY | |
2686 | We delay transactions when we've determined that the backend storage | |
2687 | isn't able to accommodate the rate of incoming writes. | |
2688 | .sp | |
2689 | If there is already a transaction waiting, we delay relative to when | |
2690 | that transaction will finish waiting. This way the calculated delay time | |
2691 | is independent of the number of threads concurrently executing | |
2692 | transactions. | |
2693 | .sp | |
2694 | If we are the only waiter, wait relative to when the transaction | |
2695 | started, rather than the current time. This credits the transaction for | |
2696 | "time already served", e.g. reading indirect blocks. | |
2697 | .sp | |
2698 | The minimum time for a transaction to take is calculated as: | |
2699 | .nf | |
2700 | min_time = zfs_delay_scale * (dirty - min) / (max - dirty) | |
2701 | min_time is then capped at 100 milliseconds. | |
2702 | .fi | |
2703 | .sp | |
2704 | The delay has two degrees of freedom that can be adjusted via tunables. The | |
2705 | percentage of dirty data at which we start to delay is defined by | |
2706 | \fBzfs_delay_min_dirty_percent\fR. This should typically be at or above | |
2707 | \fBzfs_vdev_async_write_active_max_dirty_percent\fR so that we only start to | |
2708 | delay after writing at full speed has failed to keep up with the incoming write | |
2709 | rate. The scale of the curve is defined by \fBzfs_delay_scale\fR. Roughly speaking, | |
2710 | this variable determines the amount of delay at the midpoint of the curve. | |
2711 | .sp | |
2712 | .nf | |
2713 | delay | |
2714 | 10ms +-------------------------------------------------------------*+ | |
2715 | | *| | |
2716 | 9ms + *+ | |
2717 | | *| | |
2718 | 8ms + *+ | |
2719 | | * | | |
2720 | 7ms + * + | |
2721 | | * | | |
2722 | 6ms + * + | |
2723 | | * | | |
2724 | 5ms + * + | |
2725 | | * | | |
2726 | 4ms + * + | |
2727 | | * | | |
2728 | 3ms + * + | |
2729 | | * | | |
2730 | 2ms + (midpoint) * + | |
2731 | | | ** | | |
2732 | 1ms + v *** + | |
2733 | | zfs_delay_scale ----------> ******** | | |
2734 | 0 +-------------------------------------*********----------------+ | |
2735 | 0% <- zfs_dirty_data_max -> 100% | |
2736 | .fi | |
2737 | .sp | |
2738 | Note that since the delay is added to the outstanding time remaining on the | |
2739 | most recent transaction, the delay is effectively the inverse of IOPS. | |
2740 | Here the midpoint of 500us translates to 2000 IOPS. The shape of the curve | |
2741 | was chosen such that small changes in the amount of accumulated dirty data | |
2742 | in the first 3/4 of the curve yield relatively small differences in the | |
2743 | amount of delay. | |
2744 | .sp | |
2745 | The effects can be easier to understand when the amount of delay is | |
2746 | represented on a log scale: | |
2747 | .sp | |
2748 | .nf | |
2749 | delay | |
2750 | 100ms +-------------------------------------------------------------++ | |
2751 | + + | |
2752 | | | | |
2753 | + *+ | |
2754 | 10ms + *+ | |
2755 | + ** + | |
2756 | | (midpoint) ** | | |
2757 | + | ** + | |
2758 | 1ms + v **** + | |
2759 | + zfs_delay_scale ----------> ***** + | |
2760 | | **** | | |
2761 | + **** + | |
2762 | 100us + ** + | |
2763 | + * + | |
2764 | | * | | |
2765 | + * + | |
2766 | 10us + * + | |
2767 | + + | |
2768 | | | | |
2769 | + + | |
2770 | +--------------------------------------------------------------+ | |
2771 | 0% <- zfs_dirty_data_max -> 100% | |
2772 | .fi | |
2773 | .sp | |
2774 | Note here that only as the amount of dirty data approaches its limit does | |
2775 | the delay start to increase rapidly. The goal of a properly tuned system | |
2776 | should be to keep the amount of dirty data out of that range by first | |
2777 | ensuring that the appropriate limits are set for the I/O scheduler to reach | |
2778 | optimal throughput on the backend storage, and then by changing the value | |
2779 | of \fBzfs_delay_scale\fR to increase the steepness of the curve. |