2 .\" Copyright (c) 2013 by Turbo Fredriksson <turbo@bayour.com>. All rights reserved.
3 .\" Copyright (c) 2019 by Delphix. All rights reserved.
4 .\" Copyright (c) 2019 Datto Inc.
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.
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]
17 .TH ZFS-MODULE-PARAMETERS 5 "Feb 15, 2019"
19 zfs\-module\-parameters \- ZFS module parameters
23 Description of the different parameters to the ZFS module.
25 .SS "Module parameters"
32 \fBdbuf_cache_max_bytes\fR (ulong)
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
41 Default value: \fB0\fR.
47 \fBdbuf_metadata_cache_max_bytes\fR (ulong)
50 Maximum size in bytes of the metadata dbuf cache. When \fB0\fR this value will
51 default to \fB1/2^dbuf_cache_shift\fR (1/16) of the target ARC size, otherwise
52 the provided value in bytes will be used. The behavior of the metadata dbuf
53 cache and its associated settings can be observed via the
54 \fB/proc/spl/kstat/zfs/dbufstats\fR kstat.
56 Default value: \fB0\fR.
62 \fBdbuf_cache_hiwater_pct\fR (uint)
65 The percentage over \fBdbuf_cache_max_bytes\fR when dbufs must be evicted
68 Default value: \fB10\fR%.
74 \fBdbuf_cache_lowater_pct\fR (uint)
77 The percentage below \fBdbuf_cache_max_bytes\fR when the evict thread stops
80 Default value: \fB10\fR%.
86 \fBdbuf_cache_shift\fR (int)
89 Set the size of the dbuf cache, \fBdbuf_cache_max_bytes\fR, to a log2 fraction
90 of the target arc size.
92 Default value: \fB5\fR.
98 \fBdbuf_metadata_cache_shift\fR (int)
101 Set the size of the dbuf metadata cache, \fBdbuf_metadata_cache_max_bytes\fR,
102 to a log2 fraction of the target arc size.
104 Default value: \fB6\fR.
110 \fBignore_hole_birth\fR (int)
113 When set, the hole_birth optimization will not be used, and all holes will
114 always be sent on zfs send. Useful if you suspect your datasets are affected
115 by a bug in hole_birth.
117 Use \fB1\fR for on (default) and \fB0\fR for off.
123 \fBl2arc_feed_again\fR (int)
126 Turbo L2ARC warm-up. When the L2ARC is cold the fill interval will be set as
129 Use \fB1\fR for yes (default) and \fB0\fR to disable.
135 \fBl2arc_feed_min_ms\fR (ulong)
138 Min feed interval in milliseconds. Requires \fBl2arc_feed_again=1\fR and only
139 applicable in related situations.
141 Default value: \fB200\fR.
147 \fBl2arc_feed_secs\fR (ulong)
150 Seconds between L2ARC writing
152 Default value: \fB1\fR.
158 \fBl2arc_headroom\fR (ulong)
161 How far through the ARC lists to search for L2ARC cacheable content, expressed
162 as a multiplier of \fBl2arc_write_max\fR
164 Default value: \fB2\fR.
170 \fBl2arc_headroom_boost\fR (ulong)
173 Scales \fBl2arc_headroom\fR by this percentage when L2ARC contents are being
174 successfully compressed before writing. A value of 100 disables this feature.
176 Default value: \fB200\fR%.
182 \fBl2arc_noprefetch\fR (int)
185 Do not write buffers to L2ARC if they were prefetched but not used by
188 Use \fB1\fR for yes (default) and \fB0\fR to disable.
194 \fBl2arc_norw\fR (int)
197 No reads during writes
199 Use \fB1\fR for yes and \fB0\fR for no (default).
205 \fBl2arc_write_boost\fR (ulong)
208 Cold L2ARC devices will have \fBl2arc_write_max\fR increased by this amount
209 while they remain cold.
211 Default value: \fB8,388,608\fR.
217 \fBl2arc_write_max\fR (ulong)
220 Max write bytes per interval
222 Default value: \fB8,388,608\fR.
228 \fBmetaslab_aliquot\fR (ulong)
231 Metaslab granularity, in bytes. This is roughly similar to what would be
232 referred to as the "stripe size" in traditional RAID arrays. In normal
233 operation, ZFS will try to write this amount of data to a top-level vdev
234 before moving on to the next one.
236 Default value: \fB524,288\fR.
242 \fBmetaslab_bias_enabled\fR (int)
245 Enable metaslab group biasing based on its vdev's over- or under-utilization
246 relative to the pool.
248 Use \fB1\fR for yes (default) and \fB0\fR for no.
254 \fBmetaslab_force_ganging\fR (ulong)
257 Make some blocks above a certain size be gang blocks. This option is used
258 by the test suite to facilitate testing.
260 Default value: \fB16,777,217\fR.
266 \fBzfs_metaslab_segment_weight_enabled\fR (int)
269 Enable/disable segment-based metaslab selection.
271 Use \fB1\fR for yes (default) and \fB0\fR for no.
277 \fBzfs_metaslab_switch_threshold\fR (int)
280 When using segment-based metaslab selection, continue allocating
281 from the active metaslab until \fBzfs_metaslab_switch_threshold\fR
282 worth of buckets have been exhausted.
284 Default value: \fB2\fR.
290 \fBmetaslab_debug_load\fR (int)
293 Load all metaslabs during pool import.
295 Use \fB1\fR for yes and \fB0\fR for no (default).
301 \fBmetaslab_debug_unload\fR (int)
304 Prevent metaslabs from being unloaded.
306 Use \fB1\fR for yes and \fB0\fR for no (default).
312 \fBmetaslab_fragmentation_factor_enabled\fR (int)
315 Enable use of the fragmentation metric in computing metaslab weights.
317 Use \fB1\fR for yes (default) and \fB0\fR for no.
323 \fBzfs_vdev_default_ms_count\fR (int)
326 When a vdev is added target this number of metaslabs per top-level vdev.
328 Default value: \fB200\fR.
334 \fBzfs_vdev_min_ms_count\fR (int)
337 Minimum number of metaslabs to create in a top-level vdev.
339 Default value: \fB16\fR.
345 \fBvdev_ms_count_limit\fR (int)
348 Practical upper limit of total metaslabs per top-level vdev.
350 Default value: \fB131,072\fR.
356 \fBmetaslab_preload_enabled\fR (int)
359 Enable metaslab group preloading.
361 Use \fB1\fR for yes (default) and \fB0\fR for no.
367 \fBmetaslab_lba_weighting_enabled\fR (int)
370 Give more weight to metaslabs with lower LBAs, assuming they have
371 greater bandwidth as is typically the case on a modern constant
372 angular velocity disk drive.
374 Use \fB1\fR for yes (default) and \fB0\fR for no.
380 \fBspa_config_path\fR (charp)
385 Default value: \fB/etc/zfs/zpool.cache\fR.
391 \fBspa_asize_inflation\fR (int)
394 Multiplication factor used to estimate actual disk consumption from the
395 size of data being written. The default value is a worst case estimate,
396 but lower values may be valid for a given pool depending on its
397 configuration. Pool administrators who understand the factors involved
398 may wish to specify a more realistic inflation factor, particularly if
399 they operate close to quota or capacity limits.
401 Default value: \fB24\fR.
407 \fBspa_load_print_vdev_tree\fR (int)
410 Whether to print the vdev tree in the debugging message buffer during pool import.
411 Use 0 to disable and 1 to enable.
413 Default value: \fB0\fR.
419 \fBspa_load_verify_data\fR (int)
422 Whether to traverse data blocks during an "extreme rewind" (\fB-X\fR)
423 import. Use 0 to disable and 1 to enable.
425 An extreme rewind import normally performs a full traversal of all
426 blocks in the pool for verification. If this parameter is set to 0,
427 the traversal skips non-metadata blocks. It can be toggled once the
428 import has started to stop or start the traversal of non-metadata blocks.
430 Default value: \fB1\fR.
436 \fBspa_load_verify_metadata\fR (int)
439 Whether to traverse blocks during an "extreme rewind" (\fB-X\fR)
440 pool import. Use 0 to disable and 1 to enable.
442 An extreme rewind import normally performs a full traversal of all
443 blocks in the pool for verification. If this parameter is set to 0,
444 the traversal is not performed. It can be toggled once the import has
445 started to stop or start the traversal.
447 Default value: \fB1\fR.
453 \fBspa_load_verify_maxinflight\fR (int)
456 Maximum concurrent I/Os during the traversal performed during an "extreme
457 rewind" (\fB-X\fR) pool import.
459 Default value: \fB10000\fR.
465 \fBspa_slop_shift\fR (int)
468 Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space
469 in the pool to be consumed. This ensures that we don't run the pool
470 completely out of space, due to unaccounted changes (e.g. to the MOS).
471 It also limits the worst-case time to allocate space. If we have
472 less than this amount of free space, most ZPL operations (e.g. write,
473 create) will return ENOSPC.
475 Default value: \fB5\fR.
481 \fBvdev_removal_max_span\fR (int)
484 During top-level vdev removal, chunks of data are copied from the vdev
485 which may include free space in order to trade bandwidth for IOPS.
486 This parameter determines the maximum span of free space (in bytes)
487 which will be included as "unnecessary" data in a chunk of copied data.
489 The default value here was chosen to align with
490 \fBzfs_vdev_read_gap_limit\fR, which is a similar concept when doing
491 regular reads (but there's no reason it has to be the same).
493 Default value: \fB32,768\fR.
499 \fBzfetch_array_rd_sz\fR (ulong)
502 If prefetching is enabled, disable prefetching for reads larger than this size.
504 Default value: \fB1,048,576\fR.
510 \fBzfetch_max_distance\fR (uint)
513 Max bytes to prefetch per stream (default 8MB).
515 Default value: \fB8,388,608\fR.
521 \fBzfetch_max_streams\fR (uint)
524 Max number of streams per zfetch (prefetch streams per file).
526 Default value: \fB8\fR.
532 \fBzfetch_min_sec_reap\fR (uint)
535 Min time before an active prefetch stream can be reclaimed
537 Default value: \fB2\fR.
543 \fBzfs_abd_scatter_min_size\fR (uint)
546 This is the minimum allocation size that will use scatter (page-based)
547 ABD's. Smaller allocations will use linear ABD's.
549 Default value: \fB1536\fR (512B and 1KB allocations will be linear).
555 \fBzfs_arc_dnode_limit\fR (ulong)
558 When the number of bytes consumed by dnodes in the ARC exceeds this number of
559 bytes, try to unpin some of it in response to demand for non-metadata. This
560 value acts as a ceiling to the amount of dnode metadata, and defaults to 0 which
561 indicates that a percent which is based on \fBzfs_arc_dnode_limit_percent\fR of
562 the ARC meta buffers that may be used for dnodes.
564 See also \fBzfs_arc_meta_prune\fR which serves a similar purpose but is used
565 when the amount of metadata in the ARC exceeds \fBzfs_arc_meta_limit\fR rather
566 than in response to overall demand for non-metadata.
569 Default value: \fB0\fR.
575 \fBzfs_arc_dnode_limit_percent\fR (ulong)
578 Percentage that can be consumed by dnodes of ARC meta buffers.
580 See also \fBzfs_arc_dnode_limit\fR which serves a similar purpose but has a
581 higher priority if set to nonzero value.
583 Default value: \fB10\fR%.
589 \fBzfs_arc_dnode_reduce_percent\fR (ulong)
592 Percentage of ARC dnodes to try to scan in response to demand for non-metadata
593 when the number of bytes consumed by dnodes exceeds \fBzfs_arc_dnode_limit\fR.
596 Default value: \fB10\fR% of the number of dnodes in the ARC.
602 \fBzfs_arc_average_blocksize\fR (int)
605 The ARC's buffer hash table is sized based on the assumption of an average
606 block size of \fBzfs_arc_average_blocksize\fR (default 8K). This works out
607 to roughly 1MB of hash table per 1GB of physical memory with 8-byte pointers.
608 For configurations with a known larger average block size this value can be
609 increased to reduce the memory footprint.
612 Default value: \fB8192\fR.
618 \fBzfs_arc_evict_batch_limit\fR (int)
621 Number ARC headers to evict per sub-list before proceeding to another sub-list.
622 This batch-style operation prevents entire sub-lists from being evicted at once
623 but comes at a cost of additional unlocking and locking.
625 Default value: \fB10\fR.
631 \fBzfs_arc_grow_retry\fR (int)
634 If set to a non zero value, it will replace the arc_grow_retry value with this value.
635 The arc_grow_retry value (default 5) is the number of seconds the ARC will wait before
636 trying to resume growth after a memory pressure event.
638 Default value: \fB0\fR.
644 \fBzfs_arc_lotsfree_percent\fR (int)
647 Throttle I/O when free system memory drops below this percentage of total
648 system memory. Setting this value to 0 will disable the throttle.
650 Default value: \fB10\fR%.
656 \fBzfs_arc_max\fR (ulong)
659 Max arc size of ARC in bytes. If set to 0 then it will consume 1/2 of system
660 RAM. This value must be at least 67108864 (64 megabytes).
662 This value can be changed dynamically with some caveats. It cannot be set back
663 to 0 while running and reducing it below the current ARC size will not cause
664 the ARC to shrink without memory pressure to induce shrinking.
666 Default value: \fB0\fR.
672 \fBzfs_arc_meta_adjust_restarts\fR (ulong)
675 The number of restart passes to make while scanning the ARC attempting
676 the free buffers in order to stay below the \fBzfs_arc_meta_limit\fR.
677 This value should not need to be tuned but is available to facilitate
678 performance analysis.
680 Default value: \fB4096\fR.
686 \fBzfs_arc_meta_limit\fR (ulong)
689 The maximum allowed size in bytes that meta data buffers are allowed to
690 consume in the ARC. When this limit is reached meta data buffers will
691 be reclaimed even if the overall arc_c_max has not been reached. This
692 value defaults to 0 which indicates that a percent which is based on
693 \fBzfs_arc_meta_limit_percent\fR of the ARC may be used for meta data.
695 This value my be changed dynamically except that it cannot be set back to 0
696 for a specific percent of the ARC; it must be set to an explicit value.
698 Default value: \fB0\fR.
704 \fBzfs_arc_meta_limit_percent\fR (ulong)
707 Percentage of ARC buffers that can be used for meta data.
709 See also \fBzfs_arc_meta_limit\fR which serves a similar purpose but has a
710 higher priority if set to nonzero value.
713 Default value: \fB75\fR%.
719 \fBzfs_arc_meta_min\fR (ulong)
722 The minimum allowed size in bytes that meta data buffers may consume in
723 the ARC. This value defaults to 0 which disables a floor on the amount
724 of the ARC devoted meta data.
726 Default value: \fB0\fR.
732 \fBzfs_arc_meta_prune\fR (int)
735 The number of dentries and inodes to be scanned looking for entries
736 which can be dropped. This may be required when the ARC reaches the
737 \fBzfs_arc_meta_limit\fR because dentries and inodes can pin buffers
738 in the ARC. Increasing this value will cause to dentry and inode caches
739 to be pruned more aggressively. Setting this value to 0 will disable
740 pruning the inode and dentry caches.
742 Default value: \fB10,000\fR.
748 \fBzfs_arc_meta_strategy\fR (int)
751 Define the strategy for ARC meta data buffer eviction (meta reclaim strategy).
752 A value of 0 (META_ONLY) will evict only the ARC meta data buffers.
753 A value of 1 (BALANCED) indicates that additional data buffers may be evicted if
754 that is required to in order to evict the required number of meta data buffers.
756 Default value: \fB1\fR.
762 \fBzfs_arc_min\fR (ulong)
765 Min arc size of ARC in bytes. If set to 0 then arc_c_min will default to
766 consuming the larger of 32M or 1/32 of total system memory.
768 Default value: \fB0\fR.
774 \fBzfs_arc_min_prefetch_ms\fR (int)
777 Minimum time prefetched blocks are locked in the ARC, specified in ms.
778 A value of \fB0\fR will default to 1000 ms.
780 Default value: \fB0\fR.
786 \fBzfs_arc_min_prescient_prefetch_ms\fR (int)
789 Minimum time "prescient prefetched" blocks are locked in the ARC, specified
790 in ms. These blocks are meant to be prefetched fairly aggresively ahead of
791 the code that may use them. A value of \fB0\fR will default to 6000 ms.
793 Default value: \fB0\fR.
799 \fBzfs_max_missing_tvds\fR (int)
802 Number of missing top-level vdevs which will be allowed during
803 pool import (only in read-only mode).
805 Default value: \fB0\fR
811 \fBzfs_multilist_num_sublists\fR (int)
814 To allow more fine-grained locking, each ARC state contains a series
815 of lists for both data and meta data objects. Locking is performed at
816 the level of these "sub-lists". This parameters controls the number of
817 sub-lists per ARC state, and also applies to other uses of the
818 multilist data structure.
820 Default value: \fB4\fR or the number of online CPUs, whichever is greater
826 \fBzfs_arc_overflow_shift\fR (int)
829 The ARC size is considered to be overflowing if it exceeds the current
830 ARC target size (arc_c) by a threshold determined by this parameter.
831 The threshold is calculated as a fraction of arc_c using the formula
832 "arc_c >> \fBzfs_arc_overflow_shift\fR".
834 The default value of 8 causes the ARC to be considered to be overflowing
835 if it exceeds the target size by 1/256th (0.3%) of the target size.
837 When the ARC is overflowing, new buffer allocations are stalled until
838 the reclaim thread catches up and the overflow condition no longer exists.
840 Default value: \fB8\fR.
847 \fBzfs_arc_p_min_shift\fR (int)
850 If set to a non zero value, this will update arc_p_min_shift (default 4)
852 arc_p_min_shift is used to shift of arc_c for calculating both min and max
855 Default value: \fB0\fR.
861 \fBzfs_arc_p_dampener_disable\fR (int)
864 Disable arc_p adapt dampener
866 Use \fB1\fR for yes (default) and \fB0\fR to disable.
872 \fBzfs_arc_shrink_shift\fR (int)
875 If set to a non zero value, this will update arc_shrink_shift (default 7)
878 Default value: \fB0\fR.
884 \fBzfs_arc_pc_percent\fR (uint)
887 Percent of pagecache to reclaim arc to
889 This tunable allows ZFS arc to play more nicely with the kernel's LRU
890 pagecache. It can guarantee that the arc size won't collapse under scanning
891 pressure on the pagecache, yet still allows arc to be reclaimed down to
892 zfs_arc_min if necessary. This value is specified as percent of pagecache
893 size (as measured by NR_FILE_PAGES) where that percent may exceed 100. This
894 only operates during memory pressure/reclaim.
896 Default value: \fB0\fR% (disabled).
902 \fBzfs_arc_sys_free\fR (ulong)
905 The target number of bytes the ARC should leave as free memory on the system.
906 Defaults to the larger of 1/64 of physical memory or 512K. Setting this
907 option to a non-zero value will override the default.
909 Default value: \fB0\fR.
915 \fBzfs_autoimport_disable\fR (int)
918 Disable pool import at module load by ignoring the cache file (typically \fB/etc/zfs/zpool.cache\fR).
920 Use \fB1\fR for yes (default) and \fB0\fR for no.
926 \fBzfs_checksums_per_second\fR (int)
929 Rate limit checksum events to this many per second. Note that this should
930 not be set below the zed thresholds (currently 10 checksums over 10 sec)
931 or else zed may not trigger any action.
939 \fBzfs_commit_timeout_pct\fR (int)
942 This controls the amount of time that a ZIL block (lwb) will remain "open"
943 when it isn't "full", and it has a thread waiting for it to be committed to
944 stable storage. The timeout is scaled based on a percentage of the last lwb
945 latency to avoid significantly impacting the latency of each individual
946 transaction record (itx).
948 Default value: \fB5\fR%.
954 \fBzfs_condense_indirect_vdevs_enable\fR (int)
957 Enable condensing indirect vdev mappings. When set to a non-zero value,
958 attempt to condense indirect vdev mappings if the mapping uses more than
959 \fBzfs_condense_min_mapping_bytes\fR bytes of memory and if the obsolete
960 space map object uses more than \fBzfs_condense_max_obsolete_bytes\fR
961 bytes on-disk. The condensing process is an attempt to save memory by
962 removing obsolete mappings.
964 Default value: \fB1\fR.
970 \fBzfs_condense_max_obsolete_bytes\fR (ulong)
973 Only attempt to condense indirect vdev mappings if the on-disk size
974 of the obsolete space map object is greater than this number of bytes
975 (see \fBfBzfs_condense_indirect_vdevs_enable\fR).
977 Default value: \fB1,073,741,824\fR.
983 \fBzfs_condense_min_mapping_bytes\fR (ulong)
986 Minimum size vdev mapping to attempt to condense (see
987 \fBzfs_condense_indirect_vdevs_enable\fR).
989 Default value: \fB131,072\fR.
995 \fBzfs_dbgmsg_enable\fR (int)
998 Internally ZFS keeps a small log to facilitate debugging. By default the log
999 is disabled, to enable it set this option to 1. The contents of the log can
1000 be accessed by reading the /proc/spl/kstat/zfs/dbgmsg file. Writing 0 to
1001 this proc file clears the log.
1003 Default value: \fB0\fR.
1009 \fBzfs_dbgmsg_maxsize\fR (int)
1012 The maximum size in bytes of the internal ZFS debug log.
1014 Default value: \fB4M\fR.
1020 \fBzfs_dbuf_state_index\fR (int)
1023 This feature is currently unused. It is normally used for controlling what
1024 reporting is available under /proc/spl/kstat/zfs.
1026 Default value: \fB0\fR.
1032 \fBzfs_deadman_enabled\fR (int)
1035 When a pool sync operation takes longer than \fBzfs_deadman_synctime_ms\fR
1036 milliseconds, or when an individual I/O takes longer than
1037 \fBzfs_deadman_ziotime_ms\fR milliseconds, then the operation is considered to
1038 be "hung". If \fBzfs_deadman_enabled\fR is set then the deadman behavior is
1039 invoked as described by the \fBzfs_deadman_failmode\fR module option.
1040 By default the deadman is enabled and configured to \fBwait\fR which results
1041 in "hung" I/Os only being logged. The deadman is automatically disabled
1042 when a pool gets suspended.
1044 Default value: \fB1\fR.
1050 \fBzfs_deadman_failmode\fR (charp)
1053 Controls the failure behavior when the deadman detects a "hung" I/O. Valid
1054 values are \fBwait\fR, \fBcontinue\fR, and \fBpanic\fR.
1056 \fBwait\fR - Wait for a "hung" I/O to complete. For each "hung" I/O a
1057 "deadman" event will be posted describing that I/O.
1059 \fBcontinue\fR - Attempt to recover from a "hung" I/O by re-dispatching it
1060 to the I/O pipeline if possible.
1062 \fBpanic\fR - Panic the system. This can be used to facilitate an automatic
1063 fail-over to a properly configured fail-over partner.
1065 Default value: \fBwait\fR.
1071 \fBzfs_deadman_checktime_ms\fR (int)
1074 Check time in milliseconds. This defines the frequency at which we check
1075 for hung I/O and potentially invoke the \fBzfs_deadman_failmode\fR behavior.
1077 Default value: \fB60,000\fR.
1083 \fBzfs_deadman_synctime_ms\fR (ulong)
1086 Interval in milliseconds after which the deadman is triggered and also
1087 the interval after which a pool sync operation is considered to be "hung".
1088 Once this limit is exceeded the deadman will be invoked every
1089 \fBzfs_deadman_checktime_ms\fR milliseconds until the pool sync completes.
1091 Default value: \fB600,000\fR.
1097 \fBzfs_deadman_ziotime_ms\fR (ulong)
1100 Interval in milliseconds after which the deadman is triggered and an
1101 individual I/O operation is considered to be "hung". As long as the I/O
1102 remains "hung" the deadman will be invoked every \fBzfs_deadman_checktime_ms\fR
1103 milliseconds until the I/O completes.
1105 Default value: \fB300,000\fR.
1111 \fBzfs_dedup_prefetch\fR (int)
1114 Enable prefetching dedup-ed blks
1116 Use \fB1\fR for yes and \fB0\fR to disable (default).
1122 \fBzfs_delay_min_dirty_percent\fR (int)
1125 Start to delay each transaction once there is this amount of dirty data,
1126 expressed as a percentage of \fBzfs_dirty_data_max\fR.
1127 This value should be >= zfs_vdev_async_write_active_max_dirty_percent.
1128 See the section "ZFS TRANSACTION DELAY".
1130 Default value: \fB60\fR%.
1136 \fBzfs_delay_scale\fR (int)
1139 This controls how quickly the transaction delay approaches infinity.
1140 Larger values cause longer delays for a given amount of dirty data.
1142 For the smoothest delay, this value should be about 1 billion divided
1143 by the maximum number of operations per second. This will smoothly
1144 handle between 10x and 1/10th this number.
1146 See the section "ZFS TRANSACTION DELAY".
1148 Note: \fBzfs_delay_scale\fR * \fBzfs_dirty_data_max\fR must be < 2^64.
1150 Default value: \fB500,000\fR.
1156 \fBzfs_slow_io_events_per_second\fR (int)
1159 Rate limit delay zevents (which report slow I/Os) to this many per second.
1167 \fBzfs_unlink_suspend_progress\fR (uint)
1170 When enabled, files will not be asynchronously removed from the list of pending
1171 unlinks and the space they consume will be leaked. Once this option has been
1172 disabled and the dataset is remounted, the pending unlinks will be processed
1173 and the freed space returned to the pool.
1174 This option is used by the test suite to facilitate testing.
1176 Uses \fB0\fR (default) to allow progress and \fB1\fR to pause progress.
1182 \fBzfs_delete_blocks\fR (ulong)
1185 This is the used to define a large file for the purposes of delete. Files
1186 containing more than \fBzfs_delete_blocks\fR will be deleted asynchronously
1187 while smaller files are deleted synchronously. Decreasing this value will
1188 reduce the time spent in an unlink(2) system call at the expense of a longer
1189 delay before the freed space is available.
1191 Default value: \fB20,480\fR.
1197 \fBzfs_dirty_data_max\fR (int)
1200 Determines the dirty space limit in bytes. Once this limit is exceeded, new
1201 writes are halted until space frees up. This parameter takes precedence
1202 over \fBzfs_dirty_data_max_percent\fR.
1203 See the section "ZFS TRANSACTION DELAY".
1205 Default value: \fB10\fR% of physical RAM, capped at \fBzfs_dirty_data_max_max\fR.
1211 \fBzfs_dirty_data_max_max\fR (int)
1214 Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed in bytes.
1215 This limit is only enforced at module load time, and will be ignored if
1216 \fBzfs_dirty_data_max\fR is later changed. This parameter takes
1217 precedence over \fBzfs_dirty_data_max_max_percent\fR. See the section
1218 "ZFS TRANSACTION DELAY".
1220 Default value: \fB25\fR% of physical RAM.
1226 \fBzfs_dirty_data_max_max_percent\fR (int)
1229 Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed as a
1230 percentage of physical RAM. This limit is only enforced at module load
1231 time, and will be ignored if \fBzfs_dirty_data_max\fR is later changed.
1232 The parameter \fBzfs_dirty_data_max_max\fR takes precedence over this
1233 one. See the section "ZFS TRANSACTION DELAY".
1235 Default value: \fB25\fR%.
1241 \fBzfs_dirty_data_max_percent\fR (int)
1244 Determines the dirty space limit, expressed as a percentage of all
1245 memory. Once this limit is exceeded, new writes are halted until space frees
1246 up. The parameter \fBzfs_dirty_data_max\fR takes precedence over this
1247 one. See the section "ZFS TRANSACTION DELAY".
1249 Default value: \fB10\fR%, subject to \fBzfs_dirty_data_max_max\fR.
1255 \fBzfs_dirty_data_sync_percent\fR (int)
1258 Start syncing out a transaction group if there's at least this much dirty data
1259 as a percentage of \fBzfs_dirty_data_max\fR. This should be less than
1260 \fBzfs_vdev_async_write_active_min_dirty_percent\fR.
1262 Default value: \fB20\fR% of \fBzfs_dirty_data_max\fR.
1268 \fBzfs_fletcher_4_impl\fR (string)
1271 Select a fletcher 4 implementation.
1273 Supported selectors are: \fBfastest\fR, \fBscalar\fR, \fBsse2\fR, \fBssse3\fR,
1274 \fBavx2\fR, \fBavx512f\fR, and \fBaarch64_neon\fR.
1275 All of the selectors except \fBfastest\fR and \fBscalar\fR require instruction
1276 set extensions to be available and will only appear if ZFS detects that they are
1277 present at runtime. If multiple implementations of fletcher 4 are available,
1278 the \fBfastest\fR will be chosen using a micro benchmark. Selecting \fBscalar\fR
1279 results in the original, CPU based calculation, being used. Selecting any option
1280 other than \fBfastest\fR and \fBscalar\fR results in vector instructions from
1281 the respective CPU instruction set being used.
1283 Default value: \fBfastest\fR.
1289 \fBzfs_free_bpobj_enabled\fR (int)
1292 Enable/disable the processing of the free_bpobj object.
1294 Default value: \fB1\fR.
1300 \fBzfs_async_block_max_blocks\fR (ulong)
1303 Maximum number of blocks freed in a single txg.
1305 Default value: \fB100,000\fR.
1311 \fBzfs_override_estimate_recordsize\fR (ulong)
1314 Record size calculation override for zfs send estimates.
1316 Default value: \fB0\fR.
1322 \fBzfs_vdev_async_read_max_active\fR (int)
1325 Maximum asynchronous read I/Os active to each device.
1326 See the section "ZFS I/O SCHEDULER".
1328 Default value: \fB3\fR.
1334 \fBzfs_vdev_async_read_min_active\fR (int)
1337 Minimum asynchronous read I/Os active to each device.
1338 See the section "ZFS I/O SCHEDULER".
1340 Default value: \fB1\fR.
1346 \fBzfs_vdev_async_write_active_max_dirty_percent\fR (int)
1349 When the pool has more than
1350 \fBzfs_vdev_async_write_active_max_dirty_percent\fR dirty data, use
1351 \fBzfs_vdev_async_write_max_active\fR to limit active async writes. If
1352 the dirty data is between min and max, the active I/O limit is linearly
1353 interpolated. See the section "ZFS I/O SCHEDULER".
1355 Default value: \fB60\fR%.
1361 \fBzfs_vdev_async_write_active_min_dirty_percent\fR (int)
1364 When the pool has less than
1365 \fBzfs_vdev_async_write_active_min_dirty_percent\fR dirty data, use
1366 \fBzfs_vdev_async_write_min_active\fR to limit active async writes. If
1367 the dirty data is between min and max, the active I/O limit is linearly
1368 interpolated. See the section "ZFS I/O SCHEDULER".
1370 Default value: \fB30\fR%.
1376 \fBzfs_vdev_async_write_max_active\fR (int)
1379 Maximum asynchronous write I/Os active to each device.
1380 See the section "ZFS I/O SCHEDULER".
1382 Default value: \fB10\fR.
1388 \fBzfs_vdev_async_write_min_active\fR (int)
1391 Minimum asynchronous write I/Os active to each device.
1392 See the section "ZFS I/O SCHEDULER".
1394 Lower values are associated with better latency on rotational media but poorer
1395 resilver performance. The default value of 2 was chosen as a compromise. A
1396 value of 3 has been shown to improve resilver performance further at a cost of
1397 further increasing latency.
1399 Default value: \fB2\fR.
1405 \fBzfs_vdev_initializing_max_active\fR (int)
1408 Maximum initializing I/Os active to each device.
1409 See the section "ZFS I/O SCHEDULER".
1411 Default value: \fB1\fR.
1417 \fBzfs_vdev_initializing_min_active\fR (int)
1420 Minimum initializing I/Os active to each device.
1421 See the section "ZFS I/O SCHEDULER".
1423 Default value: \fB1\fR.
1429 \fBzfs_vdev_max_active\fR (int)
1432 The maximum number of I/Os active to each device. Ideally, this will be >=
1433 the sum of each queue's max_active. It must be at least the sum of each
1434 queue's min_active. See the section "ZFS I/O SCHEDULER".
1436 Default value: \fB1,000\fR.
1442 \fBzfs_vdev_removal_max_active\fR (int)
1445 Maximum removal I/Os active to each device.
1446 See the section "ZFS I/O SCHEDULER".
1448 Default value: \fB2\fR.
1454 \fBzfs_vdev_removal_min_active\fR (int)
1457 Minimum removal I/Os active to each device.
1458 See the section "ZFS I/O SCHEDULER".
1460 Default value: \fB1\fR.
1466 \fBzfs_vdev_scrub_max_active\fR (int)
1469 Maximum scrub I/Os active to each device.
1470 See the section "ZFS I/O SCHEDULER".
1472 Default value: \fB2\fR.
1478 \fBzfs_vdev_scrub_min_active\fR (int)
1481 Minimum scrub I/Os active to each device.
1482 See the section "ZFS I/O SCHEDULER".
1484 Default value: \fB1\fR.
1490 \fBzfs_vdev_sync_read_max_active\fR (int)
1493 Maximum synchronous read I/Os active to each device.
1494 See the section "ZFS I/O SCHEDULER".
1496 Default value: \fB10\fR.
1502 \fBzfs_vdev_sync_read_min_active\fR (int)
1505 Minimum synchronous read I/Os active to each device.
1506 See the section "ZFS I/O SCHEDULER".
1508 Default value: \fB10\fR.
1514 \fBzfs_vdev_sync_write_max_active\fR (int)
1517 Maximum synchronous write I/Os active to each device.
1518 See the section "ZFS I/O SCHEDULER".
1520 Default value: \fB10\fR.
1526 \fBzfs_vdev_sync_write_min_active\fR (int)
1529 Minimum synchronous write I/Os active to each device.
1530 See the section "ZFS I/O SCHEDULER".
1532 Default value: \fB10\fR.
1538 \fBzfs_vdev_trim_max_active\fR (int)
1541 Maximum trim/discard I/Os active to each device.
1542 See the section "ZFS I/O SCHEDULER".
1544 Default value: \fB2\fR.
1550 \fBzfs_vdev_trim_min_active\fR (int)
1553 Minimum trim/discard I/Os active to each device.
1554 See the section "ZFS I/O SCHEDULER".
1556 Default value: \fB1\fR.
1562 \fBzfs_vdev_queue_depth_pct\fR (int)
1565 Maximum number of queued allocations per top-level vdev expressed as
1566 a percentage of \fBzfs_vdev_async_write_max_active\fR which allows the
1567 system to detect devices that are more capable of handling allocations
1568 and to allocate more blocks to those devices. It allows for dynamic
1569 allocation distribution when devices are imbalanced as fuller devices
1570 will tend to be slower than empty devices.
1572 See also \fBzio_dva_throttle_enabled\fR.
1574 Default value: \fB1000\fR%.
1580 \fBzfs_expire_snapshot\fR (int)
1583 Seconds to expire .zfs/snapshot
1585 Default value: \fB300\fR.
1591 \fBzfs_admin_snapshot\fR (int)
1594 Allow the creation, removal, or renaming of entries in the .zfs/snapshot
1595 directory to cause the creation, destruction, or renaming of snapshots.
1596 When enabled this functionality works both locally and over NFS exports
1597 which have the 'no_root_squash' option set. This functionality is disabled
1600 Use \fB1\fR for yes and \fB0\fR for no (default).
1606 \fBzfs_flags\fR (int)
1609 Set additional debugging flags. The following flags may be bitwise-or'd
1621 Enable dprintf entries in the debug log.
1623 2 ZFS_DEBUG_DBUF_VERIFY *
1624 Enable extra dbuf verifications.
1626 4 ZFS_DEBUG_DNODE_VERIFY *
1627 Enable extra dnode verifications.
1629 8 ZFS_DEBUG_SNAPNAMES
1630 Enable snapshot name verification.
1633 Check for illegally modified ARC buffers.
1635 64 ZFS_DEBUG_ZIO_FREE
1636 Enable verification of block frees.
1638 128 ZFS_DEBUG_HISTOGRAM_VERIFY
1639 Enable extra spacemap histogram verifications.
1641 256 ZFS_DEBUG_METASLAB_VERIFY
1642 Verify space accounting on disk matches in-core range_trees.
1644 512 ZFS_DEBUG_SET_ERROR
1645 Enable SET_ERROR and dprintf entries in the debug log.
1647 1024 ZFS_DEBUG_INDIRECT_REMAP
1648 Verify split blocks created by device removal.
1651 Verify TRIM ranges are always within the allocatable range tree.
1654 * Requires debug build.
1656 Default value: \fB0\fR.
1662 \fBzfs_free_leak_on_eio\fR (int)
1665 If destroy encounters an EIO while reading metadata (e.g. indirect
1666 blocks), space referenced by the missing metadata can not be freed.
1667 Normally this causes the background destroy to become "stalled", as
1668 it is unable to make forward progress. While in this stalled state,
1669 all remaining space to free from the error-encountering filesystem is
1670 "temporarily leaked". Set this flag to cause it to ignore the EIO,
1671 permanently leak the space from indirect blocks that can not be read,
1672 and continue to free everything else that it can.
1674 The default, "stalling" behavior is useful if the storage partially
1675 fails (i.e. some but not all i/os fail), and then later recovers. In
1676 this case, we will be able to continue pool operations while it is
1677 partially failed, and when it recovers, we can continue to free the
1678 space, with no leaks. However, note that this case is actually
1681 Typically pools either (a) fail completely (but perhaps temporarily,
1682 e.g. a top-level vdev going offline), or (b) have localized,
1683 permanent errors (e.g. disk returns the wrong data due to bit flip or
1684 firmware bug). In case (a), this setting does not matter because the
1685 pool will be suspended and the sync thread will not be able to make
1686 forward progress regardless. In case (b), because the error is
1687 permanent, the best we can do is leak the minimum amount of space,
1688 which is what setting this flag will do. Therefore, it is reasonable
1689 for this flag to normally be set, but we chose the more conservative
1690 approach of not setting it, so that there is no possibility of
1691 leaking space in the "partial temporary" failure case.
1693 Default value: \fB0\fR.
1699 \fBzfs_free_min_time_ms\fR (int)
1702 During a \fBzfs destroy\fR operation using \fBfeature@async_destroy\fR a minimum
1703 of this much time will be spent working on freeing blocks per txg.
1705 Default value: \fB1,000\fR.
1711 \fBzfs_immediate_write_sz\fR (long)
1714 Largest data block to write to zil. Larger blocks will be treated as if the
1715 dataset being written to had the property setting \fBlogbias=throughput\fR.
1717 Default value: \fB32,768\fR.
1723 \fBzfs_initialize_value\fR (ulong)
1726 Pattern written to vdev free space by \fBzpool initialize\fR.
1728 Default value: \fB16,045,690,984,833,335,022\fR (0xdeadbeefdeadbeee).
1734 \fBzfs_lua_max_instrlimit\fR (ulong)
1737 The maximum execution time limit that can be set for a ZFS channel program,
1738 specified as a number of Lua instructions.
1740 Default value: \fB100,000,000\fR.
1746 \fBzfs_lua_max_memlimit\fR (ulong)
1749 The maximum memory limit that can be set for a ZFS channel program, specified
1752 Default value: \fB104,857,600\fR.
1758 \fBzfs_max_dataset_nesting\fR (int)
1761 The maximum depth of nested datasets. This value can be tuned temporarily to
1762 fix existing datasets that exceed the predefined limit.
1764 Default value: \fB50\fR.
1770 \fBzfs_max_recordsize\fR (int)
1773 We currently support block sizes from 512 bytes to 16MB. The benefits of
1774 larger blocks, and thus larger I/O, need to be weighed against the cost of
1775 COWing a giant block to modify one byte. Additionally, very large blocks
1776 can have an impact on i/o latency, and also potentially on the memory
1777 allocator. Therefore, we do not allow the recordsize to be set larger than
1778 zfs_max_recordsize (default 1MB). Larger blocks can be created by changing
1779 this tunable, and pools with larger blocks can always be imported and used,
1780 regardless of this setting.
1782 Default value: \fB1,048,576\fR.
1788 \fBzfs_metaslab_fragmentation_threshold\fR (int)
1791 Allow metaslabs to keep their active state as long as their fragmentation
1792 percentage is less than or equal to this value. An active metaslab that
1793 exceeds this threshold will no longer keep its active status allowing
1794 better metaslabs to be selected.
1796 Default value: \fB70\fR.
1802 \fBzfs_mg_fragmentation_threshold\fR (int)
1805 Metaslab groups are considered eligible for allocations if their
1806 fragmentation metric (measured as a percentage) is less than or equal to
1807 this value. If a metaslab group exceeds this threshold then it will be
1808 skipped unless all metaslab groups within the metaslab class have also
1809 crossed this threshold.
1811 Default value: \fB85\fR.
1817 \fBzfs_mg_noalloc_threshold\fR (int)
1820 Defines a threshold at which metaslab groups should be eligible for
1821 allocations. The value is expressed as a percentage of free space
1822 beyond which a metaslab group is always eligible for allocations.
1823 If a metaslab group's free space is less than or equal to the
1824 threshold, the allocator will avoid allocating to that group
1825 unless all groups in the pool have reached the threshold. Once all
1826 groups have reached the threshold, all groups are allowed to accept
1827 allocations. The default value of 0 disables the feature and causes
1828 all metaslab groups to be eligible for allocations.
1830 This parameter allows one to deal with pools having heavily imbalanced
1831 vdevs such as would be the case when a new vdev has been added.
1832 Setting the threshold to a non-zero percentage will stop allocations
1833 from being made to vdevs that aren't filled to the specified percentage
1834 and allow lesser filled vdevs to acquire more allocations than they
1835 otherwise would under the old \fBzfs_mg_alloc_failures\fR facility.
1837 Default value: \fB0\fR.
1843 \fBzfs_ddt_data_is_special\fR (int)
1846 If enabled, ZFS will place DDT data into the special allocation class.
1848 Default value: \fB1\fR.
1854 \fBzfs_user_indirect_is_special\fR (int)
1857 If enabled, ZFS will place user data (both file and zvol) indirect blocks
1858 into the special allocation class.
1860 Default value: \fB1\fR.
1866 \fBzfs_multihost_history\fR (int)
1869 Historical statistics for the last N multihost updates will be available in
1870 \fB/proc/spl/kstat/zfs/<pool>/multihost\fR
1872 Default value: \fB0\fR.
1878 \fBzfs_multihost_interval\fR (ulong)
1881 Used to control the frequency of multihost writes which are performed when the
1882 \fBmultihost\fR pool property is on. This is one factor used to determine the
1883 length of the activity check during import.
1885 The multihost write period is \fBzfs_multihost_interval / leaf-vdevs\fR
1886 milliseconds. On average a multihost write will be issued for each leaf vdev
1887 every \fBzfs_multihost_interval\fR milliseconds. In practice, the observed
1888 period can vary with the I/O load and this observed value is the delay which is
1889 stored in the uberblock.
1891 Default value: \fB1000\fR.
1897 \fBzfs_multihost_import_intervals\fR (uint)
1900 Used to control the duration of the activity test on import. Smaller values of
1901 \fBzfs_multihost_import_intervals\fR will reduce the import time but increase
1902 the risk of failing to detect an active pool. The total activity check time is
1903 never allowed to drop below one second.
1905 On import the activity check waits a minimum amount of time determined by
1906 \fBzfs_multihost_interval * zfs_multihost_import_intervals\fR, or the same
1907 product computed on the host which last had the pool imported (whichever is
1908 greater). The activity check time may be further extended if the value of mmp
1909 delay found in the best uberblock indicates actual multihost updates happened
1910 at longer intervals than \fBzfs_multihost_interval\fR. A minimum value of
1911 \fB100ms\fR is enforced.
1913 A value of 0 is ignored and treated as if it was set to 1.
1915 Default value: \fB20\fR.
1921 \fBzfs_multihost_fail_intervals\fR (uint)
1924 Controls the behavior of the pool when multihost write failures or delays are
1927 When \fBzfs_multihost_fail_intervals = 0\fR, multihost write failures or delays
1928 are ignored. The failures will still be reported to the ZED which depending on
1929 its configuration may take action such as suspending the pool or offlining a
1933 When \fBzfs_multihost_fail_intervals > 0\fR, the pool will be suspended if
1934 \fBzfs_multihost_fail_intervals * zfs_multihost_interval\fR milliseconds pass
1935 without a successful mmp write. This guarantees the activity test will see
1936 mmp writes if the pool is imported. A value of 1 is ignored and treated as
1937 if it was set to 2. This is necessary to prevent the pool from being suspended
1938 due to normal, small I/O latency variations.
1941 Default value: \fB10\fR.
1947 \fBzfs_no_scrub_io\fR (int)
1950 Set for no scrub I/O. This results in scrubs not actually scrubbing data and
1951 simply doing a metadata crawl of the pool instead.
1953 Use \fB1\fR for yes and \fB0\fR for no (default).
1959 \fBzfs_no_scrub_prefetch\fR (int)
1962 Set to disable block prefetching for scrubs.
1964 Use \fB1\fR for yes and \fB0\fR for no (default).
1970 \fBzfs_nocacheflush\fR (int)
1973 Disable cache flush operations on disks when writing. Setting this will
1974 cause pool corruption on power loss if a volatile out-of-order write cache
1977 Use \fB1\fR for yes and \fB0\fR for no (default).
1983 \fBzfs_nopwrite_enabled\fR (int)
1988 Use \fB1\fR for yes (default) and \fB0\fR to disable.
1994 \fBzfs_dmu_offset_next_sync\fR (int)
1997 Enable forcing txg sync to find holes. When enabled forces ZFS to act
1998 like prior versions when SEEK_HOLE or SEEK_DATA flags are used, which
1999 when a dnode is dirty causes txg's to be synced so that this data can be
2002 Use \fB1\fR for yes and \fB0\fR to disable (default).
2008 \fBzfs_pd_bytes_max\fR (int)
2011 The number of bytes which should be prefetched during a pool traversal
2012 (eg: \fBzfs send\fR or other data crawling operations)
2014 Default value: \fB52,428,800\fR.
2020 \fBzfs_per_txg_dirty_frees_percent \fR (ulong)
2023 Tunable to control percentage of dirtied indirect blocks from frees allowed
2024 into one TXG. After this threshold is crossed, additional frees will wait until
2026 A value of zero will disable this throttle.
2028 Default value: \fB5\fR, set to \fB0\fR to disable.
2034 \fBzfs_prefetch_disable\fR (int)
2037 This tunable disables predictive prefetch. Note that it leaves "prescient"
2038 prefetch (e.g. prefetch for zfs send) intact. Unlike predictive prefetch,
2039 prescient prefetch never issues i/os that end up not being needed, so it
2040 can't hurt performance.
2042 Use \fB1\fR for yes and \fB0\fR for no (default).
2048 \fBzfs_qat_checksum_disable\fR (int)
2051 This tunable disables qat hardware acceleration for sha256 checksums. It
2052 may be set after the zfs modules have been loaded to initialize the qat
2053 hardware as long as support is compiled in and the qat driver is present.
2055 Use \fB1\fR for yes and \fB0\fR for no (default).
2061 \fBzfs_qat_compress_disable\fR (int)
2064 This tunable disables qat hardware acceleration for gzip compression. It
2065 may be set after the zfs modules have been loaded to initialize the qat
2066 hardware as long as support is compiled in and the qat driver is present.
2068 Use \fB1\fR for yes and \fB0\fR for no (default).
2074 \fBzfs_qat_encrypt_disable\fR (int)
2077 This tunable disables qat hardware acceleration for AES-GCM encryption. It
2078 may be set after the zfs modules have been loaded to initialize the qat
2079 hardware as long as support is compiled in and the qat driver is present.
2081 Use \fB1\fR for yes and \fB0\fR for no (default).
2087 \fBzfs_read_chunk_size\fR (long)
2090 Bytes to read per chunk
2092 Default value: \fB1,048,576\fR.
2098 \fBzfs_read_history\fR (int)
2101 Historical statistics for the last N reads will be available in
2102 \fB/proc/spl/kstat/zfs/<pool>/reads\fR
2104 Default value: \fB0\fR (no data is kept).
2110 \fBzfs_read_history_hits\fR (int)
2113 Include cache hits in read history
2115 Use \fB1\fR for yes and \fB0\fR for no (default).
2121 \fBzfs_reconstruct_indirect_combinations_max\fR (int)
2124 If an indirect split block contains more than this many possible unique
2125 combinations when being reconstructed, consider it too computationally
2126 expensive to check them all. Instead, try at most
2127 \fBzfs_reconstruct_indirect_combinations_max\fR randomly-selected
2128 combinations each time the block is accessed. This allows all segment
2129 copies to participate fairly in the reconstruction when all combinations
2130 cannot be checked and prevents repeated use of one bad copy.
2132 Default value: \fB4096\fR.
2138 \fBzfs_recover\fR (int)
2141 Set to attempt to recover from fatal errors. This should only be used as a
2142 last resort, as it typically results in leaked space, or worse.
2144 Use \fB1\fR for yes and \fB0\fR for no (default).
2150 \fBzfs_removal_ignore_errors\fR (int)
2154 Ignore hard IO errors during device removal. When set, if a device encounters
2155 a hard IO error during the removal process the removal will not be cancelled.
2156 This can result in a normally recoverable block becoming permanently damaged
2157 and is not recommended. This should only be used as a last resort when the
2158 pool cannot be returned to a healthy state prior to removing the device.
2160 Default value: \fB0\fR.
2166 \fBzfs_resilver_min_time_ms\fR (int)
2169 Resilvers are processed by the sync thread. While resilvering it will spend
2170 at least this much time working on a resilver between txg flushes.
2172 Default value: \fB3,000\fR.
2178 \fBzfs_scan_ignore_errors\fR (int)
2181 If set to a nonzero value, remove the DTL (dirty time list) upon
2182 completion of a pool scan (scrub) even if there were unrepairable
2183 errors. It is intended to be used during pool repair or recovery to
2184 stop resilvering when the pool is next imported.
2186 Default value: \fB0\fR.
2192 \fBzfs_scrub_min_time_ms\fR (int)
2195 Scrubs are processed by the sync thread. While scrubbing it will spend
2196 at least this much time working on a scrub between txg flushes.
2198 Default value: \fB1,000\fR.
2204 \fBzfs_scan_checkpoint_intval\fR (int)
2207 To preserve progress across reboots the sequential scan algorithm periodically
2208 needs to stop metadata scanning and issue all the verifications I/Os to disk.
2209 The frequency of this flushing is determined by the
2210 \fBzfs_scan_checkpoint_intval\fR tunable.
2212 Default value: \fB7200\fR seconds (every 2 hours).
2218 \fBzfs_scan_fill_weight\fR (int)
2221 This tunable affects how scrub and resilver I/O segments are ordered. A higher
2222 number indicates that we care more about how filled in a segment is, while a
2223 lower number indicates we care more about the size of the extent without
2224 considering the gaps within a segment. This value is only tunable upon module
2225 insertion. Changing the value afterwards will have no affect on scrub or
2226 resilver performance.
2228 Default value: \fB3\fR.
2234 \fBzfs_scan_issue_strategy\fR (int)
2237 Determines the order that data will be verified while scrubbing or resilvering.
2238 If set to \fB1\fR, data will be verified as sequentially as possible, given the
2239 amount of memory reserved for scrubbing (see \fBzfs_scan_mem_lim_fact\fR). This
2240 may improve scrub performance if the pool's data is very fragmented. If set to
2241 \fB2\fR, the largest mostly-contiguous chunk of found data will be verified
2242 first. By deferring scrubbing of small segments, we may later find adjacent data
2243 to coalesce and increase the segment size. If set to \fB0\fR, zfs will use
2244 strategy \fB1\fR during normal verification and strategy \fB2\fR while taking a
2247 Default value: \fB0\fR.
2253 \fBzfs_scan_legacy\fR (int)
2256 A value of 0 indicates that scrubs and resilvers will gather metadata in
2257 memory before issuing sequential I/O. A value of 1 indicates that the legacy
2258 algorithm will be used where I/O is initiated as soon as it is discovered.
2259 Changing this value to 0 will not affect scrubs or resilvers that are already
2262 Default value: \fB0\fR.
2268 \fBzfs_scan_max_ext_gap\fR (int)
2271 Indicates the largest gap in bytes between scrub / resilver I/Os that will still
2272 be considered sequential for sorting purposes. Changing this value will not
2273 affect scrubs or resilvers that are already in progress.
2275 Default value: \fB2097152 (2 MB)\fR.
2281 \fBzfs_scan_mem_lim_fact\fR (int)
2284 Maximum fraction of RAM used for I/O sorting by sequential scan algorithm.
2285 This tunable determines the hard limit for I/O sorting memory usage.
2286 When the hard limit is reached we stop scanning metadata and start issuing
2287 data verification I/O. This is done until we get below the soft limit.
2289 Default value: \fB20\fR which is 5% of RAM (1/20).
2295 \fBzfs_scan_mem_lim_soft_fact\fR (int)
2298 The fraction of the hard limit used to determined the soft limit for I/O sorting
2299 by the sequential scan algorithm. When we cross this limit from bellow no action
2300 is taken. When we cross this limit from above it is because we are issuing
2301 verification I/O. In this case (unless the metadata scan is done) we stop
2302 issuing verification I/O and start scanning metadata again until we get to the
2305 Default value: \fB20\fR which is 5% of the hard limit (1/20).
2311 \fBzfs_scan_vdev_limit\fR (int)
2314 Maximum amount of data that can be concurrently issued at once for scrubs and
2315 resilvers per leaf device, given in bytes.
2317 Default value: \fB41943040\fR.
2323 \fBzfs_send_corrupt_data\fR (int)
2326 Allow sending of corrupt data (ignore read/checksum errors when sending data)
2328 Use \fB1\fR for yes and \fB0\fR for no (default).
2334 \fBzfs_send_queue_length\fR (int)
2337 The maximum number of bytes allowed in the \fBzfs send\fR queue. This value
2338 must be at least twice the maximum block size in use.
2340 Default value: \fB16,777,216\fR.
2346 \fBzfs_recv_queue_length\fR (int)
2350 The maximum number of bytes allowed in the \fBzfs receive\fR queue. This value
2351 must be at least twice the maximum block size in use.
2353 Default value: \fB16,777,216\fR.
2359 \fBzfs_sync_pass_deferred_free\fR (int)
2362 Flushing of data to disk is done in passes. Defer frees starting in this pass
2364 Default value: \fB2\fR.
2370 \fBzfs_spa_discard_memory_limit\fR (int)
2373 Maximum memory used for prefetching a checkpoint's space map on each
2374 vdev while discarding the checkpoint.
2376 Default value: \fB16,777,216\fR.
2382 \fBzfs_sync_pass_dont_compress\fR (int)
2385 Don't compress starting in this pass
2387 Default value: \fB5\fR.
2393 \fBzfs_sync_pass_rewrite\fR (int)
2396 Rewrite new block pointers starting in this pass
2398 Default value: \fB2\fR.
2404 \fBzfs_sync_taskq_batch_pct\fR (int)
2407 This controls the number of threads used by the dp_sync_taskq. The default
2408 value of 75% will create a maximum of one thread per cpu.
2410 Default value: \fB75\fR%.
2416 \fBzfs_trim_extent_bytes_max\fR (unsigned int)
2419 Maximum size of TRIM command. Ranges larger than this will be split in to
2420 chunks no larger than \fBzfs_trim_extent_bytes_max\fR bytes before being
2421 issued to the device.
2423 Default value: \fB134,217,728\fR.
2429 \fBzfs_trim_extent_bytes_min\fR (unsigned int)
2432 Minimum size of TRIM commands. TRIM ranges smaller than this will be skipped
2433 unless they're part of a larger range which was broken in to chunks. This is
2434 done because it's common for these small TRIMs to negatively impact overall
2435 performance. This value can be set to 0 to TRIM all unallocated space.
2437 Default value: \fB32,768\fR.
2443 \fBzfs_trim_metaslab_skip\fR (unsigned int)
2446 Skip uninitialized metaslabs during the TRIM process. This option is useful
2447 for pools constructed from large thinly-provisioned devices where TRIM
2448 operations are slow. As a pool ages an increasing fraction of the pools
2449 metaslabs will be initialized progressively degrading the usefulness of
2450 this option. This setting is stored when starting a manual TRIM and will
2451 persist for the duration of the requested TRIM.
2453 Default value: \fB0\fR.
2459 \fBzfs_trim_queue_limit\fR (unsigned int)
2462 Maximum number of queued TRIMs outstanding per leaf vdev. The number of
2463 concurrent TRIM commands issued to the device is controlled by the
2464 \fBzfs_vdev_trim_min_active\fR and \fBzfs_vdev_trim_max_active\fR module
2467 Default value: \fB10\fR.
2473 \fBzfs_trim_txg_batch\fR (unsigned int)
2476 The number of transaction groups worth of frees which should be aggregated
2477 before TRIM operations are issued to the device. This setting represents a
2478 trade-off between issuing larger, more efficient TRIM operations and the
2479 delay before the recently trimmed space is available for use by the device.
2481 Increasing this value will allow frees to be aggregated for a longer time.
2482 This will result is larger TRIM operations and potentially increased memory
2483 usage. Decreasing this value will have the opposite effect. The default
2484 value of 32 was determined to be a reasonable compromise.
2486 Default value: \fB32\fR.
2492 \fBzfs_txg_history\fR (int)
2495 Historical statistics for the last N txgs will be available in
2496 \fB/proc/spl/kstat/zfs/<pool>/txgs\fR
2498 Default value: \fB0\fR.
2504 \fBzfs_txg_timeout\fR (int)
2507 Flush dirty data to disk at least every N seconds (maximum txg duration)
2509 Default value: \fB5\fR.
2515 \fBzfs_vdev_aggregate_trim\fR (int)
2518 Allow TRIM I/Os to be aggregated. This is normally not helpful because
2519 the extents to be trimmed will have been already been aggregated by the
2520 metaslab. This option is provided for debugging and performance analysis.
2522 Default value: \fB0\fR.
2528 \fBzfs_vdev_aggregation_limit\fR (int)
2531 Max vdev I/O aggregation size
2533 Default value: \fB1,048,576\fR.
2539 \fBzfs_vdev_aggregation_limit_non_rotating\fR (int)
2542 Max vdev I/O aggregation size for non-rotating media
2544 Default value: \fB131,072\fR.
2550 \fBzfs_vdev_cache_bshift\fR (int)
2553 Shift size to inflate reads too
2555 Default value: \fB16\fR (effectively 65536).
2561 \fBzfs_vdev_cache_max\fR (int)
2564 Inflate reads smaller than this value to meet the \fBzfs_vdev_cache_bshift\fR
2567 Default value: \fB16384\fR.
2573 \fBzfs_vdev_cache_size\fR (int)
2576 Total size of the per-disk cache in bytes.
2578 Currently this feature is disabled as it has been found to not be helpful
2579 for performance and in some cases harmful.
2581 Default value: \fB0\fR.
2587 \fBzfs_vdev_mirror_rotating_inc\fR (int)
2590 A number by which the balancing algorithm increments the load calculation for
2591 the purpose of selecting the least busy mirror member when an I/O immediately
2592 follows its predecessor on rotational vdevs for the purpose of making decisions
2595 Default value: \fB0\fR.
2601 \fBzfs_vdev_mirror_rotating_seek_inc\fR (int)
2604 A number by which the balancing algorithm increments the load calculation for
2605 the purpose of selecting the least busy mirror member when an I/O lacks
2606 locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within
2607 this that are not immediately following the previous I/O are incremented by
2610 Default value: \fB5\fR.
2616 \fBzfs_vdev_mirror_rotating_seek_offset\fR (int)
2619 The maximum distance for the last queued I/O in which the balancing algorithm
2620 considers an I/O to have locality.
2621 See the section "ZFS I/O SCHEDULER".
2623 Default value: \fB1048576\fR.
2629 \fBzfs_vdev_mirror_non_rotating_inc\fR (int)
2632 A number by which the balancing algorithm increments the load calculation for
2633 the purpose of selecting the least busy mirror member on non-rotational vdevs
2634 when I/Os do not immediately follow one another.
2636 Default value: \fB0\fR.
2642 \fBzfs_vdev_mirror_non_rotating_seek_inc\fR (int)
2645 A number by which the balancing algorithm increments the load calculation for
2646 the purpose of selecting the least busy mirror member when an I/O lacks
2647 locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within
2648 this that are not immediately following the previous I/O are incremented by
2651 Default value: \fB1\fR.
2657 \fBzfs_vdev_read_gap_limit\fR (int)
2660 Aggregate read I/O operations if the gap on-disk between them is within this
2663 Default value: \fB32,768\fR.
2669 \fBzfs_vdev_scheduler\fR (charp)
2672 Set the Linux I/O scheduler on whole disk vdevs to this scheduler. Valid options
2673 are noop, cfq, bfq & deadline
2675 Default value: \fBnoop\fR.
2681 \fBzfs_vdev_write_gap_limit\fR (int)
2684 Aggregate write I/O over gap
2686 Default value: \fB4,096\fR.
2692 \fBzfs_vdev_raidz_impl\fR (string)
2695 Parameter for selecting raidz parity implementation to use.
2697 Options marked (always) below may be selected on module load as they are
2698 supported on all systems.
2699 The remaining options may only be set after the module is loaded, as they
2700 are available only if the implementations are compiled in and supported
2701 on the running system.
2703 Once the module is loaded, the content of
2704 /sys/module/zfs/parameters/zfs_vdev_raidz_impl will show available options
2705 with the currently selected one enclosed in [].
2706 Possible options are:
2707 fastest - (always) implementation selected using built-in benchmark
2708 original - (always) original raidz implementation
2709 scalar - (always) scalar raidz implementation
2710 sse2 - implementation using SSE2 instruction set (64bit x86 only)
2711 ssse3 - implementation using SSSE3 instruction set (64bit x86 only)
2712 avx2 - implementation using AVX2 instruction set (64bit x86 only)
2713 avx512f - implementation using AVX512F instruction set (64bit x86 only)
2714 avx512bw - implementation using AVX512F & AVX512BW instruction sets (64bit x86 only)
2715 aarch64_neon - implementation using NEON (Aarch64/64 bit ARMv8 only)
2716 aarch64_neonx2 - implementation using NEON with more unrolling (Aarch64/64 bit ARMv8 only)
2718 Default value: \fBfastest\fR.
2724 \fBzfs_zevent_cols\fR (int)
2727 When zevents are logged to the console use this as the word wrap width.
2729 Default value: \fB80\fR.
2735 \fBzfs_zevent_console\fR (int)
2738 Log events to the console
2740 Use \fB1\fR for yes and \fB0\fR for no (default).
2746 \fBzfs_zevent_len_max\fR (int)
2749 Max event queue length. A value of 0 will result in a calculated value which
2750 increases with the number of CPUs in the system (minimum 64 events). Events
2751 in the queue can be viewed with the \fBzpool events\fR command.
2753 Default value: \fB0\fR.
2759 \fBzfs_zil_clean_taskq_maxalloc\fR (int)
2762 The maximum number of taskq entries that are allowed to be cached. When this
2763 limit is exceeded transaction records (itxs) will be cleaned synchronously.
2765 Default value: \fB1048576\fR.
2771 \fBzfs_zil_clean_taskq_minalloc\fR (int)
2774 The number of taskq entries that are pre-populated when the taskq is first
2775 created and are immediately available for use.
2777 Default value: \fB1024\fR.
2783 \fBzfs_zil_clean_taskq_nthr_pct\fR (int)
2786 This controls the number of threads used by the dp_zil_clean_taskq. The default
2787 value of 100% will create a maximum of one thread per cpu.
2789 Default value: \fB100\fR%.
2795 \fBzil_nocacheflush\fR (int)
2798 Disable the cache flush commands that are normally sent to the disk(s) by
2799 the ZIL after an LWB write has completed. Setting this will cause ZIL
2800 corruption on power loss if a volatile out-of-order write cache is enabled.
2802 Use \fB1\fR for yes and \fB0\fR for no (default).
2808 \fBzil_replay_disable\fR (int)
2811 Disable intent logging replay. Can be disabled for recovery from corrupted
2814 Use \fB1\fR for yes and \fB0\fR for no (default).
2820 \fBzil_slog_bulk\fR (ulong)
2823 Limit SLOG write size per commit executed with synchronous priority.
2824 Any writes above that will be executed with lower (asynchronous) priority
2825 to limit potential SLOG device abuse by single active ZIL writer.
2827 Default value: \fB786,432\fR.
2833 \fBzio_deadman_log_all\fR (int)
2836 If non-zero, the zio deadman will produce debugging messages (see
2837 \fBzfs_dbgmsg_enable\fR) for all zios, rather than only for leaf
2838 zios possessing a vdev. This is meant to be used by developers to gain
2839 diagnostic information for hang conditions which don't involve a mutex
2840 or other locking primitive; typically conditions in which a thread in
2841 the zio pipeline is looping indefinitely.
2843 Default value: \fB0\fR.
2849 \fBzio_decompress_fail_fraction\fR (int)
2852 If non-zero, this value represents the denominator of the probability that zfs
2853 should induce a decompression failure. For instance, for a 5% decompression
2854 failure rate, this value should be set to 20.
2856 Default value: \fB0\fR.
2862 \fBzio_slow_io_ms\fR (int)
2865 When an I/O operation takes more than \fBzio_slow_io_ms\fR milliseconds to
2866 complete is marked as a slow I/O. Each slow I/O causes a delay zevent. Slow
2867 I/O counters can be seen with "zpool status -s".
2870 Default value: \fB30,000\fR.
2876 \fBzio_dva_throttle_enabled\fR (int)
2879 Throttle block allocations in the I/O pipeline. This allows for
2880 dynamic allocation distribution when devices are imbalanced.
2881 When enabled, the maximum number of pending allocations per top-level vdev
2882 is limited by \fBzfs_vdev_queue_depth_pct\fR.
2884 Default value: \fB1\fR.
2890 \fBzio_requeue_io_start_cut_in_line\fR (int)
2893 Prioritize requeued I/O
2895 Default value: \fB0\fR.
2901 \fBzio_taskq_batch_pct\fR (uint)
2904 Percentage of online CPUs (or CPU cores, etc) which will run a worker thread
2905 for I/O. These workers are responsible for I/O work such as compression and
2906 checksum calculations. Fractional number of CPUs will be rounded down.
2908 The default value of 75 was chosen to avoid using all CPUs which can result in
2909 latency issues and inconsistent application performance, especially when high
2910 compression is enabled.
2912 Default value: \fB75\fR.
2918 \fBzvol_inhibit_dev\fR (uint)
2921 Do not create zvol device nodes. This may slightly improve startup time on
2922 systems with a very large number of zvols.
2924 Use \fB1\fR for yes and \fB0\fR for no (default).
2930 \fBzvol_major\fR (uint)
2933 Major number for zvol block devices
2935 Default value: \fB230\fR.
2941 \fBzvol_max_discard_blocks\fR (ulong)
2944 Discard (aka TRIM) operations done on zvols will be done in batches of this
2945 many blocks, where block size is determined by the \fBvolblocksize\fR property
2948 Default value: \fB16,384\fR.
2954 \fBzvol_prefetch_bytes\fR (uint)
2957 When adding a zvol to the system prefetch \fBzvol_prefetch_bytes\fR
2958 from the start and end of the volume. Prefetching these regions
2959 of the volume is desirable because they are likely to be accessed
2960 immediately by \fBblkid(8)\fR or by the kernel scanning for a partition
2963 Default value: \fB131,072\fR.
2969 \fBzvol_request_sync\fR (uint)
2972 When processing I/O requests for a zvol submit them synchronously. This
2973 effectively limits the queue depth to 1 for each I/O submitter. When set
2974 to 0 requests are handled asynchronously by a thread pool. The number of
2975 requests which can be handled concurrently is controller by \fBzvol_threads\fR.
2977 Default value: \fB0\fR.
2983 \fBzvol_threads\fR (uint)
2986 Max number of threads which can handle zvol I/O requests concurrently.
2988 Default value: \fB32\fR.
2994 \fBzvol_volmode\fR (uint)
2997 Defines zvol block devices behaviour when \fBvolmode\fR is set to \fBdefault\fR.
2998 Valid values are \fB1\fR (full), \fB2\fR (dev) and \fB3\fR (none).
3000 Default value: \fB1\fR.
3003 .SH ZFS I/O SCHEDULER
3004 ZFS issues I/O operations to leaf vdevs to satisfy and complete I/Os.
3005 The I/O scheduler determines when and in what order those operations are
3006 issued. The I/O scheduler divides operations into five I/O classes
3007 prioritized in the following order: sync read, sync write, async read,
3008 async write, and scrub/resilver. Each queue defines the minimum and
3009 maximum number of concurrent operations that may be issued to the
3010 device. In addition, the device has an aggregate maximum,
3011 \fBzfs_vdev_max_active\fR. Note that the sum of the per-queue minimums
3012 must not exceed the aggregate maximum. If the sum of the per-queue
3013 maximums exceeds the aggregate maximum, then the number of active I/Os
3014 may reach \fBzfs_vdev_max_active\fR, in which case no further I/Os will
3015 be issued regardless of whether all per-queue minimums have been met.
3017 For many physical devices, throughput increases with the number of
3018 concurrent operations, but latency typically suffers. Further, physical
3019 devices typically have a limit at which more concurrent operations have no
3020 effect on throughput or can actually cause it to decrease.
3022 The scheduler selects the next operation to issue by first looking for an
3023 I/O class whose minimum has not been satisfied. Once all are satisfied and
3024 the aggregate maximum has not been hit, the scheduler looks for classes
3025 whose maximum has not been satisfied. Iteration through the I/O classes is
3026 done in the order specified above. No further operations are issued if the
3027 aggregate maximum number of concurrent operations has been hit or if there
3028 are no operations queued for an I/O class that has not hit its maximum.
3029 Every time an I/O is queued or an operation completes, the I/O scheduler
3030 looks for new operations to issue.
3032 In general, smaller max_active's will lead to lower latency of synchronous
3033 operations. Larger max_active's may lead to higher overall throughput,
3034 depending on underlying storage.
3036 The ratio of the queues' max_actives determines the balance of performance
3037 between reads, writes, and scrubs. E.g., increasing
3038 \fBzfs_vdev_scrub_max_active\fR will cause the scrub or resilver to complete
3039 more quickly, but reads and writes to have higher latency and lower throughput.
3041 All I/O classes have a fixed maximum number of outstanding operations
3042 except for the async write class. Asynchronous writes represent the data
3043 that is committed to stable storage during the syncing stage for
3044 transaction groups. Transaction groups enter the syncing state
3045 periodically so the number of queued async writes will quickly burst up
3046 and then bleed down to zero. Rather than servicing them as quickly as
3047 possible, the I/O scheduler changes the maximum number of active async
3048 write I/Os according to the amount of dirty data in the pool. Since
3049 both throughput and latency typically increase with the number of
3050 concurrent operations issued to physical devices, reducing the
3051 burstiness in the number of concurrent operations also stabilizes the
3052 response time of operations from other -- and in particular synchronous
3053 -- queues. In broad strokes, the I/O scheduler will issue more
3054 concurrent operations from the async write queue as there's more dirty
3059 The number of concurrent operations issued for the async write I/O class
3060 follows a piece-wise linear function defined by a few adjustable points.
3063 | o---------| <-- zfs_vdev_async_write_max_active
3070 |-------o | | <-- zfs_vdev_async_write_min_active
3071 0|_______^______|_________|
3072 0% | | 100% of zfs_dirty_data_max
3074 | `-- zfs_vdev_async_write_active_max_dirty_percent
3075 `--------- zfs_vdev_async_write_active_min_dirty_percent
3078 Until the amount of dirty data exceeds a minimum percentage of the dirty
3079 data allowed in the pool, the I/O scheduler will limit the number of
3080 concurrent operations to the minimum. As that threshold is crossed, the
3081 number of concurrent operations issued increases linearly to the maximum at
3082 the specified maximum percentage of the dirty data allowed in the pool.
3084 Ideally, the amount of dirty data on a busy pool will stay in the sloped
3085 part of the function between \fBzfs_vdev_async_write_active_min_dirty_percent\fR
3086 and \fBzfs_vdev_async_write_active_max_dirty_percent\fR. If it exceeds the
3087 maximum percentage, this indicates that the rate of incoming data is
3088 greater than the rate that the backend storage can handle. In this case, we
3089 must further throttle incoming writes, as described in the next section.
3091 .SH ZFS TRANSACTION DELAY
3092 We delay transactions when we've determined that the backend storage
3093 isn't able to accommodate the rate of incoming writes.
3095 If there is already a transaction waiting, we delay relative to when
3096 that transaction will finish waiting. This way the calculated delay time
3097 is independent of the number of threads concurrently executing
3100 If we are the only waiter, wait relative to when the transaction
3101 started, rather than the current time. This credits the transaction for
3102 "time already served", e.g. reading indirect blocks.
3104 The minimum time for a transaction to take is calculated as:
3106 min_time = zfs_delay_scale * (dirty - min) / (max - dirty)
3107 min_time is then capped at 100 milliseconds.
3110 The delay has two degrees of freedom that can be adjusted via tunables. The
3111 percentage of dirty data at which we start to delay is defined by
3112 \fBzfs_delay_min_dirty_percent\fR. This should typically be at or above
3113 \fBzfs_vdev_async_write_active_max_dirty_percent\fR so that we only start to
3114 delay after writing at full speed has failed to keep up with the incoming write
3115 rate. The scale of the curve is defined by \fBzfs_delay_scale\fR. Roughly speaking,
3116 this variable determines the amount of delay at the midpoint of the curve.
3120 10ms +-------------------------------------------------------------*+
3136 2ms + (midpoint) * +
3139 | zfs_delay_scale ----------> ******** |
3140 0 +-------------------------------------*********----------------+
3141 0% <- zfs_dirty_data_max -> 100%
3144 Note that since the delay is added to the outstanding time remaining on the
3145 most recent transaction, the delay is effectively the inverse of IOPS.
3146 Here the midpoint of 500us translates to 2000 IOPS. The shape of the curve
3147 was chosen such that small changes in the amount of accumulated dirty data
3148 in the first 3/4 of the curve yield relatively small differences in the
3151 The effects can be easier to understand when the amount of delay is
3152 represented on a log scale:
3156 100ms +-------------------------------------------------------------++
3165 + zfs_delay_scale ----------> ***** +
3176 +--------------------------------------------------------------+
3177 0% <- zfs_dirty_data_max -> 100%
3180 Note here that only as the amount of dirty data approaches its limit does
3181 the delay start to increase rapidly. The goal of a properly tuned system
3182 should be to keep the amount of dirty data out of that range by first
3183 ensuring that the appropriate limits are set for the I/O scheduler to reach
3184 optimal throughput on the backend storage, and then by changing the value
3185 of \fBzfs_delay_scale\fR to increase the steepness of the curve.