2 .\" Copyright (c) 2013 by Turbo Fredriksson <turbo@bayour.com>. All rights reserved.
3 .\" Copyright (c) 2017 Datto Inc.
4 .\" Copyright (c) 2018 by Delphix. All rights reserved.
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 "Oct 28, 2017"
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_cache_hiwater_pct\fR (uint)
50 The percentage over \fBdbuf_cache_max_bytes\fR when dbufs must be evicted
53 Default value: \fB10\fR%.
59 \fBdbuf_cache_lowater_pct\fR (uint)
62 The percentage below \fBdbuf_cache_max_bytes\fR when the evict thread stops
65 Default value: \fB10\fR%.
71 \fBdbuf_cache_shift\fR (int)
74 Set the size of the dbuf cache, \fBdbuf_cache_max_bytes\fR, to a log2 fraction
75 of the target arc size.
77 Default value: \fB5\fR.
83 \fBignore_hole_birth\fR (int)
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.
90 Use \fB1\fR for on (default) and \fB0\fR for off.
96 \fBl2arc_feed_again\fR (int)
99 Turbo L2ARC warm-up. When the L2ARC is cold the fill interval will be set as
102 Use \fB1\fR for yes (default) and \fB0\fR to disable.
108 \fBl2arc_feed_min_ms\fR (ulong)
111 Min feed interval in milliseconds. Requires \fBl2arc_feed_again=1\fR and only
112 applicable in related situations.
114 Default value: \fB200\fR.
120 \fBl2arc_feed_secs\fR (ulong)
123 Seconds between L2ARC writing
125 Default value: \fB1\fR.
131 \fBl2arc_headroom\fR (ulong)
134 How far through the ARC lists to search for L2ARC cacheable content, expressed
135 as a multiplier of \fBl2arc_write_max\fR
137 Default value: \fB2\fR.
143 \fBl2arc_headroom_boost\fR (ulong)
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.
149 Default value: \fB200\fR%.
155 \fBl2arc_noprefetch\fR (int)
158 Do not write buffers to L2ARC if they were prefetched but not used by
161 Use \fB1\fR for yes (default) and \fB0\fR to disable.
167 \fBl2arc_norw\fR (int)
170 No reads during writes
172 Use \fB1\fR for yes and \fB0\fR for no (default).
178 \fBl2arc_write_boost\fR (ulong)
181 Cold L2ARC devices will have \fBl2arc_write_max\fR increased by this amount
182 while they remain cold.
184 Default value: \fB8,388,608\fR.
190 \fBl2arc_write_max\fR (ulong)
193 Max write bytes per interval
195 Default value: \fB8,388,608\fR.
201 \fBmetaslab_aliquot\fR (ulong)
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.
209 Default value: \fB524,288\fR.
215 \fBmetaslab_bias_enabled\fR (int)
218 Enable metaslab group biasing based on its vdev's over- or under-utilization
219 relative to the pool.
221 Use \fB1\fR for yes (default) and \fB0\fR for no.
227 \fBmetaslab_force_ganging\fR (ulong)
230 Make some blocks above a certain size be gang blocks. This option is used
231 by the test suite to facilitate testing.
233 Default value: \fB16,777,217\fR.
239 \fBzfs_metaslab_segment_weight_enabled\fR (int)
242 Enable/disable segment-based metaslab selection.
244 Use \fB1\fR for yes (default) and \fB0\fR for no.
250 \fBzfs_metaslab_switch_threshold\fR (int)
253 When using segment-based metaslab selection, continue allocating
254 from the active metaslab until \fBzfs_metaslab_switch_threshold\fR
255 worth of buckets have been exhausted.
257 Default value: \fB2\fR.
263 \fBmetaslab_debug_load\fR (int)
266 Load all metaslabs during pool import.
268 Use \fB1\fR for yes and \fB0\fR for no (default).
274 \fBmetaslab_debug_unload\fR (int)
277 Prevent metaslabs from being unloaded.
279 Use \fB1\fR for yes and \fB0\fR for no (default).
285 \fBmetaslab_fragmentation_factor_enabled\fR (int)
288 Enable use of the fragmentation metric in computing metaslab weights.
290 Use \fB1\fR for yes (default) and \fB0\fR for no.
296 \fBvdev_max_ms_count\fR (int)
299 When a vdev is added target this number of metaslabs per top-level vdev.
301 Default value: \fB200\fR.
307 \fBvdev_min_ms_count\fR (int)
310 Minimum number of metaslabs to create in a top-level vdev.
312 Default value: \fB16\fR.
318 \fBvdev_ms_count_limit\fR (int)
321 Practical upper limit of total metaslabs per top-level vdev.
323 Default value: \fB131,072\fR.
329 \fBmetaslab_preload_enabled\fR (int)
332 Enable metaslab group preloading.
334 Use \fB1\fR for yes (default) and \fB0\fR for no.
340 \fBmetaslab_lba_weighting_enabled\fR (int)
343 Give more weight to metaslabs with lower LBAs, assuming they have
344 greater bandwidth as is typically the case on a modern constant
345 angular velocity disk drive.
347 Use \fB1\fR for yes (default) and \fB0\fR for no.
353 \fBspa_config_path\fR (charp)
358 Default value: \fB/etc/zfs/zpool.cache\fR.
364 \fBspa_asize_inflation\fR (int)
367 Multiplication factor used to estimate actual disk consumption from the
368 size of data being written. The default value is a worst case estimate,
369 but lower values may be valid for a given pool depending on its
370 configuration. Pool administrators who understand the factors involved
371 may wish to specify a more realistic inflation factor, particularly if
372 they operate close to quota or capacity limits.
374 Default value: \fB24\fR.
380 \fBspa_load_print_vdev_tree\fR (int)
383 Whether to print the vdev tree in the debugging message buffer during pool import.
384 Use 0 to disable and 1 to enable.
386 Default value: \fB0\fR.
392 \fBspa_load_verify_data\fR (int)
395 Whether to traverse data blocks during an "extreme rewind" (\fB-X\fR)
396 import. Use 0 to disable and 1 to enable.
398 An extreme rewind import normally performs a full traversal of all
399 blocks in the pool for verification. If this parameter is set to 0,
400 the traversal skips non-metadata blocks. It can be toggled once the
401 import has started to stop or start the traversal of non-metadata blocks.
403 Default value: \fB1\fR.
409 \fBspa_load_verify_metadata\fR (int)
412 Whether to traverse blocks during an "extreme rewind" (\fB-X\fR)
413 pool import. Use 0 to disable and 1 to enable.
415 An extreme rewind import normally performs a full traversal of all
416 blocks in the pool for verification. If this parameter is set to 0,
417 the traversal is not performed. It can be toggled once the import has
418 started to stop or start the traversal.
420 Default value: \fB1\fR.
426 \fBspa_load_verify_maxinflight\fR (int)
429 Maximum concurrent I/Os during the traversal performed during an "extreme
430 rewind" (\fB-X\fR) pool import.
432 Default value: \fB10000\fR.
438 \fBspa_slop_shift\fR (int)
441 Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space
442 in the pool to be consumed. This ensures that we don't run the pool
443 completely out of space, due to unaccounted changes (e.g. to the MOS).
444 It also limits the worst-case time to allocate space. If we have
445 less than this amount of free space, most ZPL operations (e.g. write,
446 create) will return ENOSPC.
448 Default value: \fB5\fR.
454 \fBvdev_removal_max_span\fR (int)
457 During top-level vdev removal, chunks of data are copied from the vdev
458 which may include free space in order to trade bandwidth for IOPS.
459 This parameter determines the maximum span of free space (in bytes)
460 which will be included as "unnecessary" data in a chunk of copied data.
462 The default value here was chosen to align with
463 \fBzfs_vdev_read_gap_limit\fR, which is a similar concept when doing
464 regular reads (but there's no reason it has to be the same).
466 Default value: \fB32,768\fR.
472 \fBzfetch_array_rd_sz\fR (ulong)
475 If prefetching is enabled, disable prefetching for reads larger than this size.
477 Default value: \fB1,048,576\fR.
483 \fBzfetch_max_distance\fR (uint)
486 Max bytes to prefetch per stream (default 8MB).
488 Default value: \fB8,388,608\fR.
494 \fBzfetch_max_streams\fR (uint)
497 Max number of streams per zfetch (prefetch streams per file).
499 Default value: \fB8\fR.
505 \fBzfetch_min_sec_reap\fR (uint)
508 Min time before an active prefetch stream can be reclaimed
510 Default value: \fB2\fR.
516 \fBzfs_arc_dnode_limit\fR (ulong)
519 When the number of bytes consumed by dnodes in the ARC exceeds this number of
520 bytes, try to unpin some of it in response to demand for non-metadata. This
521 value acts as a ceiling to the amount of dnode metadata, and defaults to 0 which
522 indicates that a percent which is based on \fBzfs_arc_dnode_limit_percent\fR of
523 the ARC meta buffers that may be used for dnodes.
525 See also \fBzfs_arc_meta_prune\fR which serves a similar purpose but is used
526 when the amount of metadata in the ARC exceeds \fBzfs_arc_meta_limit\fR rather
527 than in response to overall demand for non-metadata.
530 Default value: \fB0\fR.
536 \fBzfs_arc_dnode_limit_percent\fR (ulong)
539 Percentage that can be consumed by dnodes of ARC meta buffers.
541 See also \fBzfs_arc_dnode_limit\fR which serves a similar purpose but has a
542 higher priority if set to nonzero value.
544 Default value: \fB10\fR%.
550 \fBzfs_arc_dnode_reduce_percent\fR (ulong)
553 Percentage of ARC dnodes to try to scan in response to demand for non-metadata
554 when the number of bytes consumed by dnodes exceeds \fBzfs_arc_dnode_limit\fR.
557 Default value: \fB10\fR% of the number of dnodes in the ARC.
563 \fBzfs_arc_average_blocksize\fR (int)
566 The ARC's buffer hash table is sized based on the assumption of an average
567 block size of \fBzfs_arc_average_blocksize\fR (default 8K). This works out
568 to roughly 1MB of hash table per 1GB of physical memory with 8-byte pointers.
569 For configurations with a known larger average block size this value can be
570 increased to reduce the memory footprint.
573 Default value: \fB8192\fR.
579 \fBzfs_arc_evict_batch_limit\fR (int)
582 Number ARC headers to evict per sub-list before proceeding to another sub-list.
583 This batch-style operation prevents entire sub-lists from being evicted at once
584 but comes at a cost of additional unlocking and locking.
586 Default value: \fB10\fR.
592 \fBzfs_arc_grow_retry\fR (int)
595 If set to a non zero value, it will replace the arc_grow_retry value with this value.
596 The arc_grow_retry value (default 5) is the number of seconds the ARC will wait before
597 trying to resume growth after a memory pressure event.
599 Default value: \fB0\fR.
605 \fBzfs_arc_lotsfree_percent\fR (int)
608 Throttle I/O when free system memory drops below this percentage of total
609 system memory. Setting this value to 0 will disable the throttle.
611 Default value: \fB10\fR%.
617 \fBzfs_arc_max\fR (ulong)
620 Max arc size of ARC in bytes. If set to 0 then it will consume 1/2 of system
621 RAM. This value must be at least 67108864 (64 megabytes).
623 This value can be changed dynamically with some caveats. It cannot be set back
624 to 0 while running and reducing it below the current ARC size will not cause
625 the ARC to shrink without memory pressure to induce shrinking.
627 Default value: \fB0\fR.
633 \fBzfs_arc_meta_adjust_restarts\fR (ulong)
636 The number of restart passes to make while scanning the ARC attempting
637 the free buffers in order to stay below the \fBzfs_arc_meta_limit\fR.
638 This value should not need to be tuned but is available to facilitate
639 performance analysis.
641 Default value: \fB4096\fR.
647 \fBzfs_arc_meta_limit\fR (ulong)
650 The maximum allowed size in bytes that meta data buffers are allowed to
651 consume in the ARC. When this limit is reached meta data buffers will
652 be reclaimed even if the overall arc_c_max has not been reached. This
653 value defaults to 0 which indicates that a percent which is based on
654 \fBzfs_arc_meta_limit_percent\fR of the ARC may be used for meta data.
656 This value my be changed dynamically except that it cannot be set back to 0
657 for a specific percent of the ARC; it must be set to an explicit value.
659 Default value: \fB0\fR.
665 \fBzfs_arc_meta_limit_percent\fR (ulong)
668 Percentage of ARC buffers that can be used for meta data.
670 See also \fBzfs_arc_meta_limit\fR which serves a similar purpose but has a
671 higher priority if set to nonzero value.
674 Default value: \fB75\fR%.
680 \fBzfs_arc_meta_min\fR (ulong)
683 The minimum allowed size in bytes that meta data buffers may consume in
684 the ARC. This value defaults to 0 which disables a floor on the amount
685 of the ARC devoted meta data.
687 Default value: \fB0\fR.
693 \fBzfs_arc_meta_prune\fR (int)
696 The number of dentries and inodes to be scanned looking for entries
697 which can be dropped. This may be required when the ARC reaches the
698 \fBzfs_arc_meta_limit\fR because dentries and inodes can pin buffers
699 in the ARC. Increasing this value will cause to dentry and inode caches
700 to be pruned more aggressively. Setting this value to 0 will disable
701 pruning the inode and dentry caches.
703 Default value: \fB10,000\fR.
709 \fBzfs_arc_meta_strategy\fR (int)
712 Define the strategy for ARC meta data buffer eviction (meta reclaim strategy).
713 A value of 0 (META_ONLY) will evict only the ARC meta data buffers.
714 A value of 1 (BALANCED) indicates that additional data buffers may be evicted if
715 that is required to in order to evict the required number of meta data buffers.
717 Default value: \fB1\fR.
723 \fBzfs_arc_min\fR (ulong)
726 Min arc size of ARC in bytes. If set to 0 then arc_c_min will default to
727 consuming the larger of 32M or 1/32 of total system memory.
729 Default value: \fB0\fR.
735 \fBzfs_arc_min_prefetch_ms\fR (int)
738 Minimum time prefetched blocks are locked in the ARC, specified in ms.
739 A value of \fB0\fR will default to 1000 ms.
741 Default value: \fB0\fR.
747 \fBzfs_arc_min_prescient_prefetch_ms\fR (int)
750 Minimum time "prescient prefetched" blocks are locked in the ARC, specified
751 in ms. These blocks are meant to be prefetched fairly aggresively ahead of
752 the code that may use them. A value of \fB0\fR will default to 6000 ms.
754 Default value: \fB0\fR.
760 \fBzfs_max_missing_tvds\fR (int)
763 Number of missing top-level vdevs which will be allowed during
764 pool import (only in read-only mode).
766 Default value: \fB0\fR
772 \fBzfs_multilist_num_sublists\fR (int)
775 To allow more fine-grained locking, each ARC state contains a series
776 of lists for both data and meta data objects. Locking is performed at
777 the level of these "sub-lists". This parameters controls the number of
778 sub-lists per ARC state, and also applies to other uses of the
779 multilist data structure.
781 Default value: \fB4\fR or the number of online CPUs, whichever is greater
787 \fBzfs_arc_overflow_shift\fR (int)
790 The ARC size is considered to be overflowing if it exceeds the current
791 ARC target size (arc_c) by a threshold determined by this parameter.
792 The threshold is calculated as a fraction of arc_c using the formula
793 "arc_c >> \fBzfs_arc_overflow_shift\fR".
795 The default value of 8 causes the ARC to be considered to be overflowing
796 if it exceeds the target size by 1/256th (0.3%) of the target size.
798 When the ARC is overflowing, new buffer allocations are stalled until
799 the reclaim thread catches up and the overflow condition no longer exists.
801 Default value: \fB8\fR.
808 \fBzfs_arc_p_min_shift\fR (int)
811 If set to a non zero value, this will update arc_p_min_shift (default 4)
813 arc_p_min_shift is used to shift of arc_c for calculating both min and max
816 Default value: \fB0\fR.
822 \fBzfs_arc_p_dampener_disable\fR (int)
825 Disable arc_p adapt dampener
827 Use \fB1\fR for yes (default) and \fB0\fR to disable.
833 \fBzfs_arc_shrink_shift\fR (int)
836 If set to a non zero value, this will update arc_shrink_shift (default 7)
839 Default value: \fB0\fR.
845 \fBzfs_arc_pc_percent\fR (uint)
848 Percent of pagecache to reclaim arc to
850 This tunable allows ZFS arc to play more nicely with the kernel's LRU
851 pagecache. It can guarantee that the arc size won't collapse under scanning
852 pressure on the pagecache, yet still allows arc to be reclaimed down to
853 zfs_arc_min if necessary. This value is specified as percent of pagecache
854 size (as measured by NR_FILE_PAGES) where that percent may exceed 100. This
855 only operates during memory pressure/reclaim.
857 Default value: \fB0\fR% (disabled).
863 \fBzfs_arc_sys_free\fR (ulong)
866 The target number of bytes the ARC should leave as free memory on the system.
867 Defaults to the larger of 1/64 of physical memory or 512K. Setting this
868 option to a non-zero value will override the default.
870 Default value: \fB0\fR.
876 \fBzfs_autoimport_disable\fR (int)
879 Disable pool import at module load by ignoring the cache file (typically \fB/etc/zfs/zpool.cache\fR).
881 Use \fB1\fR for yes (default) and \fB0\fR for no.
887 \fBzfs_checksums_per_second\fR (int)
890 Rate limit checksum events to this many per second. Note that this should
891 not be set below the zed thresholds (currently 10 checksums over 10 sec)
892 or else zed may not trigger any action.
900 \fBzfs_commit_timeout_pct\fR (int)
903 This controls the amount of time that a ZIL block (lwb) will remain "open"
904 when it isn't "full", and it has a thread waiting for it to be committed to
905 stable storage. The timeout is scaled based on a percentage of the last lwb
906 latency to avoid significantly impacting the latency of each individual
907 transaction record (itx).
909 Default value: \fB5\fR%.
915 \fBzfs_condense_indirect_vdevs_enable\fR (int)
918 Enable condensing indirect vdev mappings. When set to a non-zero value,
919 attempt to condense indirect vdev mappings if the mapping uses more than
920 \fBzfs_condense_min_mapping_bytes\fR bytes of memory and if the obsolete
921 space map object uses more than \fBzfs_condense_max_obsolete_bytes\fR
922 bytes on-disk. The condensing process is an attempt to save memory by
923 removing obsolete mappings.
925 Default value: \fB1\fR.
931 \fBzfs_condense_max_obsolete_bytes\fR (ulong)
934 Only attempt to condense indirect vdev mappings if the on-disk size
935 of the obsolete space map object is greater than this number of bytes
936 (see \fBfBzfs_condense_indirect_vdevs_enable\fR).
938 Default value: \fB1,073,741,824\fR.
944 \fBzfs_condense_min_mapping_bytes\fR (ulong)
947 Minimum size vdev mapping to attempt to condense (see
948 \fBzfs_condense_indirect_vdevs_enable\fR).
950 Default value: \fB131,072\fR.
956 \fBzfs_dbgmsg_enable\fR (int)
959 Internally ZFS keeps a small log to facilitate debugging. By default the log
960 is disabled, to enable it set this option to 1. The contents of the log can
961 be accessed by reading the /proc/spl/kstat/zfs/dbgmsg file. Writing 0 to
962 this proc file clears the log.
964 Default value: \fB0\fR.
970 \fBzfs_dbgmsg_maxsize\fR (int)
973 The maximum size in bytes of the internal ZFS debug log.
975 Default value: \fB4M\fR.
981 \fBzfs_dbuf_state_index\fR (int)
984 This feature is currently unused. It is normally used for controlling what
985 reporting is available under /proc/spl/kstat/zfs.
987 Default value: \fB0\fR.
993 \fBzfs_deadman_enabled\fR (int)
996 When a pool sync operation takes longer than \fBzfs_deadman_synctime_ms\fR
997 milliseconds, or when an individual I/O takes longer than
998 \fBzfs_deadman_ziotime_ms\fR milliseconds, then the operation is considered to
999 be "hung". If \fBzfs_deadman_enabled\fR is set then the deadman behavior is
1000 invoked as described by the \fBzfs_deadman_failmode\fR module option.
1001 By default the deadman is enabled and configured to \fBwait\fR which results
1002 in "hung" I/Os only being logged. The deadman is automatically disabled
1003 when a pool gets suspended.
1005 Default value: \fB1\fR.
1011 \fBzfs_deadman_failmode\fR (charp)
1014 Controls the failure behavior when the deadman detects a "hung" I/O. Valid
1015 values are \fBwait\fR, \fBcontinue\fR, and \fBpanic\fR.
1017 \fBwait\fR - Wait for a "hung" I/O to complete. For each "hung" I/O a
1018 "deadman" event will be posted describing that I/O.
1020 \fBcontinue\fR - Attempt to recover from a "hung" I/O by re-dispatching it
1021 to the I/O pipeline if possible.
1023 \fBpanic\fR - Panic the system. This can be used to facilitate an automatic
1024 fail-over to a properly configured fail-over partner.
1026 Default value: \fBwait\fR.
1032 \fBzfs_deadman_checktime_ms\fR (int)
1035 Check time in milliseconds. This defines the frequency at which we check
1036 for hung I/O and potentially invoke the \fBzfs_deadman_failmode\fR behavior.
1038 Default value: \fB60,000\fR.
1044 \fBzfs_deadman_synctime_ms\fR (ulong)
1047 Interval in milliseconds after which the deadman is triggered and also
1048 the interval after which a pool sync operation is considered to be "hung".
1049 Once this limit is exceeded the deadman will be invoked every
1050 \fBzfs_deadman_checktime_ms\fR milliseconds until the pool sync completes.
1052 Default value: \fB600,000\fR.
1058 \fBzfs_deadman_ziotime_ms\fR (ulong)
1061 Interval in milliseconds after which the deadman is triggered and an
1062 individual IO operation is considered to be "hung". As long as the I/O
1063 remains "hung" the deadman will be invoked every \fBzfs_deadman_checktime_ms\fR
1064 milliseconds until the I/O completes.
1066 Default value: \fB300,000\fR.
1072 \fBzfs_dedup_prefetch\fR (int)
1075 Enable prefetching dedup-ed blks
1077 Use \fB1\fR for yes and \fB0\fR to disable (default).
1083 \fBzfs_delay_min_dirty_percent\fR (int)
1086 Start to delay each transaction once there is this amount of dirty data,
1087 expressed as a percentage of \fBzfs_dirty_data_max\fR.
1088 This value should be >= zfs_vdev_async_write_active_max_dirty_percent.
1089 See the section "ZFS TRANSACTION DELAY".
1091 Default value: \fB60\fR%.
1097 \fBzfs_delay_scale\fR (int)
1100 This controls how quickly the transaction delay approaches infinity.
1101 Larger values cause longer delays for a given amount of dirty data.
1103 For the smoothest delay, this value should be about 1 billion divided
1104 by the maximum number of operations per second. This will smoothly
1105 handle between 10x and 1/10th this number.
1107 See the section "ZFS TRANSACTION DELAY".
1109 Note: \fBzfs_delay_scale\fR * \fBzfs_dirty_data_max\fR must be < 2^64.
1111 Default value: \fB500,000\fR.
1117 \fBzfs_delays_per_second\fR (int)
1120 Rate limit IO delay events to this many per second.
1128 \fBzfs_delete_blocks\fR (ulong)
1131 This is the used to define a large file for the purposes of delete. Files
1132 containing more than \fBzfs_delete_blocks\fR will be deleted asynchronously
1133 while smaller files are deleted synchronously. Decreasing this value will
1134 reduce the time spent in an unlink(2) system call at the expense of a longer
1135 delay before the freed space is available.
1137 Default value: \fB20,480\fR.
1143 \fBzfs_dirty_data_max\fR (int)
1146 Determines the dirty space limit in bytes. Once this limit is exceeded, new
1147 writes are halted until space frees up. This parameter takes precedence
1148 over \fBzfs_dirty_data_max_percent\fR.
1149 See the section "ZFS TRANSACTION DELAY".
1151 Default value: \fB10\fR% of physical RAM, capped at \fBzfs_dirty_data_max_max\fR.
1157 \fBzfs_dirty_data_max_max\fR (int)
1160 Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed in bytes.
1161 This limit is only enforced at module load time, and will be ignored if
1162 \fBzfs_dirty_data_max\fR is later changed. This parameter takes
1163 precedence over \fBzfs_dirty_data_max_max_percent\fR. See the section
1164 "ZFS TRANSACTION DELAY".
1166 Default value: \fB25\fR% of physical RAM.
1172 \fBzfs_dirty_data_max_max_percent\fR (int)
1175 Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed as a
1176 percentage of physical RAM. This limit is only enforced at module load
1177 time, and will be ignored if \fBzfs_dirty_data_max\fR is later changed.
1178 The parameter \fBzfs_dirty_data_max_max\fR takes precedence over this
1179 one. See the section "ZFS TRANSACTION DELAY".
1181 Default value: \fB25\fR%.
1187 \fBzfs_dirty_data_max_percent\fR (int)
1190 Determines the dirty space limit, expressed as a percentage of all
1191 memory. Once this limit is exceeded, new writes are halted until space frees
1192 up. The parameter \fBzfs_dirty_data_max\fR takes precedence over this
1193 one. See the section "ZFS TRANSACTION DELAY".
1195 Default value: \fB10\fR%, subject to \fBzfs_dirty_data_max_max\fR.
1201 \fBzfs_dirty_data_sync\fR (int)
1204 Start syncing out a transaction group if there is at least this much dirty data.
1206 Default value: \fB67,108,864\fR.
1212 \fBzfs_fletcher_4_impl\fR (string)
1215 Select a fletcher 4 implementation.
1217 Supported selectors are: \fBfastest\fR, \fBscalar\fR, \fBsse2\fR, \fBssse3\fR,
1218 \fBavx2\fR, \fBavx512f\fR, and \fBaarch64_neon\fR.
1219 All of the selectors except \fBfastest\fR and \fBscalar\fR require instruction
1220 set extensions to be available and will only appear if ZFS detects that they are
1221 present at runtime. If multiple implementations of fletcher 4 are available,
1222 the \fBfastest\fR will be chosen using a micro benchmark. Selecting \fBscalar\fR
1223 results in the original, CPU based calculation, being used. Selecting any option
1224 other than \fBfastest\fR and \fBscalar\fR results in vector instructions from
1225 the respective CPU instruction set being used.
1227 Default value: \fBfastest\fR.
1233 \fBzfs_free_bpobj_enabled\fR (int)
1236 Enable/disable the processing of the free_bpobj object.
1238 Default value: \fB1\fR.
1244 \fBzfs_async_block_max_blocks\fR (ulong)
1247 Maximum number of blocks freed in a single txg.
1249 Default value: \fB100,000\fR.
1255 \fBzfs_override_estimate_recordsize\fR (ulong)
1258 Record size calculation override for zfs send estimates.
1260 Default value: \fB0\fR.
1266 \fBzfs_vdev_async_read_max_active\fR (int)
1269 Maximum asynchronous read I/Os active to each device.
1270 See the section "ZFS I/O SCHEDULER".
1272 Default value: \fB3\fR.
1278 \fBzfs_vdev_async_read_min_active\fR (int)
1281 Minimum asynchronous read I/Os active to each device.
1282 See the section "ZFS I/O SCHEDULER".
1284 Default value: \fB1\fR.
1290 \fBzfs_vdev_async_write_active_max_dirty_percent\fR (int)
1293 When the pool has more than
1294 \fBzfs_vdev_async_write_active_max_dirty_percent\fR dirty data, use
1295 \fBzfs_vdev_async_write_max_active\fR to limit active async writes. If
1296 the dirty data is between min and max, the active I/O limit is linearly
1297 interpolated. See the section "ZFS I/O SCHEDULER".
1299 Default value: \fB60\fR%.
1305 \fBzfs_vdev_async_write_active_min_dirty_percent\fR (int)
1308 When the pool has less than
1309 \fBzfs_vdev_async_write_active_min_dirty_percent\fR dirty data, use
1310 \fBzfs_vdev_async_write_min_active\fR to limit active async writes. If
1311 the dirty data is between min and max, the active I/O limit is linearly
1312 interpolated. See the section "ZFS I/O SCHEDULER".
1314 Default value: \fB30\fR%.
1320 \fBzfs_vdev_async_write_max_active\fR (int)
1323 Maximum asynchronous write I/Os active to each device.
1324 See the section "ZFS I/O SCHEDULER".
1326 Default value: \fB10\fR.
1332 \fBzfs_vdev_async_write_min_active\fR (int)
1335 Minimum asynchronous write I/Os active to each device.
1336 See the section "ZFS I/O SCHEDULER".
1338 Lower values are associated with better latency on rotational media but poorer
1339 resilver performance. The default value of 2 was chosen as a compromise. A
1340 value of 3 has been shown to improve resilver performance further at a cost of
1341 further increasing latency.
1343 Default value: \fB2\fR.
1349 \fBzfs_vdev_max_active\fR (int)
1352 The maximum number of I/Os active to each device. Ideally, this will be >=
1353 the sum of each queue's max_active. It must be at least the sum of each
1354 queue's min_active. See the section "ZFS I/O SCHEDULER".
1356 Default value: \fB1,000\fR.
1362 \fBzfs_vdev_scrub_max_active\fR (int)
1365 Maximum scrub I/Os active to each device.
1366 See the section "ZFS I/O SCHEDULER".
1368 Default value: \fB2\fR.
1374 \fBzfs_vdev_scrub_min_active\fR (int)
1377 Minimum scrub I/Os active to each device.
1378 See the section "ZFS I/O SCHEDULER".
1380 Default value: \fB1\fR.
1386 \fBzfs_vdev_sync_read_max_active\fR (int)
1389 Maximum synchronous read I/Os active to each device.
1390 See the section "ZFS I/O SCHEDULER".
1392 Default value: \fB10\fR.
1398 \fBzfs_vdev_sync_read_min_active\fR (int)
1401 Minimum synchronous read I/Os active to each device.
1402 See the section "ZFS I/O SCHEDULER".
1404 Default value: \fB10\fR.
1410 \fBzfs_vdev_sync_write_max_active\fR (int)
1413 Maximum synchronous write I/Os active to each device.
1414 See the section "ZFS I/O SCHEDULER".
1416 Default value: \fB10\fR.
1422 \fBzfs_vdev_sync_write_min_active\fR (int)
1425 Minimum synchronous write I/Os active to each device.
1426 See the section "ZFS I/O SCHEDULER".
1428 Default value: \fB10\fR.
1434 \fBzfs_vdev_queue_depth_pct\fR (int)
1437 Maximum number of queued allocations per top-level vdev expressed as
1438 a percentage of \fBzfs_vdev_async_write_max_active\fR which allows the
1439 system to detect devices that are more capable of handling allocations
1440 and to allocate more blocks to those devices. It allows for dynamic
1441 allocation distribution when devices are imbalanced as fuller devices
1442 will tend to be slower than empty devices.
1444 See also \fBzio_dva_throttle_enabled\fR.
1446 Default value: \fB1000\fR%.
1452 \fBzfs_expire_snapshot\fR (int)
1455 Seconds to expire .zfs/snapshot
1457 Default value: \fB300\fR.
1463 \fBzfs_admin_snapshot\fR (int)
1466 Allow the creation, removal, or renaming of entries in the .zfs/snapshot
1467 directory to cause the creation, destruction, or renaming of snapshots.
1468 When enabled this functionality works both locally and over NFS exports
1469 which have the 'no_root_squash' option set. This functionality is disabled
1472 Use \fB1\fR for yes and \fB0\fR for no (default).
1478 \fBzfs_flags\fR (int)
1481 Set additional debugging flags. The following flags may be bitwise-or'd
1493 Enable dprintf entries in the debug log.
1495 2 ZFS_DEBUG_DBUF_VERIFY *
1496 Enable extra dbuf verifications.
1498 4 ZFS_DEBUG_DNODE_VERIFY *
1499 Enable extra dnode verifications.
1501 8 ZFS_DEBUG_SNAPNAMES
1502 Enable snapshot name verification.
1505 Check for illegally modified ARC buffers.
1507 64 ZFS_DEBUG_ZIO_FREE
1508 Enable verification of block frees.
1510 128 ZFS_DEBUG_HISTOGRAM_VERIFY
1511 Enable extra spacemap histogram verifications.
1513 256 ZFS_DEBUG_METASLAB_VERIFY
1514 Verify space accounting on disk matches in-core range_trees.
1516 512 ZFS_DEBUG_SET_ERROR
1517 Enable SET_ERROR and dprintf entries in the debug log.
1520 * Requires debug build.
1522 Default value: \fB0\fR.
1528 \fBzfs_free_leak_on_eio\fR (int)
1531 If destroy encounters an EIO while reading metadata (e.g. indirect
1532 blocks), space referenced by the missing metadata can not be freed.
1533 Normally this causes the background destroy to become "stalled", as
1534 it is unable to make forward progress. While in this stalled state,
1535 all remaining space to free from the error-encountering filesystem is
1536 "temporarily leaked". Set this flag to cause it to ignore the EIO,
1537 permanently leak the space from indirect blocks that can not be read,
1538 and continue to free everything else that it can.
1540 The default, "stalling" behavior is useful if the storage partially
1541 fails (i.e. some but not all i/os fail), and then later recovers. In
1542 this case, we will be able to continue pool operations while it is
1543 partially failed, and when it recovers, we can continue to free the
1544 space, with no leaks. However, note that this case is actually
1547 Typically pools either (a) fail completely (but perhaps temporarily,
1548 e.g. a top-level vdev going offline), or (b) have localized,
1549 permanent errors (e.g. disk returns the wrong data due to bit flip or
1550 firmware bug). In case (a), this setting does not matter because the
1551 pool will be suspended and the sync thread will not be able to make
1552 forward progress regardless. In case (b), because the error is
1553 permanent, the best we can do is leak the minimum amount of space,
1554 which is what setting this flag will do. Therefore, it is reasonable
1555 for this flag to normally be set, but we chose the more conservative
1556 approach of not setting it, so that there is no possibility of
1557 leaking space in the "partial temporary" failure case.
1559 Default value: \fB0\fR.
1565 \fBzfs_free_min_time_ms\fR (int)
1568 During a \fBzfs destroy\fR operation using \fBfeature@async_destroy\fR a minimum
1569 of this much time will be spent working on freeing blocks per txg.
1571 Default value: \fB1,000\fR.
1577 \fBzfs_immediate_write_sz\fR (long)
1580 Largest data block to write to zil. Larger blocks will be treated as if the
1581 dataset being written to had the property setting \fBlogbias=throughput\fR.
1583 Default value: \fB32,768\fR.
1589 \fBzfs_lua_max_instrlimit\fR (ulong)
1592 The maximum execution time limit that can be set for a ZFS channel program,
1593 specified as a number of Lua instructions.
1595 Default value: \fB100,000,000\fR.
1601 \fBzfs_lua_max_memlimit\fR (ulong)
1604 The maximum memory limit that can be set for a ZFS channel program, specified
1607 Default value: \fB104,857,600\fR.
1613 \fBzfs_max_dataset_nesting\fR (int)
1616 The maximum depth of nested datasets. This value can be tuned temporarily to
1617 fix existing datasets that exceed the predefined limit.
1619 Default value: \fB50\fR.
1625 \fBzfs_max_recordsize\fR (int)
1628 We currently support block sizes from 512 bytes to 16MB. The benefits of
1629 larger blocks, and thus larger IO, need to be weighed against the cost of
1630 COWing a giant block to modify one byte. Additionally, very large blocks
1631 can have an impact on i/o latency, and also potentially on the memory
1632 allocator. Therefore, we do not allow the recordsize to be set larger than
1633 zfs_max_recordsize (default 1MB). Larger blocks can be created by changing
1634 this tunable, and pools with larger blocks can always be imported and used,
1635 regardless of this setting.
1637 Default value: \fB1,048,576\fR.
1643 \fBzfs_metaslab_fragmentation_threshold\fR (int)
1646 Allow metaslabs to keep their active state as long as their fragmentation
1647 percentage is less than or equal to this value. An active metaslab that
1648 exceeds this threshold will no longer keep its active status allowing
1649 better metaslabs to be selected.
1651 Default value: \fB70\fR.
1657 \fBzfs_mg_fragmentation_threshold\fR (int)
1660 Metaslab groups are considered eligible for allocations if their
1661 fragmentation metric (measured as a percentage) is less than or equal to
1662 this value. If a metaslab group exceeds this threshold then it will be
1663 skipped unless all metaslab groups within the metaslab class have also
1664 crossed this threshold.
1666 Default value: \fB85\fR.
1672 \fBzfs_mg_noalloc_threshold\fR (int)
1675 Defines a threshold at which metaslab groups should be eligible for
1676 allocations. The value is expressed as a percentage of free space
1677 beyond which a metaslab group is always eligible for allocations.
1678 If a metaslab group's free space is less than or equal to the
1679 threshold, the allocator will avoid allocating to that group
1680 unless all groups in the pool have reached the threshold. Once all
1681 groups have reached the threshold, all groups are allowed to accept
1682 allocations. The default value of 0 disables the feature and causes
1683 all metaslab groups to be eligible for allocations.
1685 This parameter allows one to deal with pools having heavily imbalanced
1686 vdevs such as would be the case when a new vdev has been added.
1687 Setting the threshold to a non-zero percentage will stop allocations
1688 from being made to vdevs that aren't filled to the specified percentage
1689 and allow lesser filled vdevs to acquire more allocations than they
1690 otherwise would under the old \fBzfs_mg_alloc_failures\fR facility.
1692 Default value: \fB0\fR.
1698 \fBzfs_multihost_history\fR (int)
1701 Historical statistics for the last N multihost updates will be available in
1702 \fB/proc/spl/kstat/zfs/<pool>/multihost\fR
1704 Default value: \fB0\fR.
1710 \fBzfs_multihost_interval\fR (ulong)
1713 Used to control the frequency of multihost writes which are performed when the
1714 \fBmultihost\fR pool property is on. This is one factor used to determine
1715 the length of the activity check during import.
1717 The multihost write period is \fBzfs_multihost_interval / leaf-vdevs\fR milliseconds.
1718 This means that on average a multihost write will be issued for each leaf vdev every
1719 \fBzfs_multihost_interval\fR milliseconds. In practice, the observed period can
1720 vary with the I/O load and this observed value is the delay which is stored in
1723 On import the activity check waits a minimum amount of time determined by
1724 \fBzfs_multihost_interval * zfs_multihost_import_intervals\fR. The activity
1725 check time may be further extended if the value of mmp delay found in the best
1726 uberblock indicates actual multihost updates happened at longer intervals than
1727 \fBzfs_multihost_interval\fR. A minimum value of \fB100ms\fR is enforced.
1729 Default value: \fB1000\fR.
1735 \fBzfs_multihost_import_intervals\fR (uint)
1738 Used to control the duration of the activity test on import. Smaller values of
1739 \fBzfs_multihost_import_intervals\fR will reduce the import time but increase
1740 the risk of failing to detect an active pool. The total activity check time is
1741 never allowed to drop below one second. A value of 0 is ignored and treated as
1744 Default value: \fB10\fR.
1750 \fBzfs_multihost_fail_intervals\fR (uint)
1753 Controls the behavior of the pool when multihost write failures are detected.
1755 When \fBzfs_multihost_fail_intervals = 0\fR then multihost write failures are ignored.
1756 The failures will still be reported to the ZED which depending on its
1757 configuration may take action such as suspending the pool or offlining a device.
1759 When \fBzfs_multihost_fail_intervals > 0\fR then sequential multihost write failures
1760 will cause the pool to be suspended. This occurs when
1761 \fBzfs_multihost_fail_intervals * zfs_multihost_interval\fR milliseconds have
1762 passed since the last successful multihost write. This guarantees the activity test
1763 will see multihost writes if the pool is imported.
1765 Default value: \fB5\fR.
1771 \fBzfs_no_scrub_io\fR (int)
1774 Set for no scrub I/O. This results in scrubs not actually scrubbing data and
1775 simply doing a metadata crawl of the pool instead.
1777 Use \fB1\fR for yes and \fB0\fR for no (default).
1783 \fBzfs_no_scrub_prefetch\fR (int)
1786 Set to disable block prefetching for scrubs.
1788 Use \fB1\fR for yes and \fB0\fR for no (default).
1794 \fBzfs_nocacheflush\fR (int)
1797 Disable cache flush operations on disks when writing. Beware, this may cause
1798 corruption if disks re-order writes.
1800 Use \fB1\fR for yes and \fB0\fR for no (default).
1806 \fBzfs_nopwrite_enabled\fR (int)
1811 Use \fB1\fR for yes (default) and \fB0\fR to disable.
1817 \fBzfs_dmu_offset_next_sync\fR (int)
1820 Enable forcing txg sync to find holes. When enabled forces ZFS to act
1821 like prior versions when SEEK_HOLE or SEEK_DATA flags are used, which
1822 when a dnode is dirty causes txg's to be synced so that this data can be
1825 Use \fB1\fR for yes and \fB0\fR to disable (default).
1831 \fBzfs_pd_bytes_max\fR (int)
1834 The number of bytes which should be prefetched during a pool traversal
1835 (eg: \fBzfs send\fR or other data crawling operations)
1837 Default value: \fB52,428,800\fR.
1843 \fBzfs_per_txg_dirty_frees_percent \fR (ulong)
1846 Tunable to control percentage of dirtied blocks from frees in one TXG.
1847 After this threshold is crossed, additional dirty blocks from frees
1848 wait until the next TXG.
1849 A value of zero will disable this throttle.
1851 Default value: \fB30\fR and \fB0\fR to disable.
1859 \fBzfs_prefetch_disable\fR (int)
1862 This tunable disables predictive prefetch. Note that it leaves "prescient"
1863 prefetch (e.g. prefetch for zfs send) intact. Unlike predictive prefetch,
1864 prescient prefetch never issues i/os that end up not being needed, so it
1865 can't hurt performance.
1867 Use \fB1\fR for yes and \fB0\fR for no (default).
1873 \fBzfs_read_chunk_size\fR (long)
1876 Bytes to read per chunk
1878 Default value: \fB1,048,576\fR.
1884 \fBzfs_read_history\fR (int)
1887 Historical statistics for the last N reads will be available in
1888 \fB/proc/spl/kstat/zfs/<pool>/reads\fR
1890 Default value: \fB0\fR (no data is kept).
1896 \fBzfs_read_history_hits\fR (int)
1899 Include cache hits in read history
1901 Use \fB1\fR for yes and \fB0\fR for no (default).
1907 \fBzfs_reconstruct_indirect_combinations_max\fR (int)
1910 If an indirect split block contains more than this many possible unique
1911 combinations when being reconstructed, consider it too computationally
1912 expensive to check them all. Instead, try at most
1913 \fBzfs_reconstruct_indirect_combinations_max\fR randomly-selected
1914 combinations each time the block is accessed. This allows all segment
1915 copies to participate fairly in the reconstruction when all combinations
1916 cannot be checked and prevents repeated use of one bad copy.
1918 Default value: \fB100\fR.
1924 \fBzfs_recover\fR (int)
1927 Set to attempt to recover from fatal errors. This should only be used as a
1928 last resort, as it typically results in leaked space, or worse.
1930 Use \fB1\fR for yes and \fB0\fR for no (default).
1936 \fBzfs_resilver_min_time_ms\fR (int)
1939 Resilvers are processed by the sync thread. While resilvering it will spend
1940 at least this much time working on a resilver between txg flushes.
1942 Default value: \fB3,000\fR.
1948 \fBzfs_scan_ignore_errors\fR (int)
1951 If set to a nonzero value, remove the DTL (dirty time list) upon
1952 completion of a pool scan (scrub) even if there were unrepairable
1953 errors. It is intended to be used during pool repair or recovery to
1954 stop resilvering when the pool is next imported.
1956 Default value: \fB0\fR.
1962 \fBzfs_scrub_min_time_ms\fR (int)
1965 Scrubs are processed by the sync thread. While scrubbing it will spend
1966 at least this much time working on a scrub between txg flushes.
1968 Default value: \fB1,000\fR.
1974 \fBzfs_scan_checkpoint_intval\fR (int)
1977 To preserve progress across reboots the sequential scan algorithm periodically
1978 needs to stop metadata scanning and issue all the verifications I/Os to disk.
1979 The frequency of this flushing is determined by the
1980 \fBzfs_scan_checkpoint_intval\fR tunable.
1982 Default value: \fB7200\fR seconds (every 2 hours).
1988 \fBzfs_scan_fill_weight\fR (int)
1991 This tunable affects how scrub and resilver I/O segments are ordered. A higher
1992 number indicates that we care more about how filled in a segment is, while a
1993 lower number indicates we care more about the size of the extent without
1994 considering the gaps within a segment. This value is only tunable upon module
1995 insertion. Changing the value afterwards will have no affect on scrub or
1996 resilver performance.
1998 Default value: \fB3\fR.
2004 \fBzfs_scan_issue_strategy\fR (int)
2007 Determines the order that data will be verified while scrubbing or resilvering.
2008 If set to \fB1\fR, data will be verified as sequentially as possible, given the
2009 amount of memory reserved for scrubbing (see \fBzfs_scan_mem_lim_fact\fR). This
2010 may improve scrub performance if the pool's data is very fragmented. If set to
2011 \fB2\fR, the largest mostly-contiguous chunk of found data will be verified
2012 first. By deferring scrubbing of small segments, we may later find adjacent data
2013 to coalesce and increase the segment size. If set to \fB0\fR, zfs will use
2014 strategy \fB1\fR during normal verification and strategy \fB2\fR while taking a
2017 Default value: \fB0\fR.
2023 \fBzfs_scan_legacy\fR (int)
2026 A value of 0 indicates that scrubs and resilvers will gather metadata in
2027 memory before issuing sequential I/O. A value of 1 indicates that the legacy
2028 algorithm will be used where I/O is initiated as soon as it is discovered.
2029 Changing this value to 0 will not affect scrubs or resilvers that are already
2032 Default value: \fB0\fR.
2038 \fBzfs_scan_max_ext_gap\fR (int)
2041 Indicates the largest gap in bytes between scrub / resilver I/Os that will still
2042 be considered sequential for sorting purposes. Changing this value will not
2043 affect scrubs or resilvers that are already in progress.
2045 Default value: \fB2097152 (2 MB)\fR.
2051 \fBzfs_scan_mem_lim_fact\fR (int)
2054 Maximum fraction of RAM used for I/O sorting by sequential scan algorithm.
2055 This tunable determines the hard limit for I/O sorting memory usage.
2056 When the hard limit is reached we stop scanning metadata and start issuing
2057 data verification I/O. This is done until we get below the soft limit.
2059 Default value: \fB20\fR which is 5% of RAM (1/20).
2065 \fBzfs_scan_mem_lim_soft_fact\fR (int)
2068 The fraction of the hard limit used to determined the soft limit for I/O sorting
2069 by the sequential scan algorithm. When we cross this limit from bellow no action
2070 is taken. When we cross this limit from above it is because we are issuing
2071 verification I/O. In this case (unless the metadata scan is done) we stop
2072 issuing verification I/O and start scanning metadata again until we get to the
2075 Default value: \fB20\fR which is 5% of the hard limit (1/20).
2081 \fBzfs_scan_vdev_limit\fR (int)
2084 Maximum amount of data that can be concurrently issued at once for scrubs and
2085 resilvers per leaf device, given in bytes.
2087 Default value: \fB41943040\fR.
2093 \fBzfs_send_corrupt_data\fR (int)
2096 Allow sending of corrupt data (ignore read/checksum errors when sending data)
2098 Use \fB1\fR for yes and \fB0\fR for no (default).
2104 \fBzfs_send_queue_length\fR (int)
2107 The maximum number of bytes allowed in the \fBzfs send\fR queue. This value
2108 must be at least twice the maximum block size in use.
2110 Default value: \fB16,777,216\fR.
2116 \fBzfs_recv_queue_length\fR (int)
2120 The maximum number of bytes allowed in the \fBzfs receive\fR queue. This value
2121 must be at least twice the maximum block size in use.
2123 Default value: \fB16,777,216\fR.
2129 \fBzfs_sync_pass_deferred_free\fR (int)
2132 Flushing of data to disk is done in passes. Defer frees starting in this pass
2134 Default value: \fB2\fR.
2140 \fBzfs_spa_discard_memory_limit\fR (int)
2143 Maximum memory used for prefetching a checkpoint's space map on each
2144 vdev while discarding the checkpoint.
2146 Default value: \fB16,777,216\fR.
2152 \fBzfs_sync_pass_dont_compress\fR (int)
2155 Don't compress starting in this pass
2157 Default value: \fB5\fR.
2163 \fBzfs_sync_pass_rewrite\fR (int)
2166 Rewrite new block pointers starting in this pass
2168 Default value: \fB2\fR.
2174 \fBzfs_sync_taskq_batch_pct\fR (int)
2177 This controls the number of threads used by the dp_sync_taskq. The default
2178 value of 75% will create a maximum of one thread per cpu.
2180 Default value: \fB75\fR%.
2186 \fBzfs_txg_history\fR (int)
2189 Historical statistics for the last N txgs will be available in
2190 \fB/proc/spl/kstat/zfs/<pool>/txgs\fR
2192 Default value: \fB0\fR.
2198 \fBzfs_txg_timeout\fR (int)
2201 Flush dirty data to disk at least every N seconds (maximum txg duration)
2203 Default value: \fB5\fR.
2209 \fBzfs_vdev_aggregation_limit\fR (int)
2212 Max vdev I/O aggregation size
2214 Default value: \fB131,072\fR.
2220 \fBzfs_vdev_cache_bshift\fR (int)
2223 Shift size to inflate reads too
2225 Default value: \fB16\fR (effectively 65536).
2231 \fBzfs_vdev_cache_max\fR (int)
2234 Inflate reads smaller than this value to meet the \fBzfs_vdev_cache_bshift\fR
2237 Default value: \fB16384\fR.
2243 \fBzfs_vdev_cache_size\fR (int)
2246 Total size of the per-disk cache in bytes.
2248 Currently this feature is disabled as it has been found to not be helpful
2249 for performance and in some cases harmful.
2251 Default value: \fB0\fR.
2257 \fBzfs_vdev_mirror_rotating_inc\fR (int)
2260 A number by which the balancing algorithm increments the load calculation for
2261 the purpose of selecting the least busy mirror member when an I/O immediately
2262 follows its predecessor on rotational vdevs for the purpose of making decisions
2265 Default value: \fB0\fR.
2271 \fBzfs_vdev_mirror_rotating_seek_inc\fR (int)
2274 A number by which the balancing algorithm increments the load calculation for
2275 the purpose of selecting the least busy mirror member when an I/O lacks
2276 locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within
2277 this that are not immediately following the previous I/O are incremented by
2280 Default value: \fB5\fR.
2286 \fBzfs_vdev_mirror_rotating_seek_offset\fR (int)
2289 The maximum distance for the last queued I/O in which the balancing algorithm
2290 considers an I/O to have locality.
2291 See the section "ZFS I/O SCHEDULER".
2293 Default value: \fB1048576\fR.
2299 \fBzfs_vdev_mirror_non_rotating_inc\fR (int)
2302 A number by which the balancing algorithm increments the load calculation for
2303 the purpose of selecting the least busy mirror member on non-rotational vdevs
2304 when I/Os do not immediately follow one another.
2306 Default value: \fB0\fR.
2312 \fBzfs_vdev_mirror_non_rotating_seek_inc\fR (int)
2315 A number by which the balancing algorithm increments the load calculation for
2316 the purpose of selecting the least busy mirror member when an I/O lacks
2317 locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within
2318 this that are not immediately following the previous I/O are incremented by
2321 Default value: \fB1\fR.
2327 \fBzfs_vdev_read_gap_limit\fR (int)
2330 Aggregate read I/O operations if the gap on-disk between them is within this
2333 Default value: \fB32,768\fR.
2339 \fBzfs_vdev_scheduler\fR (charp)
2342 Set the Linux I/O scheduler on whole disk vdevs to this scheduler. Valid options
2343 are noop, cfq, bfq & deadline
2345 Default value: \fBnoop\fR.
2351 \fBzfs_vdev_write_gap_limit\fR (int)
2354 Aggregate write I/O over gap
2356 Default value: \fB4,096\fR.
2362 \fBzfs_vdev_raidz_impl\fR (string)
2365 Parameter for selecting raidz parity implementation to use.
2367 Options marked (always) below may be selected on module load as they are
2368 supported on all systems.
2369 The remaining options may only be set after the module is loaded, as they
2370 are available only if the implementations are compiled in and supported
2371 on the running system.
2373 Once the module is loaded, the content of
2374 /sys/module/zfs/parameters/zfs_vdev_raidz_impl will show available options
2375 with the currently selected one enclosed in [].
2376 Possible options are:
2377 fastest - (always) implementation selected using built-in benchmark
2378 original - (always) original raidz implementation
2379 scalar - (always) scalar raidz implementation
2380 sse2 - implementation using SSE2 instruction set (64bit x86 only)
2381 ssse3 - implementation using SSSE3 instruction set (64bit x86 only)
2382 avx2 - implementation using AVX2 instruction set (64bit x86 only)
2383 avx512f - implementation using AVX512F instruction set (64bit x86 only)
2384 avx512bw - implementation using AVX512F & AVX512BW instruction sets (64bit x86 only)
2385 aarch64_neon - implementation using NEON (Aarch64/64 bit ARMv8 only)
2386 aarch64_neonx2 - implementation using NEON with more unrolling (Aarch64/64 bit ARMv8 only)
2388 Default value: \fBfastest\fR.
2394 \fBzfs_zevent_cols\fR (int)
2397 When zevents are logged to the console use this as the word wrap width.
2399 Default value: \fB80\fR.
2405 \fBzfs_zevent_console\fR (int)
2408 Log events to the console
2410 Use \fB1\fR for yes and \fB0\fR for no (default).
2416 \fBzfs_zevent_len_max\fR (int)
2419 Max event queue length. A value of 0 will result in a calculated value which
2420 increases with the number of CPUs in the system (minimum 64 events). Events
2421 in the queue can be viewed with the \fBzpool events\fR command.
2423 Default value: \fB0\fR.
2429 \fBzfs_zil_clean_taskq_maxalloc\fR (int)
2432 The maximum number of taskq entries that are allowed to be cached. When this
2433 limit is exceeded transaction records (itxs) will be cleaned synchronously.
2435 Default value: \fB1048576\fR.
2441 \fBzfs_zil_clean_taskq_minalloc\fR (int)
2444 The number of taskq entries that are pre-populated when the taskq is first
2445 created and are immediately available for use.
2447 Default value: \fB1024\fR.
2453 \fBzfs_zil_clean_taskq_nthr_pct\fR (int)
2456 This controls the number of threads used by the dp_zil_clean_taskq. The default
2457 value of 100% will create a maximum of one thread per cpu.
2459 Default value: \fB100\fR%.
2465 \fBzil_replay_disable\fR (int)
2468 Disable intent logging replay. Can be disabled for recovery from corrupted
2471 Use \fB1\fR for yes and \fB0\fR for no (default).
2477 \fBzil_slog_bulk\fR (ulong)
2480 Limit SLOG write size per commit executed with synchronous priority.
2481 Any writes above that will be executed with lower (asynchronous) priority
2482 to limit potential SLOG device abuse by single active ZIL writer.
2484 Default value: \fB786,432\fR.
2490 \fBzio_delay_max\fR (int)
2493 A zevent will be logged if a ZIO operation takes more than N milliseconds to
2494 complete. Note that this is only a logging facility, not a timeout on
2497 Default value: \fB30,000\fR.
2503 \fBzio_dva_throttle_enabled\fR (int)
2506 Throttle block allocations in the ZIO pipeline. This allows for
2507 dynamic allocation distribution when devices are imbalanced.
2508 When enabled, the maximum number of pending allocations per top-level vdev
2509 is limited by \fBzfs_vdev_queue_depth_pct\fR.
2511 Default value: \fB1\fR.
2517 \fBzio_requeue_io_start_cut_in_line\fR (int)
2520 Prioritize requeued I/O
2522 Default value: \fB0\fR.
2528 \fBzio_taskq_batch_pct\fR (uint)
2531 Percentage of online CPUs (or CPU cores, etc) which will run a worker thread
2532 for IO. These workers are responsible for IO work such as compression and
2533 checksum calculations. Fractional number of CPUs will be rounded down.
2535 The default value of 75 was chosen to avoid using all CPUs which can result in
2536 latency issues and inconsistent application performance, especially when high
2537 compression is enabled.
2539 Default value: \fB75\fR.
2545 \fBzvol_inhibit_dev\fR (uint)
2548 Do not create zvol device nodes. This may slightly improve startup time on
2549 systems with a very large number of zvols.
2551 Use \fB1\fR for yes and \fB0\fR for no (default).
2557 \fBzvol_major\fR (uint)
2560 Major number for zvol block devices
2562 Default value: \fB230\fR.
2568 \fBzvol_max_discard_blocks\fR (ulong)
2571 Discard (aka TRIM) operations done on zvols will be done in batches of this
2572 many blocks, where block size is determined by the \fBvolblocksize\fR property
2575 Default value: \fB16,384\fR.
2581 \fBzvol_prefetch_bytes\fR (uint)
2584 When adding a zvol to the system prefetch \fBzvol_prefetch_bytes\fR
2585 from the start and end of the volume. Prefetching these regions
2586 of the volume is desirable because they are likely to be accessed
2587 immediately by \fBblkid(8)\fR or by the kernel scanning for a partition
2590 Default value: \fB131,072\fR.
2596 \fBzvol_request_sync\fR (uint)
2599 When processing I/O requests for a zvol submit them synchronously. This
2600 effectively limits the queue depth to 1 for each I/O submitter. When set
2601 to 0 requests are handled asynchronously by a thread pool. The number of
2602 requests which can be handled concurrently is controller by \fBzvol_threads\fR.
2604 Default value: \fB0\fR.
2610 \fBzvol_threads\fR (uint)
2613 Max number of threads which can handle zvol I/O requests concurrently.
2615 Default value: \fB32\fR.
2621 \fBzvol_volmode\fR (uint)
2624 Defines zvol block devices behaviour when \fBvolmode\fR is set to \fBdefault\fR.
2625 Valid values are \fB1\fR (full), \fB2\fR (dev) and \fB3\fR (none).
2627 Default value: \fB1\fR.
2633 \fBzfs_qat_disable\fR (int)
2636 This tunable disables qat hardware acceleration for gzip compression and.
2637 AES-GCM encryption. It is available only if qat acceleration is compiled in
2638 and the qat driver is present.
2640 Use \fB1\fR for yes and \fB0\fR for no (default).
2643 .SH ZFS I/O SCHEDULER
2644 ZFS issues I/O operations to leaf vdevs to satisfy and complete I/Os.
2645 The I/O scheduler determines when and in what order those operations are
2646 issued. The I/O scheduler divides operations into five I/O classes
2647 prioritized in the following order: sync read, sync write, async read,
2648 async write, and scrub/resilver. Each queue defines the minimum and
2649 maximum number of concurrent operations that may be issued to the
2650 device. In addition, the device has an aggregate maximum,
2651 \fBzfs_vdev_max_active\fR. Note that the sum of the per-queue minimums
2652 must not exceed the aggregate maximum. If the sum of the per-queue
2653 maximums exceeds the aggregate maximum, then the number of active I/Os
2654 may reach \fBzfs_vdev_max_active\fR, in which case no further I/Os will
2655 be issued regardless of whether all per-queue minimums have been met.
2657 For many physical devices, throughput increases with the number of
2658 concurrent operations, but latency typically suffers. Further, physical
2659 devices typically have a limit at which more concurrent operations have no
2660 effect on throughput or can actually cause it to decrease.
2662 The scheduler selects the next operation to issue by first looking for an
2663 I/O class whose minimum has not been satisfied. Once all are satisfied and
2664 the aggregate maximum has not been hit, the scheduler looks for classes
2665 whose maximum has not been satisfied. Iteration through the I/O classes is
2666 done in the order specified above. No further operations are issued if the
2667 aggregate maximum number of concurrent operations has been hit or if there
2668 are no operations queued for an I/O class that has not hit its maximum.
2669 Every time an I/O is queued or an operation completes, the I/O scheduler
2670 looks for new operations to issue.
2672 In general, smaller max_active's will lead to lower latency of synchronous
2673 operations. Larger max_active's may lead to higher overall throughput,
2674 depending on underlying storage.
2676 The ratio of the queues' max_actives determines the balance of performance
2677 between reads, writes, and scrubs. E.g., increasing
2678 \fBzfs_vdev_scrub_max_active\fR will cause the scrub or resilver to complete
2679 more quickly, but reads and writes to have higher latency and lower throughput.
2681 All I/O classes have a fixed maximum number of outstanding operations
2682 except for the async write class. Asynchronous writes represent the data
2683 that is committed to stable storage during the syncing stage for
2684 transaction groups. Transaction groups enter the syncing state
2685 periodically so the number of queued async writes will quickly burst up
2686 and then bleed down to zero. Rather than servicing them as quickly as
2687 possible, the I/O scheduler changes the maximum number of active async
2688 write I/Os according to the amount of dirty data in the pool. Since
2689 both throughput and latency typically increase with the number of
2690 concurrent operations issued to physical devices, reducing the
2691 burstiness in the number of concurrent operations also stabilizes the
2692 response time of operations from other -- and in particular synchronous
2693 -- queues. In broad strokes, the I/O scheduler will issue more
2694 concurrent operations from the async write queue as there's more dirty
2699 The number of concurrent operations issued for the async write I/O class
2700 follows a piece-wise linear function defined by a few adjustable points.
2703 | o---------| <-- zfs_vdev_async_write_max_active
2710 |-------o | | <-- zfs_vdev_async_write_min_active
2711 0|_______^______|_________|
2712 0% | | 100% of zfs_dirty_data_max
2714 | `-- zfs_vdev_async_write_active_max_dirty_percent
2715 `--------- zfs_vdev_async_write_active_min_dirty_percent
2718 Until the amount of dirty data exceeds a minimum percentage of the dirty
2719 data allowed in the pool, the I/O scheduler will limit the number of
2720 concurrent operations to the minimum. As that threshold is crossed, the
2721 number of concurrent operations issued increases linearly to the maximum at
2722 the specified maximum percentage of the dirty data allowed in the pool.
2724 Ideally, the amount of dirty data on a busy pool will stay in the sloped
2725 part of the function between \fBzfs_vdev_async_write_active_min_dirty_percent\fR
2726 and \fBzfs_vdev_async_write_active_max_dirty_percent\fR. If it exceeds the
2727 maximum percentage, this indicates that the rate of incoming data is
2728 greater than the rate that the backend storage can handle. In this case, we
2729 must further throttle incoming writes, as described in the next section.
2731 .SH ZFS TRANSACTION DELAY
2732 We delay transactions when we've determined that the backend storage
2733 isn't able to accommodate the rate of incoming writes.
2735 If there is already a transaction waiting, we delay relative to when
2736 that transaction will finish waiting. This way the calculated delay time
2737 is independent of the number of threads concurrently executing
2740 If we are the only waiter, wait relative to when the transaction
2741 started, rather than the current time. This credits the transaction for
2742 "time already served", e.g. reading indirect blocks.
2744 The minimum time for a transaction to take is calculated as:
2746 min_time = zfs_delay_scale * (dirty - min) / (max - dirty)
2747 min_time is then capped at 100 milliseconds.
2750 The delay has two degrees of freedom that can be adjusted via tunables. The
2751 percentage of dirty data at which we start to delay is defined by
2752 \fBzfs_delay_min_dirty_percent\fR. This should typically be at or above
2753 \fBzfs_vdev_async_write_active_max_dirty_percent\fR so that we only start to
2754 delay after writing at full speed has failed to keep up with the incoming write
2755 rate. The scale of the curve is defined by \fBzfs_delay_scale\fR. Roughly speaking,
2756 this variable determines the amount of delay at the midpoint of the curve.
2760 10ms +-------------------------------------------------------------*+
2776 2ms + (midpoint) * +
2779 | zfs_delay_scale ----------> ******** |
2780 0 +-------------------------------------*********----------------+
2781 0% <- zfs_dirty_data_max -> 100%
2784 Note that since the delay is added to the outstanding time remaining on the
2785 most recent transaction, the delay is effectively the inverse of IOPS.
2786 Here the midpoint of 500us translates to 2000 IOPS. The shape of the curve
2787 was chosen such that small changes in the amount of accumulated dirty data
2788 in the first 3/4 of the curve yield relatively small differences in the
2791 The effects can be easier to understand when the amount of delay is
2792 represented on a log scale:
2796 100ms +-------------------------------------------------------------++
2805 + zfs_delay_scale ----------> ***** +
2816 +--------------------------------------------------------------+
2817 0% <- zfs_dirty_data_max -> 100%
2820 Note here that only as the amount of dirty data approaches its limit does
2821 the delay start to increase rapidly. The goal of a properly tuned system
2822 should be to keep the amount of dirty data out of that range by first
2823 ensuring that the appropriate limits are set for the I/O scheduler to reach
2824 optimal throughput on the backend storage, and then by changing the value
2825 of \fBzfs_delay_scale\fR to increase the steepness of the curve.