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, it will be divided into approximately (but no more than) this number of metaslabs.
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 \fBmetaslab_preload_enabled\fR (int)
321 Enable metaslab group preloading.
323 Use \fB1\fR for yes (default) and \fB0\fR for no.
329 \fBmetaslab_lba_weighting_enabled\fR (int)
332 Give more weight to metaslabs with lower LBAs, assuming they have
333 greater bandwidth as is typically the case on a modern constant
334 angular velocity disk drive.
336 Use \fB1\fR for yes (default) and \fB0\fR for no.
342 \fBspa_config_path\fR (charp)
347 Default value: \fB/etc/zfs/zpool.cache\fR.
353 \fBspa_asize_inflation\fR (int)
356 Multiplication factor used to estimate actual disk consumption from the
357 size of data being written. The default value is a worst case estimate,
358 but lower values may be valid for a given pool depending on its
359 configuration. Pool administrators who understand the factors involved
360 may wish to specify a more realistic inflation factor, particularly if
361 they operate close to quota or capacity limits.
363 Default value: \fB24\fR.
369 \fBspa_load_print_vdev_tree\fR (int)
372 Whether to print the vdev tree in the debugging message buffer during pool import.
373 Use 0 to disable and 1 to enable.
375 Default value: \fB0\fR.
381 \fBspa_load_verify_data\fR (int)
384 Whether to traverse data blocks during an "extreme rewind" (\fB-X\fR)
385 import. Use 0 to disable and 1 to enable.
387 An extreme rewind import normally performs a full traversal of all
388 blocks in the pool for verification. If this parameter is set to 0,
389 the traversal skips non-metadata blocks. It can be toggled once the
390 import has started to stop or start the traversal of non-metadata blocks.
392 Default value: \fB1\fR.
398 \fBspa_load_verify_metadata\fR (int)
401 Whether to traverse blocks during an "extreme rewind" (\fB-X\fR)
402 pool import. Use 0 to disable and 1 to enable.
404 An extreme rewind import normally performs a full traversal of all
405 blocks in the pool for verification. If this parameter is set to 0,
406 the traversal is not performed. It can be toggled once the import has
407 started to stop or start the traversal.
409 Default value: \fB1\fR.
415 \fBspa_load_verify_maxinflight\fR (int)
418 Maximum concurrent I/Os during the traversal performed during an "extreme
419 rewind" (\fB-X\fR) pool import.
421 Default value: \fB10000\fR.
427 \fBspa_slop_shift\fR (int)
430 Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space
431 in the pool to be consumed. This ensures that we don't run the pool
432 completely out of space, due to unaccounted changes (e.g. to the MOS).
433 It also limits the worst-case time to allocate space. If we have
434 less than this amount of free space, most ZPL operations (e.g. write,
435 create) will return ENOSPC.
437 Default value: \fB5\fR.
443 \fBvdev_removal_max_span\fR (int)
446 During top-level vdev removal, chunks of data are copied from the vdev
447 which may include free space in order to trade bandwidth for IOPS.
448 This parameter determines the maximum span of free space (in bytes)
449 which will be included as "unnecessary" data in a chunk of copied data.
451 The default value here was chosen to align with
452 \fBzfs_vdev_read_gap_limit\fR, which is a similar concept when doing
453 regular reads (but there's no reason it has to be the same).
455 Default value: \fB32,768\fR.
461 \fBzfetch_array_rd_sz\fR (ulong)
464 If prefetching is enabled, disable prefetching for reads larger than this size.
466 Default value: \fB1,048,576\fR.
472 \fBzfetch_max_distance\fR (uint)
475 Max bytes to prefetch per stream (default 8MB).
477 Default value: \fB8,388,608\fR.
483 \fBzfetch_max_streams\fR (uint)
486 Max number of streams per zfetch (prefetch streams per file).
488 Default value: \fB8\fR.
494 \fBzfetch_min_sec_reap\fR (uint)
497 Min time before an active prefetch stream can be reclaimed
499 Default value: \fB2\fR.
505 \fBzfs_arc_dnode_limit\fR (ulong)
508 When the number of bytes consumed by dnodes in the ARC exceeds this number of
509 bytes, try to unpin some of it in response to demand for non-metadata. This
510 value acts as a ceiling to the amount of dnode metadata, and defaults to 0 which
511 indicates that a percent which is based on \fBzfs_arc_dnode_limit_percent\fR of
512 the ARC meta buffers that may be used for dnodes.
514 See also \fBzfs_arc_meta_prune\fR which serves a similar purpose but is used
515 when the amount of metadata in the ARC exceeds \fBzfs_arc_meta_limit\fR rather
516 than in response to overall demand for non-metadata.
519 Default value: \fB0\fR.
525 \fBzfs_arc_dnode_limit_percent\fR (ulong)
528 Percentage that can be consumed by dnodes of ARC meta buffers.
530 See also \fBzfs_arc_dnode_limit\fR which serves a similar purpose but has a
531 higher priority if set to nonzero value.
533 Default value: \fB10\fR%.
539 \fBzfs_arc_dnode_reduce_percent\fR (ulong)
542 Percentage of ARC dnodes to try to scan in response to demand for non-metadata
543 when the number of bytes consumed by dnodes exceeds \fBzfs_arc_dnode_limit\fR.
546 Default value: \fB10\fR% of the number of dnodes in the ARC.
552 \fBzfs_arc_average_blocksize\fR (int)
555 The ARC's buffer hash table is sized based on the assumption of an average
556 block size of \fBzfs_arc_average_blocksize\fR (default 8K). This works out
557 to roughly 1MB of hash table per 1GB of physical memory with 8-byte pointers.
558 For configurations with a known larger average block size this value can be
559 increased to reduce the memory footprint.
562 Default value: \fB8192\fR.
568 \fBzfs_arc_evict_batch_limit\fR (int)
571 Number ARC headers to evict per sub-list before proceeding to another sub-list.
572 This batch-style operation prevents entire sub-lists from being evicted at once
573 but comes at a cost of additional unlocking and locking.
575 Default value: \fB10\fR.
581 \fBzfs_arc_grow_retry\fR (int)
584 If set to a non zero value, it will replace the arc_grow_retry value with this value.
585 The arc_grow_retry value (default 5) is the number of seconds the ARC will wait before
586 trying to resume growth after a memory pressure event.
588 Default value: \fB0\fR.
594 \fBzfs_arc_lotsfree_percent\fR (int)
597 Throttle I/O when free system memory drops below this percentage of total
598 system memory. Setting this value to 0 will disable the throttle.
600 Default value: \fB10\fR%.
606 \fBzfs_arc_max\fR (ulong)
609 Max arc size of ARC in bytes. If set to 0 then it will consume 1/2 of system
610 RAM. This value must be at least 67108864 (64 megabytes).
612 This value can be changed dynamically with some caveats. It cannot be set back
613 to 0 while running and reducing it below the current ARC size will not cause
614 the ARC to shrink without memory pressure to induce shrinking.
616 Default value: \fB0\fR.
622 \fBzfs_arc_meta_adjust_restarts\fR (ulong)
625 The number of restart passes to make while scanning the ARC attempting
626 the free buffers in order to stay below the \fBzfs_arc_meta_limit\fR.
627 This value should not need to be tuned but is available to facilitate
628 performance analysis.
630 Default value: \fB4096\fR.
636 \fBzfs_arc_meta_limit\fR (ulong)
639 The maximum allowed size in bytes that meta data buffers are allowed to
640 consume in the ARC. When this limit is reached meta data buffers will
641 be reclaimed even if the overall arc_c_max has not been reached. This
642 value defaults to 0 which indicates that a percent which is based on
643 \fBzfs_arc_meta_limit_percent\fR of the ARC may be used for meta data.
645 This value my be changed dynamically except that it cannot be set back to 0
646 for a specific percent of the ARC; it must be set to an explicit value.
648 Default value: \fB0\fR.
654 \fBzfs_arc_meta_limit_percent\fR (ulong)
657 Percentage of ARC buffers that can be used for meta data.
659 See also \fBzfs_arc_meta_limit\fR which serves a similar purpose but has a
660 higher priority if set to nonzero value.
663 Default value: \fB75\fR%.
669 \fBzfs_arc_meta_min\fR (ulong)
672 The minimum allowed size in bytes that meta data buffers may consume in
673 the ARC. This value defaults to 0 which disables a floor on the amount
674 of the ARC devoted meta data.
676 Default value: \fB0\fR.
682 \fBzfs_arc_meta_prune\fR (int)
685 The number of dentries and inodes to be scanned looking for entries
686 which can be dropped. This may be required when the ARC reaches the
687 \fBzfs_arc_meta_limit\fR because dentries and inodes can pin buffers
688 in the ARC. Increasing this value will cause to dentry and inode caches
689 to be pruned more aggressively. Setting this value to 0 will disable
690 pruning the inode and dentry caches.
692 Default value: \fB10,000\fR.
698 \fBzfs_arc_meta_strategy\fR (int)
701 Define the strategy for ARC meta data buffer eviction (meta reclaim strategy).
702 A value of 0 (META_ONLY) will evict only the ARC meta data buffers.
703 A value of 1 (BALANCED) indicates that additional data buffers may be evicted if
704 that is required to in order to evict the required number of meta data buffers.
706 Default value: \fB1\fR.
712 \fBzfs_arc_min\fR (ulong)
715 Min arc size of ARC in bytes. If set to 0 then arc_c_min will default to
716 consuming the larger of 32M or 1/32 of total system memory.
718 Default value: \fB0\fR.
724 \fBzfs_arc_min_prefetch_ms\fR (int)
727 Minimum time prefetched blocks are locked in the ARC, specified in ms.
728 A value of \fB0\fR will default to 1000 ms.
730 Default value: \fB0\fR.
736 \fBzfs_arc_min_prescient_prefetch_ms\fR (int)
739 Minimum time "prescient prefetched" blocks are locked in the ARC, specified
740 in ms. These blocks are meant to be prefetched fairly aggresively ahead of
741 the code that may use them. A value of \fB0\fR will default to 6000 ms.
743 Default value: \fB0\fR.
749 \fBzfs_max_missing_tvds\fR (int)
752 Number of missing top-level vdevs which will be allowed during
753 pool import (only in read-only mode).
755 Default value: \fB0\fR
761 \fBzfs_multilist_num_sublists\fR (int)
764 To allow more fine-grained locking, each ARC state contains a series
765 of lists for both data and meta data objects. Locking is performed at
766 the level of these "sub-lists". This parameters controls the number of
767 sub-lists per ARC state, and also applies to other uses of the
768 multilist data structure.
770 Default value: \fB4\fR or the number of online CPUs, whichever is greater
776 \fBzfs_arc_overflow_shift\fR (int)
779 The ARC size is considered to be overflowing if it exceeds the current
780 ARC target size (arc_c) by a threshold determined by this parameter.
781 The threshold is calculated as a fraction of arc_c using the formula
782 "arc_c >> \fBzfs_arc_overflow_shift\fR".
784 The default value of 8 causes the ARC to be considered to be overflowing
785 if it exceeds the target size by 1/256th (0.3%) of the target size.
787 When the ARC is overflowing, new buffer allocations are stalled until
788 the reclaim thread catches up and the overflow condition no longer exists.
790 Default value: \fB8\fR.
797 \fBzfs_arc_p_min_shift\fR (int)
800 If set to a non zero value, this will update arc_p_min_shift (default 4)
802 arc_p_min_shift is used to shift of arc_c for calculating both min and max
805 Default value: \fB0\fR.
811 \fBzfs_arc_p_dampener_disable\fR (int)
814 Disable arc_p adapt dampener
816 Use \fB1\fR for yes (default) and \fB0\fR to disable.
822 \fBzfs_arc_shrink_shift\fR (int)
825 If set to a non zero value, this will update arc_shrink_shift (default 7)
828 Default value: \fB0\fR.
834 \fBzfs_arc_pc_percent\fR (uint)
837 Percent of pagecache to reclaim arc to
839 This tunable allows ZFS arc to play more nicely with the kernel's LRU
840 pagecache. It can guarantee that the arc size won't collapse under scanning
841 pressure on the pagecache, yet still allows arc to be reclaimed down to
842 zfs_arc_min if necessary. This value is specified as percent of pagecache
843 size (as measured by NR_FILE_PAGES) where that percent may exceed 100. This
844 only operates during memory pressure/reclaim.
846 Default value: \fB0\fR% (disabled).
852 \fBzfs_arc_sys_free\fR (ulong)
855 The target number of bytes the ARC should leave as free memory on the system.
856 Defaults to the larger of 1/64 of physical memory or 512K. Setting this
857 option to a non-zero value will override the default.
859 Default value: \fB0\fR.
865 \fBzfs_autoimport_disable\fR (int)
868 Disable pool import at module load by ignoring the cache file (typically \fB/etc/zfs/zpool.cache\fR).
870 Use \fB1\fR for yes (default) and \fB0\fR for no.
876 \fBzfs_checksums_per_second\fR (int)
879 Rate limit checksum events to this many per second. Note that this should
880 not be set below the zed thresholds (currently 10 checksums over 10 sec)
881 or else zed may not trigger any action.
889 \fBzfs_commit_timeout_pct\fR (int)
892 This controls the amount of time that a ZIL block (lwb) will remain "open"
893 when it isn't "full", and it has a thread waiting for it to be committed to
894 stable storage. The timeout is scaled based on a percentage of the last lwb
895 latency to avoid significantly impacting the latency of each individual
896 transaction record (itx).
898 Default value: \fB5\fR%.
904 \fBzfs_condense_indirect_vdevs_enable\fR (int)
907 Enable condensing indirect vdev mappings. When set to a non-zero value,
908 attempt to condense indirect vdev mappings if the mapping uses more than
909 \fBzfs_condense_min_mapping_bytes\fR bytes of memory and if the obsolete
910 space map object uses more than \fBzfs_condense_max_obsolete_bytes\fR
911 bytes on-disk. The condensing process is an attempt to save memory by
912 removing obsolete mappings.
914 Default value: \fB1\fR.
920 \fBzfs_condense_max_obsolete_bytes\fR (ulong)
923 Only attempt to condense indirect vdev mappings if the on-disk size
924 of the obsolete space map object is greater than this number of bytes
925 (see \fBfBzfs_condense_indirect_vdevs_enable\fR).
927 Default value: \fB1,073,741,824\fR.
933 \fBzfs_condense_min_mapping_bytes\fR (ulong)
936 Minimum size vdev mapping to attempt to condense (see
937 \fBzfs_condense_indirect_vdevs_enable\fR).
939 Default value: \fB131,072\fR.
945 \fBzfs_dbgmsg_enable\fR (int)
948 Internally ZFS keeps a small log to facilitate debugging. By default the log
949 is disabled, to enable it set this option to 1. The contents of the log can
950 be accessed by reading the /proc/spl/kstat/zfs/dbgmsg file. Writing 0 to
951 this proc file clears the log.
953 Default value: \fB0\fR.
959 \fBzfs_dbgmsg_maxsize\fR (int)
962 The maximum size in bytes of the internal ZFS debug log.
964 Default value: \fB4M\fR.
970 \fBzfs_dbuf_state_index\fR (int)
973 This feature is currently unused. It is normally used for controlling what
974 reporting is available under /proc/spl/kstat/zfs.
976 Default value: \fB0\fR.
982 \fBzfs_deadman_enabled\fR (int)
985 When a pool sync operation takes longer than \fBzfs_deadman_synctime_ms\fR
986 milliseconds, or when an individual I/O takes longer than
987 \fBzfs_deadman_ziotime_ms\fR milliseconds, then the operation is considered to
988 be "hung". If \fBzfs_deadman_enabled\fR is set then the deadman behavior is
989 invoked as described by the \fBzfs_deadman_failmode\fR module option.
990 By default the deadman is enabled and configured to \fBwait\fR which results
991 in "hung" I/Os only being logged. The deadman is automatically disabled
992 when a pool gets suspended.
994 Default value: \fB1\fR.
1000 \fBzfs_deadman_failmode\fR (charp)
1003 Controls the failure behavior when the deadman detects a "hung" I/O. Valid
1004 values are \fBwait\fR, \fBcontinue\fR, and \fBpanic\fR.
1006 \fBwait\fR - Wait for a "hung" I/O to complete. For each "hung" I/O a
1007 "deadman" event will be posted describing that I/O.
1009 \fBcontinue\fR - Attempt to recover from a "hung" I/O by re-dispatching it
1010 to the I/O pipeline if possible.
1012 \fBpanic\fR - Panic the system. This can be used to facilitate an automatic
1013 fail-over to a properly configured fail-over partner.
1015 Default value: \fBwait\fR.
1021 \fBzfs_deadman_checktime_ms\fR (int)
1024 Check time in milliseconds. This defines the frequency at which we check
1025 for hung I/O and potentially invoke the \fBzfs_deadman_failmode\fR behavior.
1027 Default value: \fB60,000\fR.
1033 \fBzfs_deadman_synctime_ms\fR (ulong)
1036 Interval in milliseconds after which the deadman is triggered and also
1037 the interval after which a pool sync operation is considered to be "hung".
1038 Once this limit is exceeded the deadman will be invoked every
1039 \fBzfs_deadman_checktime_ms\fR milliseconds until the pool sync completes.
1041 Default value: \fB600,000\fR.
1047 \fBzfs_deadman_ziotime_ms\fR (ulong)
1050 Interval in milliseconds after which the deadman is triggered and an
1051 individual IO operation is considered to be "hung". As long as the I/O
1052 remains "hung" the deadman will be invoked every \fBzfs_deadman_checktime_ms\fR
1053 milliseconds until the I/O completes.
1055 Default value: \fB300,000\fR.
1061 \fBzfs_dedup_prefetch\fR (int)
1064 Enable prefetching dedup-ed blks
1066 Use \fB1\fR for yes and \fB0\fR to disable (default).
1072 \fBzfs_delay_min_dirty_percent\fR (int)
1075 Start to delay each transaction once there is this amount of dirty data,
1076 expressed as a percentage of \fBzfs_dirty_data_max\fR.
1077 This value should be >= zfs_vdev_async_write_active_max_dirty_percent.
1078 See the section "ZFS TRANSACTION DELAY".
1080 Default value: \fB60\fR%.
1086 \fBzfs_delay_scale\fR (int)
1089 This controls how quickly the transaction delay approaches infinity.
1090 Larger values cause longer delays for a given amount of dirty data.
1092 For the smoothest delay, this value should be about 1 billion divided
1093 by the maximum number of operations per second. This will smoothly
1094 handle between 10x and 1/10th this number.
1096 See the section "ZFS TRANSACTION DELAY".
1098 Note: \fBzfs_delay_scale\fR * \fBzfs_dirty_data_max\fR must be < 2^64.
1100 Default value: \fB500,000\fR.
1106 \fBzfs_delays_per_second\fR (int)
1109 Rate limit IO delay events to this many per second.
1117 \fBzfs_delete_blocks\fR (ulong)
1120 This is the used to define a large file for the purposes of delete. Files
1121 containing more than \fBzfs_delete_blocks\fR will be deleted asynchronously
1122 while smaller files are deleted synchronously. Decreasing this value will
1123 reduce the time spent in an unlink(2) system call at the expense of a longer
1124 delay before the freed space is available.
1126 Default value: \fB20,480\fR.
1132 \fBzfs_dirty_data_max\fR (int)
1135 Determines the dirty space limit in bytes. Once this limit is exceeded, new
1136 writes are halted until space frees up. This parameter takes precedence
1137 over \fBzfs_dirty_data_max_percent\fR.
1138 See the section "ZFS TRANSACTION DELAY".
1140 Default value: \fB10\fR% of physical RAM, capped at \fBzfs_dirty_data_max_max\fR.
1146 \fBzfs_dirty_data_max_max\fR (int)
1149 Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed in bytes.
1150 This limit is only enforced at module load time, and will be ignored if
1151 \fBzfs_dirty_data_max\fR is later changed. This parameter takes
1152 precedence over \fBzfs_dirty_data_max_max_percent\fR. See the section
1153 "ZFS TRANSACTION DELAY".
1155 Default value: \fB25\fR% of physical RAM.
1161 \fBzfs_dirty_data_max_max_percent\fR (int)
1164 Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed as a
1165 percentage of physical RAM. This limit is only enforced at module load
1166 time, and will be ignored if \fBzfs_dirty_data_max\fR is later changed.
1167 The parameter \fBzfs_dirty_data_max_max\fR takes precedence over this
1168 one. See the section "ZFS TRANSACTION DELAY".
1170 Default value: \fB25\fR%.
1176 \fBzfs_dirty_data_max_percent\fR (int)
1179 Determines the dirty space limit, expressed as a percentage of all
1180 memory. Once this limit is exceeded, new writes are halted until space frees
1181 up. The parameter \fBzfs_dirty_data_max\fR takes precedence over this
1182 one. See the section "ZFS TRANSACTION DELAY".
1184 Default value: \fB10\fR%, subject to \fBzfs_dirty_data_max_max\fR.
1190 \fBzfs_dirty_data_sync\fR (int)
1193 Start syncing out a transaction group if there is at least this much dirty data.
1195 Default value: \fB67,108,864\fR.
1201 \fBzfs_fletcher_4_impl\fR (string)
1204 Select a fletcher 4 implementation.
1206 Supported selectors are: \fBfastest\fR, \fBscalar\fR, \fBsse2\fR, \fBssse3\fR,
1207 \fBavx2\fR, \fBavx512f\fR, and \fBaarch64_neon\fR.
1208 All of the selectors except \fBfastest\fR and \fBscalar\fR require instruction
1209 set extensions to be available and will only appear if ZFS detects that they are
1210 present at runtime. If multiple implementations of fletcher 4 are available,
1211 the \fBfastest\fR will be chosen using a micro benchmark. Selecting \fBscalar\fR
1212 results in the original, CPU based calculation, being used. Selecting any option
1213 other than \fBfastest\fR and \fBscalar\fR results in vector instructions from
1214 the respective CPU instruction set being used.
1216 Default value: \fBfastest\fR.
1222 \fBzfs_free_bpobj_enabled\fR (int)
1225 Enable/disable the processing of the free_bpobj object.
1227 Default value: \fB1\fR.
1233 \fBzfs_async_block_max_blocks\fR (ulong)
1236 Maximum number of blocks freed in a single txg.
1238 Default value: \fB100,000\fR.
1244 \fBzfs_override_estimate_recordsize\fR (ulong)
1247 Record size calculation override for zfs send estimates.
1249 Default value: \fB0\fR.
1255 \fBzfs_vdev_async_read_max_active\fR (int)
1258 Maximum asynchronous read I/Os active to each device.
1259 See the section "ZFS I/O SCHEDULER".
1261 Default value: \fB3\fR.
1267 \fBzfs_vdev_async_read_min_active\fR (int)
1270 Minimum asynchronous read I/Os active to each device.
1271 See the section "ZFS I/O SCHEDULER".
1273 Default value: \fB1\fR.
1279 \fBzfs_vdev_async_write_active_max_dirty_percent\fR (int)
1282 When the pool has more than
1283 \fBzfs_vdev_async_write_active_max_dirty_percent\fR dirty data, use
1284 \fBzfs_vdev_async_write_max_active\fR to limit active async writes. If
1285 the dirty data is between min and max, the active I/O limit is linearly
1286 interpolated. See the section "ZFS I/O SCHEDULER".
1288 Default value: \fB60\fR%.
1294 \fBzfs_vdev_async_write_active_min_dirty_percent\fR (int)
1297 When the pool has less than
1298 \fBzfs_vdev_async_write_active_min_dirty_percent\fR dirty data, use
1299 \fBzfs_vdev_async_write_min_active\fR to limit active async writes. If
1300 the dirty data is between min and max, the active I/O limit is linearly
1301 interpolated. See the section "ZFS I/O SCHEDULER".
1303 Default value: \fB30\fR%.
1309 \fBzfs_vdev_async_write_max_active\fR (int)
1312 Maximum asynchronous write I/Os active to each device.
1313 See the section "ZFS I/O SCHEDULER".
1315 Default value: \fB10\fR.
1321 \fBzfs_vdev_async_write_min_active\fR (int)
1324 Minimum asynchronous write I/Os active to each device.
1325 See the section "ZFS I/O SCHEDULER".
1327 Lower values are associated with better latency on rotational media but poorer
1328 resilver performance. The default value of 2 was chosen as a compromise. A
1329 value of 3 has been shown to improve resilver performance further at a cost of
1330 further increasing latency.
1332 Default value: \fB2\fR.
1338 \fBzfs_vdev_max_active\fR (int)
1341 The maximum number of I/Os active to each device. Ideally, this will be >=
1342 the sum of each queue's max_active. It must be at least the sum of each
1343 queue's min_active. See the section "ZFS I/O SCHEDULER".
1345 Default value: \fB1,000\fR.
1351 \fBzfs_vdev_scrub_max_active\fR (int)
1354 Maximum scrub I/Os active to each device.
1355 See the section "ZFS I/O SCHEDULER".
1357 Default value: \fB2\fR.
1363 \fBzfs_vdev_scrub_min_active\fR (int)
1366 Minimum scrub I/Os active to each device.
1367 See the section "ZFS I/O SCHEDULER".
1369 Default value: \fB1\fR.
1375 \fBzfs_vdev_sync_read_max_active\fR (int)
1378 Maximum synchronous read I/Os active to each device.
1379 See the section "ZFS I/O SCHEDULER".
1381 Default value: \fB10\fR.
1387 \fBzfs_vdev_sync_read_min_active\fR (int)
1390 Minimum synchronous read I/Os active to each device.
1391 See the section "ZFS I/O SCHEDULER".
1393 Default value: \fB10\fR.
1399 \fBzfs_vdev_sync_write_max_active\fR (int)
1402 Maximum synchronous write I/Os active to each device.
1403 See the section "ZFS I/O SCHEDULER".
1405 Default value: \fB10\fR.
1411 \fBzfs_vdev_sync_write_min_active\fR (int)
1414 Minimum synchronous write I/Os active to each device.
1415 See the section "ZFS I/O SCHEDULER".
1417 Default value: \fB10\fR.
1423 \fBzfs_vdev_queue_depth_pct\fR (int)
1426 Maximum number of queued allocations per top-level vdev expressed as
1427 a percentage of \fBzfs_vdev_async_write_max_active\fR which allows the
1428 system to detect devices that are more capable of handling allocations
1429 and to allocate more blocks to those devices. It allows for dynamic
1430 allocation distribution when devices are imbalanced as fuller devices
1431 will tend to be slower than empty devices.
1433 See also \fBzio_dva_throttle_enabled\fR.
1435 Default value: \fB1000\fR%.
1441 \fBzfs_expire_snapshot\fR (int)
1444 Seconds to expire .zfs/snapshot
1446 Default value: \fB300\fR.
1452 \fBzfs_admin_snapshot\fR (int)
1455 Allow the creation, removal, or renaming of entries in the .zfs/snapshot
1456 directory to cause the creation, destruction, or renaming of snapshots.
1457 When enabled this functionality works both locally and over NFS exports
1458 which have the 'no_root_squash' option set. This functionality is disabled
1461 Use \fB1\fR for yes and \fB0\fR for no (default).
1467 \fBzfs_flags\fR (int)
1470 Set additional debugging flags. The following flags may be bitwise-or'd
1482 Enable dprintf entries in the debug log.
1484 2 ZFS_DEBUG_DBUF_VERIFY *
1485 Enable extra dbuf verifications.
1487 4 ZFS_DEBUG_DNODE_VERIFY *
1488 Enable extra dnode verifications.
1490 8 ZFS_DEBUG_SNAPNAMES
1491 Enable snapshot name verification.
1494 Check for illegally modified ARC buffers.
1496 64 ZFS_DEBUG_ZIO_FREE
1497 Enable verification of block frees.
1499 128 ZFS_DEBUG_HISTOGRAM_VERIFY
1500 Enable extra spacemap histogram verifications.
1502 256 ZFS_DEBUG_METASLAB_VERIFY
1503 Verify space accounting on disk matches in-core range_trees.
1505 512 ZFS_DEBUG_SET_ERROR
1506 Enable SET_ERROR and dprintf entries in the debug log.
1509 * Requires debug build.
1511 Default value: \fB0\fR.
1517 \fBzfs_free_leak_on_eio\fR (int)
1520 If destroy encounters an EIO while reading metadata (e.g. indirect
1521 blocks), space referenced by the missing metadata can not be freed.
1522 Normally this causes the background destroy to become "stalled", as
1523 it is unable to make forward progress. While in this stalled state,
1524 all remaining space to free from the error-encountering filesystem is
1525 "temporarily leaked". Set this flag to cause it to ignore the EIO,
1526 permanently leak the space from indirect blocks that can not be read,
1527 and continue to free everything else that it can.
1529 The default, "stalling" behavior is useful if the storage partially
1530 fails (i.e. some but not all i/os fail), and then later recovers. In
1531 this case, we will be able to continue pool operations while it is
1532 partially failed, and when it recovers, we can continue to free the
1533 space, with no leaks. However, note that this case is actually
1536 Typically pools either (a) fail completely (but perhaps temporarily,
1537 e.g. a top-level vdev going offline), or (b) have localized,
1538 permanent errors (e.g. disk returns the wrong data due to bit flip or
1539 firmware bug). In case (a), this setting does not matter because the
1540 pool will be suspended and the sync thread will not be able to make
1541 forward progress regardless. In case (b), because the error is
1542 permanent, the best we can do is leak the minimum amount of space,
1543 which is what setting this flag will do. Therefore, it is reasonable
1544 for this flag to normally be set, but we chose the more conservative
1545 approach of not setting it, so that there is no possibility of
1546 leaking space in the "partial temporary" failure case.
1548 Default value: \fB0\fR.
1554 \fBzfs_free_min_time_ms\fR (int)
1557 During a \fBzfs destroy\fR operation using \fBfeature@async_destroy\fR a minimum
1558 of this much time will be spent working on freeing blocks per txg.
1560 Default value: \fB1,000\fR.
1566 \fBzfs_immediate_write_sz\fR (long)
1569 Largest data block to write to zil. Larger blocks will be treated as if the
1570 dataset being written to had the property setting \fBlogbias=throughput\fR.
1572 Default value: \fB32,768\fR.
1578 \fBzfs_lua_max_instrlimit\fR (ulong)
1581 The maximum execution time limit that can be set for a ZFS channel program,
1582 specified as a number of Lua instructions.
1584 Default value: \fB100,000,000\fR.
1590 \fBzfs_lua_max_memlimit\fR (ulong)
1593 The maximum memory limit that can be set for a ZFS channel program, specified
1596 Default value: \fB104,857,600\fR.
1602 \fBzfs_max_dataset_nesting\fR (int)
1605 The maximum depth of nested datasets. This value can be tuned temporarily to
1606 fix existing datasets that exceed the predefined limit.
1608 Default value: \fB50\fR.
1614 \fBzfs_max_recordsize\fR (int)
1617 We currently support block sizes from 512 bytes to 16MB. The benefits of
1618 larger blocks, and thus larger IO, need to be weighed against the cost of
1619 COWing a giant block to modify one byte. Additionally, very large blocks
1620 can have an impact on i/o latency, and also potentially on the memory
1621 allocator. Therefore, we do not allow the recordsize to be set larger than
1622 zfs_max_recordsize (default 1MB). Larger blocks can be created by changing
1623 this tunable, and pools with larger blocks can always be imported and used,
1624 regardless of this setting.
1626 Default value: \fB1,048,576\fR.
1632 \fBzfs_metaslab_fragmentation_threshold\fR (int)
1635 Allow metaslabs to keep their active state as long as their fragmentation
1636 percentage is less than or equal to this value. An active metaslab that
1637 exceeds this threshold will no longer keep its active status allowing
1638 better metaslabs to be selected.
1640 Default value: \fB70\fR.
1646 \fBzfs_mg_fragmentation_threshold\fR (int)
1649 Metaslab groups are considered eligible for allocations if their
1650 fragmentation metric (measured as a percentage) is less than or equal to
1651 this value. If a metaslab group exceeds this threshold then it will be
1652 skipped unless all metaslab groups within the metaslab class have also
1653 crossed this threshold.
1655 Default value: \fB85\fR.
1661 \fBzfs_mg_noalloc_threshold\fR (int)
1664 Defines a threshold at which metaslab groups should be eligible for
1665 allocations. The value is expressed as a percentage of free space
1666 beyond which a metaslab group is always eligible for allocations.
1667 If a metaslab group's free space is less than or equal to the
1668 threshold, the allocator will avoid allocating to that group
1669 unless all groups in the pool have reached the threshold. Once all
1670 groups have reached the threshold, all groups are allowed to accept
1671 allocations. The default value of 0 disables the feature and causes
1672 all metaslab groups to be eligible for allocations.
1674 This parameter allows one to deal with pools having heavily imbalanced
1675 vdevs such as would be the case when a new vdev has been added.
1676 Setting the threshold to a non-zero percentage will stop allocations
1677 from being made to vdevs that aren't filled to the specified percentage
1678 and allow lesser filled vdevs to acquire more allocations than they
1679 otherwise would under the old \fBzfs_mg_alloc_failures\fR facility.
1681 Default value: \fB0\fR.
1687 \fBzfs_multihost_history\fR (int)
1690 Historical statistics for the last N multihost updates will be available in
1691 \fB/proc/spl/kstat/zfs/<pool>/multihost\fR
1693 Default value: \fB0\fR.
1699 \fBzfs_multihost_interval\fR (ulong)
1702 Used to control the frequency of multihost writes which are performed when the
1703 \fBmultihost\fR pool property is on. This is one factor used to determine
1704 the length of the activity check during import.
1706 The multihost write period is \fBzfs_multihost_interval / leaf-vdevs\fR milliseconds.
1707 This means that on average a multihost write will be issued for each leaf vdev every
1708 \fBzfs_multihost_interval\fR milliseconds. In practice, the observed period can
1709 vary with the I/O load and this observed value is the delay which is stored in
1712 On import the activity check waits a minimum amount of time determined by
1713 \fBzfs_multihost_interval * zfs_multihost_import_intervals\fR. The activity
1714 check time may be further extended if the value of mmp delay found in the best
1715 uberblock indicates actual multihost updates happened at longer intervals than
1716 \fBzfs_multihost_interval\fR. A minimum value of \fB100ms\fR is enforced.
1718 Default value: \fB1000\fR.
1724 \fBzfs_multihost_import_intervals\fR (uint)
1727 Used to control the duration of the activity test on import. Smaller values of
1728 \fBzfs_multihost_import_intervals\fR will reduce the import time but increase
1729 the risk of failing to detect an active pool. The total activity check time is
1730 never allowed to drop below one second. A value of 0 is ignored and treated as
1733 Default value: \fB10\fR.
1739 \fBzfs_multihost_fail_intervals\fR (uint)
1742 Controls the behavior of the pool when multihost write failures are detected.
1744 When \fBzfs_multihost_fail_intervals = 0\fR then multihost write failures are ignored.
1745 The failures will still be reported to the ZED which depending on its
1746 configuration may take action such as suspending the pool or offlining a device.
1748 When \fBzfs_multihost_fail_intervals > 0\fR then sequential multihost write failures
1749 will cause the pool to be suspended. This occurs when
1750 \fBzfs_multihost_fail_intervals * zfs_multihost_interval\fR milliseconds have
1751 passed since the last successful multihost write. This guarantees the activity test
1752 will see multihost writes if the pool is imported.
1754 Default value: \fB5\fR.
1760 \fBzfs_no_scrub_io\fR (int)
1763 Set for no scrub I/O. This results in scrubs not actually scrubbing data and
1764 simply doing a metadata crawl of the pool instead.
1766 Use \fB1\fR for yes and \fB0\fR for no (default).
1772 \fBzfs_no_scrub_prefetch\fR (int)
1775 Set to disable block prefetching for scrubs.
1777 Use \fB1\fR for yes and \fB0\fR for no (default).
1783 \fBzfs_nocacheflush\fR (int)
1786 Disable cache flush operations on disks when writing. Beware, this may cause
1787 corruption if disks re-order writes.
1789 Use \fB1\fR for yes and \fB0\fR for no (default).
1795 \fBzfs_nopwrite_enabled\fR (int)
1800 Use \fB1\fR for yes (default) and \fB0\fR to disable.
1806 \fBzfs_dmu_offset_next_sync\fR (int)
1809 Enable forcing txg sync to find holes. When enabled forces ZFS to act
1810 like prior versions when SEEK_HOLE or SEEK_DATA flags are used, which
1811 when a dnode is dirty causes txg's to be synced so that this data can be
1814 Use \fB1\fR for yes and \fB0\fR to disable (default).
1820 \fBzfs_pd_bytes_max\fR (int)
1823 The number of bytes which should be prefetched during a pool traversal
1824 (eg: \fBzfs send\fR or other data crawling operations)
1826 Default value: \fB52,428,800\fR.
1832 \fBzfs_per_txg_dirty_frees_percent \fR (ulong)
1835 Tunable to control percentage of dirtied blocks from frees in one TXG.
1836 After this threshold is crossed, additional dirty blocks from frees
1837 wait until the next TXG.
1838 A value of zero will disable this throttle.
1840 Default value: \fB30\fR and \fB0\fR to disable.
1848 \fBzfs_prefetch_disable\fR (int)
1851 This tunable disables predictive prefetch. Note that it leaves "prescient"
1852 prefetch (e.g. prefetch for zfs send) intact. Unlike predictive prefetch,
1853 prescient prefetch never issues i/os that end up not being needed, so it
1854 can't hurt performance.
1856 Use \fB1\fR for yes and \fB0\fR for no (default).
1862 \fBzfs_read_chunk_size\fR (long)
1865 Bytes to read per chunk
1867 Default value: \fB1,048,576\fR.
1873 \fBzfs_read_history\fR (int)
1876 Historical statistics for the last N reads will be available in
1877 \fB/proc/spl/kstat/zfs/<pool>/reads\fR
1879 Default value: \fB0\fR (no data is kept).
1885 \fBzfs_read_history_hits\fR (int)
1888 Include cache hits in read history
1890 Use \fB1\fR for yes and \fB0\fR for no (default).
1896 \fBzfs_reconstruct_indirect_combinations_max\fR (int)
1899 If an indirect split block contains more than this many possible unique
1900 combinations when being reconstructed, consider it too computationally
1901 expensive to check them all. Instead, try at most
1902 \fBzfs_reconstruct_indirect_combinations_max\fR randomly-selected
1903 combinations each time the block is accessed. This allows all segment
1904 copies to participate fairly in the reconstruction when all combinations
1905 cannot be checked and prevents repeated use of one bad copy.
1907 Default value: \fB100\fR.
1913 \fBzfs_recover\fR (int)
1916 Set to attempt to recover from fatal errors. This should only be used as a
1917 last resort, as it typically results in leaked space, or worse.
1919 Use \fB1\fR for yes and \fB0\fR for no (default).
1925 \fBzfs_resilver_min_time_ms\fR (int)
1928 Resilvers are processed by the sync thread. While resilvering it will spend
1929 at least this much time working on a resilver between txg flushes.
1931 Default value: \fB3,000\fR.
1937 \fBzfs_scan_ignore_errors\fR (int)
1940 If set to a nonzero value, remove the DTL (dirty time list) upon
1941 completion of a pool scan (scrub) even if there were unrepairable
1942 errors. It is intended to be used during pool repair or recovery to
1943 stop resilvering when the pool is next imported.
1945 Default value: \fB0\fR.
1951 \fBzfs_scrub_min_time_ms\fR (int)
1954 Scrubs are processed by the sync thread. While scrubbing it will spend
1955 at least this much time working on a scrub between txg flushes.
1957 Default value: \fB1,000\fR.
1963 \fBzfs_scan_checkpoint_intval\fR (int)
1966 To preserve progress across reboots the sequential scan algorithm periodically
1967 needs to stop metadata scanning and issue all the verifications I/Os to disk.
1968 The frequency of this flushing is determined by the
1969 \fBzfs_scan_checkpoint_intval\fR tunable.
1971 Default value: \fB7200\fR seconds (every 2 hours).
1977 \fBzfs_scan_fill_weight\fR (int)
1980 This tunable affects how scrub and resilver I/O segments are ordered. A higher
1981 number indicates that we care more about how filled in a segment is, while a
1982 lower number indicates we care more about the size of the extent without
1983 considering the gaps within a segment. This value is only tunable upon module
1984 insertion. Changing the value afterwards will have no affect on scrub or
1985 resilver performance.
1987 Default value: \fB3\fR.
1993 \fBzfs_scan_issue_strategy\fR (int)
1996 Determines the order that data will be verified while scrubbing or resilvering.
1997 If set to \fB1\fR, data will be verified as sequentially as possible, given the
1998 amount of memory reserved for scrubbing (see \fBzfs_scan_mem_lim_fact\fR). This
1999 may improve scrub performance if the pool's data is very fragmented. If set to
2000 \fB2\fR, the largest mostly-contiguous chunk of found data will be verified
2001 first. By deferring scrubbing of small segments, we may later find adjacent data
2002 to coalesce and increase the segment size. If set to \fB0\fR, zfs will use
2003 strategy \fB1\fR during normal verification and strategy \fB2\fR while taking a
2006 Default value: \fB0\fR.
2012 \fBzfs_scan_legacy\fR (int)
2015 A value of 0 indicates that scrubs and resilvers will gather metadata in
2016 memory before issuing sequential I/O. A value of 1 indicates that the legacy
2017 algorithm will be used where I/O is initiated as soon as it is discovered.
2018 Changing this value to 0 will not affect scrubs or resilvers that are already
2021 Default value: \fB0\fR.
2027 \fBzfs_scan_max_ext_gap\fR (int)
2030 Indicates the largest gap in bytes between scrub / resilver I/Os that will still
2031 be considered sequential for sorting purposes. Changing this value will not
2032 affect scrubs or resilvers that are already in progress.
2034 Default value: \fB2097152 (2 MB)\fR.
2040 \fBzfs_scan_mem_lim_fact\fR (int)
2043 Maximum fraction of RAM used for I/O sorting by sequential scan algorithm.
2044 This tunable determines the hard limit for I/O sorting memory usage.
2045 When the hard limit is reached we stop scanning metadata and start issuing
2046 data verification I/O. This is done until we get below the soft limit.
2048 Default value: \fB20\fR which is 5% of RAM (1/20).
2054 \fBzfs_scan_mem_lim_soft_fact\fR (int)
2057 The fraction of the hard limit used to determined the soft limit for I/O sorting
2058 by the sequential scan algorithm. When we cross this limit from bellow no action
2059 is taken. When we cross this limit from above it is because we are issuing
2060 verification I/O. In this case (unless the metadata scan is done) we stop
2061 issuing verification I/O and start scanning metadata again until we get to the
2064 Default value: \fB20\fR which is 5% of the hard limit (1/20).
2070 \fBzfs_scan_vdev_limit\fR (int)
2073 Maximum amount of data that can be concurrently issued at once for scrubs and
2074 resilvers per leaf device, given in bytes.
2076 Default value: \fB41943040\fR.
2082 \fBzfs_send_corrupt_data\fR (int)
2085 Allow sending of corrupt data (ignore read/checksum errors when sending data)
2087 Use \fB1\fR for yes and \fB0\fR for no (default).
2093 \fBzfs_send_queue_length\fR (int)
2096 The maximum number of bytes allowed in the \fBzfs send\fR queue. This value
2097 must be at least twice the maximum block size in use.
2099 Default value: \fB16,777,216\fR.
2105 \fBzfs_recv_queue_length\fR (int)
2109 The maximum number of bytes allowed in the \fBzfs receive\fR queue. This value
2110 must be at least twice the maximum block size in use.
2112 Default value: \fB16,777,216\fR.
2118 \fBzfs_sync_pass_deferred_free\fR (int)
2121 Flushing of data to disk is done in passes. Defer frees starting in this pass
2123 Default value: \fB2\fR.
2129 \fBzfs_spa_discard_memory_limit\fR (int)
2132 Maximum memory used for prefetching a checkpoint's space map on each
2133 vdev while discarding the checkpoint.
2135 Default value: \fB16,777,216\fR.
2141 \fBzfs_sync_pass_dont_compress\fR (int)
2144 Don't compress starting in this pass
2146 Default value: \fB5\fR.
2152 \fBzfs_sync_pass_rewrite\fR (int)
2155 Rewrite new block pointers starting in this pass
2157 Default value: \fB2\fR.
2163 \fBzfs_sync_taskq_batch_pct\fR (int)
2166 This controls the number of threads used by the dp_sync_taskq. The default
2167 value of 75% will create a maximum of one thread per cpu.
2169 Default value: \fB75\fR%.
2175 \fBzfs_txg_history\fR (int)
2178 Historical statistics for the last N txgs will be available in
2179 \fB/proc/spl/kstat/zfs/<pool>/txgs\fR
2181 Default value: \fB0\fR.
2187 \fBzfs_txg_timeout\fR (int)
2190 Flush dirty data to disk at least every N seconds (maximum txg duration)
2192 Default value: \fB5\fR.
2198 \fBzfs_vdev_aggregation_limit\fR (int)
2201 Max vdev I/O aggregation size
2203 Default value: \fB131,072\fR.
2209 \fBzfs_vdev_cache_bshift\fR (int)
2212 Shift size to inflate reads too
2214 Default value: \fB16\fR (effectively 65536).
2220 \fBzfs_vdev_cache_max\fR (int)
2223 Inflate reads smaller than this value to meet the \fBzfs_vdev_cache_bshift\fR
2226 Default value: \fB16384\fR.
2232 \fBzfs_vdev_cache_size\fR (int)
2235 Total size of the per-disk cache in bytes.
2237 Currently this feature is disabled as it has been found to not be helpful
2238 for performance and in some cases harmful.
2240 Default value: \fB0\fR.
2246 \fBzfs_vdev_mirror_rotating_inc\fR (int)
2249 A number by which the balancing algorithm increments the load calculation for
2250 the purpose of selecting the least busy mirror member when an I/O immediately
2251 follows its predecessor on rotational vdevs for the purpose of making decisions
2254 Default value: \fB0\fR.
2260 \fBzfs_vdev_mirror_rotating_seek_inc\fR (int)
2263 A number by which the balancing algorithm increments the load calculation for
2264 the purpose of selecting the least busy mirror member when an I/O lacks
2265 locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within
2266 this that are not immediately following the previous I/O are incremented by
2269 Default value: \fB5\fR.
2275 \fBzfs_vdev_mirror_rotating_seek_offset\fR (int)
2278 The maximum distance for the last queued I/O in which the balancing algorithm
2279 considers an I/O to have locality.
2280 See the section "ZFS I/O SCHEDULER".
2282 Default value: \fB1048576\fR.
2288 \fBzfs_vdev_mirror_non_rotating_inc\fR (int)
2291 A number by which the balancing algorithm increments the load calculation for
2292 the purpose of selecting the least busy mirror member on non-rotational vdevs
2293 when I/Os do not immediately follow one another.
2295 Default value: \fB0\fR.
2301 \fBzfs_vdev_mirror_non_rotating_seek_inc\fR (int)
2304 A number by which the balancing algorithm increments the load calculation for
2305 the purpose of selecting the least busy mirror member when an I/O lacks
2306 locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within
2307 this that are not immediately following the previous I/O are incremented by
2310 Default value: \fB1\fR.
2316 \fBzfs_vdev_read_gap_limit\fR (int)
2319 Aggregate read I/O operations if the gap on-disk between them is within this
2322 Default value: \fB32,768\fR.
2328 \fBzfs_vdev_scheduler\fR (charp)
2331 Set the Linux I/O scheduler on whole disk vdevs to this scheduler. Valid options
2332 are noop, cfq, bfq & deadline
2334 Default value: \fBnoop\fR.
2340 \fBzfs_vdev_write_gap_limit\fR (int)
2343 Aggregate write I/O over gap
2345 Default value: \fB4,096\fR.
2351 \fBzfs_vdev_raidz_impl\fR (string)
2354 Parameter for selecting raidz parity implementation to use.
2356 Options marked (always) below may be selected on module load as they are
2357 supported on all systems.
2358 The remaining options may only be set after the module is loaded, as they
2359 are available only if the implementations are compiled in and supported
2360 on the running system.
2362 Once the module is loaded, the content of
2363 /sys/module/zfs/parameters/zfs_vdev_raidz_impl will show available options
2364 with the currently selected one enclosed in [].
2365 Possible options are:
2366 fastest - (always) implementation selected using built-in benchmark
2367 original - (always) original raidz implementation
2368 scalar - (always) scalar raidz implementation
2369 sse2 - implementation using SSE2 instruction set (64bit x86 only)
2370 ssse3 - implementation using SSSE3 instruction set (64bit x86 only)
2371 avx2 - implementation using AVX2 instruction set (64bit x86 only)
2372 avx512f - implementation using AVX512F instruction set (64bit x86 only)
2373 avx512bw - implementation using AVX512F & AVX512BW instruction sets (64bit x86 only)
2374 aarch64_neon - implementation using NEON (Aarch64/64 bit ARMv8 only)
2375 aarch64_neonx2 - implementation using NEON with more unrolling (Aarch64/64 bit ARMv8 only)
2377 Default value: \fBfastest\fR.
2383 \fBzfs_zevent_cols\fR (int)
2386 When zevents are logged to the console use this as the word wrap width.
2388 Default value: \fB80\fR.
2394 \fBzfs_zevent_console\fR (int)
2397 Log events to the console
2399 Use \fB1\fR for yes and \fB0\fR for no (default).
2405 \fBzfs_zevent_len_max\fR (int)
2408 Max event queue length. A value of 0 will result in a calculated value which
2409 increases with the number of CPUs in the system (minimum 64 events). Events
2410 in the queue can be viewed with the \fBzpool events\fR command.
2412 Default value: \fB0\fR.
2418 \fBzfs_zil_clean_taskq_maxalloc\fR (int)
2421 The maximum number of taskq entries that are allowed to be cached. When this
2422 limit is exceeded transaction records (itxs) will be cleaned synchronously.
2424 Default value: \fB1048576\fR.
2430 \fBzfs_zil_clean_taskq_minalloc\fR (int)
2433 The number of taskq entries that are pre-populated when the taskq is first
2434 created and are immediately available for use.
2436 Default value: \fB1024\fR.
2442 \fBzfs_zil_clean_taskq_nthr_pct\fR (int)
2445 This controls the number of threads used by the dp_zil_clean_taskq. The default
2446 value of 100% will create a maximum of one thread per cpu.
2448 Default value: \fB100\fR%.
2454 \fBzil_replay_disable\fR (int)
2457 Disable intent logging replay. Can be disabled for recovery from corrupted
2460 Use \fB1\fR for yes and \fB0\fR for no (default).
2466 \fBzil_slog_bulk\fR (ulong)
2469 Limit SLOG write size per commit executed with synchronous priority.
2470 Any writes above that will be executed with lower (asynchronous) priority
2471 to limit potential SLOG device abuse by single active ZIL writer.
2473 Default value: \fB786,432\fR.
2479 \fBzio_delay_max\fR (int)
2482 A zevent will be logged if a ZIO operation takes more than N milliseconds to
2483 complete. Note that this is only a logging facility, not a timeout on
2486 Default value: \fB30,000\fR.
2492 \fBzio_dva_throttle_enabled\fR (int)
2495 Throttle block allocations in the ZIO pipeline. This allows for
2496 dynamic allocation distribution when devices are imbalanced.
2497 When enabled, the maximum number of pending allocations per top-level vdev
2498 is limited by \fBzfs_vdev_queue_depth_pct\fR.
2500 Default value: \fB1\fR.
2506 \fBzio_requeue_io_start_cut_in_line\fR (int)
2509 Prioritize requeued I/O
2511 Default value: \fB0\fR.
2517 \fBzio_taskq_batch_pct\fR (uint)
2520 Percentage of online CPUs (or CPU cores, etc) which will run a worker thread
2521 for IO. These workers are responsible for IO work such as compression and
2522 checksum calculations. Fractional number of CPUs will be rounded down.
2524 The default value of 75 was chosen to avoid using all CPUs which can result in
2525 latency issues and inconsistent application performance, especially when high
2526 compression is enabled.
2528 Default value: \fB75\fR.
2534 \fBzvol_inhibit_dev\fR (uint)
2537 Do not create zvol device nodes. This may slightly improve startup time on
2538 systems with a very large number of zvols.
2540 Use \fB1\fR for yes and \fB0\fR for no (default).
2546 \fBzvol_major\fR (uint)
2549 Major number for zvol block devices
2551 Default value: \fB230\fR.
2557 \fBzvol_max_discard_blocks\fR (ulong)
2560 Discard (aka TRIM) operations done on zvols will be done in batches of this
2561 many blocks, where block size is determined by the \fBvolblocksize\fR property
2564 Default value: \fB16,384\fR.
2570 \fBzvol_prefetch_bytes\fR (uint)
2573 When adding a zvol to the system prefetch \fBzvol_prefetch_bytes\fR
2574 from the start and end of the volume. Prefetching these regions
2575 of the volume is desirable because they are likely to be accessed
2576 immediately by \fBblkid(8)\fR or by the kernel scanning for a partition
2579 Default value: \fB131,072\fR.
2585 \fBzvol_request_sync\fR (uint)
2588 When processing I/O requests for a zvol submit them synchronously. This
2589 effectively limits the queue depth to 1 for each I/O submitter. When set
2590 to 0 requests are handled asynchronously by a thread pool. The number of
2591 requests which can be handled concurrently is controller by \fBzvol_threads\fR.
2593 Default value: \fB0\fR.
2599 \fBzvol_threads\fR (uint)
2602 Max number of threads which can handle zvol I/O requests concurrently.
2604 Default value: \fB32\fR.
2610 \fBzvol_volmode\fR (uint)
2613 Defines zvol block devices behaviour when \fBvolmode\fR is set to \fBdefault\fR.
2614 Valid values are \fB1\fR (full), \fB2\fR (dev) and \fB3\fR (none).
2616 Default value: \fB1\fR.
2622 \fBzfs_qat_disable\fR (int)
2625 This tunable disables qat hardware acceleration for gzip compression and.
2626 AES-GCM encryption. It is available only if qat acceleration is compiled in
2627 and the qat driver is present.
2629 Use \fB1\fR for yes and \fB0\fR for no (default).
2632 .SH ZFS I/O SCHEDULER
2633 ZFS issues I/O operations to leaf vdevs to satisfy and complete I/Os.
2634 The I/O scheduler determines when and in what order those operations are
2635 issued. The I/O scheduler divides operations into five I/O classes
2636 prioritized in the following order: sync read, sync write, async read,
2637 async write, and scrub/resilver. Each queue defines the minimum and
2638 maximum number of concurrent operations that may be issued to the
2639 device. In addition, the device has an aggregate maximum,
2640 \fBzfs_vdev_max_active\fR. Note that the sum of the per-queue minimums
2641 must not exceed the aggregate maximum. If the sum of the per-queue
2642 maximums exceeds the aggregate maximum, then the number of active I/Os
2643 may reach \fBzfs_vdev_max_active\fR, in which case no further I/Os will
2644 be issued regardless of whether all per-queue minimums have been met.
2646 For many physical devices, throughput increases with the number of
2647 concurrent operations, but latency typically suffers. Further, physical
2648 devices typically have a limit at which more concurrent operations have no
2649 effect on throughput or can actually cause it to decrease.
2651 The scheduler selects the next operation to issue by first looking for an
2652 I/O class whose minimum has not been satisfied. Once all are satisfied and
2653 the aggregate maximum has not been hit, the scheduler looks for classes
2654 whose maximum has not been satisfied. Iteration through the I/O classes is
2655 done in the order specified above. No further operations are issued if the
2656 aggregate maximum number of concurrent operations has been hit or if there
2657 are no operations queued for an I/O class that has not hit its maximum.
2658 Every time an I/O is queued or an operation completes, the I/O scheduler
2659 looks for new operations to issue.
2661 In general, smaller max_active's will lead to lower latency of synchronous
2662 operations. Larger max_active's may lead to higher overall throughput,
2663 depending on underlying storage.
2665 The ratio of the queues' max_actives determines the balance of performance
2666 between reads, writes, and scrubs. E.g., increasing
2667 \fBzfs_vdev_scrub_max_active\fR will cause the scrub or resilver to complete
2668 more quickly, but reads and writes to have higher latency and lower throughput.
2670 All I/O classes have a fixed maximum number of outstanding operations
2671 except for the async write class. Asynchronous writes represent the data
2672 that is committed to stable storage during the syncing stage for
2673 transaction groups. Transaction groups enter the syncing state
2674 periodically so the number of queued async writes will quickly burst up
2675 and then bleed down to zero. Rather than servicing them as quickly as
2676 possible, the I/O scheduler changes the maximum number of active async
2677 write I/Os according to the amount of dirty data in the pool. Since
2678 both throughput and latency typically increase with the number of
2679 concurrent operations issued to physical devices, reducing the
2680 burstiness in the number of concurrent operations also stabilizes the
2681 response time of operations from other -- and in particular synchronous
2682 -- queues. In broad strokes, the I/O scheduler will issue more
2683 concurrent operations from the async write queue as there's more dirty
2688 The number of concurrent operations issued for the async write I/O class
2689 follows a piece-wise linear function defined by a few adjustable points.
2692 | o---------| <-- zfs_vdev_async_write_max_active
2699 |-------o | | <-- zfs_vdev_async_write_min_active
2700 0|_______^______|_________|
2701 0% | | 100% of zfs_dirty_data_max
2703 | `-- zfs_vdev_async_write_active_max_dirty_percent
2704 `--------- zfs_vdev_async_write_active_min_dirty_percent
2707 Until the amount of dirty data exceeds a minimum percentage of the dirty
2708 data allowed in the pool, the I/O scheduler will limit the number of
2709 concurrent operations to the minimum. As that threshold is crossed, the
2710 number of concurrent operations issued increases linearly to the maximum at
2711 the specified maximum percentage of the dirty data allowed in the pool.
2713 Ideally, the amount of dirty data on a busy pool will stay in the sloped
2714 part of the function between \fBzfs_vdev_async_write_active_min_dirty_percent\fR
2715 and \fBzfs_vdev_async_write_active_max_dirty_percent\fR. If it exceeds the
2716 maximum percentage, this indicates that the rate of incoming data is
2717 greater than the rate that the backend storage can handle. In this case, we
2718 must further throttle incoming writes, as described in the next section.
2720 .SH ZFS TRANSACTION DELAY
2721 We delay transactions when we've determined that the backend storage
2722 isn't able to accommodate the rate of incoming writes.
2724 If there is already a transaction waiting, we delay relative to when
2725 that transaction will finish waiting. This way the calculated delay time
2726 is independent of the number of threads concurrently executing
2729 If we are the only waiter, wait relative to when the transaction
2730 started, rather than the current time. This credits the transaction for
2731 "time already served", e.g. reading indirect blocks.
2733 The minimum time for a transaction to take is calculated as:
2735 min_time = zfs_delay_scale * (dirty - min) / (max - dirty)
2736 min_time is then capped at 100 milliseconds.
2739 The delay has two degrees of freedom that can be adjusted via tunables. The
2740 percentage of dirty data at which we start to delay is defined by
2741 \fBzfs_delay_min_dirty_percent\fR. This should typically be at or above
2742 \fBzfs_vdev_async_write_active_max_dirty_percent\fR so that we only start to
2743 delay after writing at full speed has failed to keep up with the incoming write
2744 rate. The scale of the curve is defined by \fBzfs_delay_scale\fR. Roughly speaking,
2745 this variable determines the amount of delay at the midpoint of the curve.
2749 10ms +-------------------------------------------------------------*+
2765 2ms + (midpoint) * +
2768 | zfs_delay_scale ----------> ******** |
2769 0 +-------------------------------------*********----------------+
2770 0% <- zfs_dirty_data_max -> 100%
2773 Note that since the delay is added to the outstanding time remaining on the
2774 most recent transaction, the delay is effectively the inverse of IOPS.
2775 Here the midpoint of 500us translates to 2000 IOPS. The shape of the curve
2776 was chosen such that small changes in the amount of accumulated dirty data
2777 in the first 3/4 of the curve yield relatively small differences in the
2780 The effects can be easier to understand when the amount of delay is
2781 represented on a log scale:
2785 100ms +-------------------------------------------------------------++
2794 + zfs_delay_scale ----------> ***** +
2805 +--------------------------------------------------------------+
2806 0% <- zfs_dirty_data_max -> 100%
2809 Note here that only as the amount of dirty data approaches its limit does
2810 the delay start to increase rapidly. The goal of a properly tuned system
2811 should be to keep the amount of dirty data out of that range by first
2812 ensuring that the appropriate limits are set for the I/O scheduler to reach
2813 optimal throughput on the backend storage, and then by changing the value
2814 of \fBzfs_delay_scale\fR to increase the steepness of the curve.