2 .\" Copyright (c) 2013 by Turbo Fredriksson <turbo@bayour.com>. All rights reserved.
3 .\" Copyright (c) 2017 Datto Inc.
4 .\" The contents of this file are subject to the terms of the Common Development
5 .\" and Distribution License (the "License"). You may not use this file except
6 .\" in compliance with the License. You can obtain a copy of the license at
7 .\" usr/src/OPENSOLARIS.LICENSE or http://www.opensolaris.org/os/licensing.
9 .\" See the License for the specific language governing permissions and
10 .\" limitations under the License. When distributing Covered Code, include this
11 .\" CDDL HEADER in each file and include the License file at
12 .\" usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this
13 .\" CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your
14 .\" own identifying information:
15 .\" Portions Copyright [yyyy] [name of copyright owner]
16 .TH ZFS-MODULE-PARAMETERS 5 "Oct 28, 2017"
18 zfs\-module\-parameters \- ZFS module parameters
22 Description of the different parameters to the ZFS module.
24 .SS "Module parameters"
31 \fBignore_hole_birth\fR (int)
34 When set, the hole_birth optimization will not be used, and all holes will
35 always be sent on zfs send. Useful if you suspect your datasets are affected
36 by a bug in hole_birth.
38 Use \fB1\fR for on (default) and \fB0\fR for off.
44 \fBl2arc_feed_again\fR (int)
47 Turbo L2ARC warm-up. When the L2ARC is cold the fill interval will be set as
50 Use \fB1\fR for yes (default) and \fB0\fR to disable.
56 \fBl2arc_feed_min_ms\fR (ulong)
59 Min feed interval in milliseconds. Requires \fBl2arc_feed_again=1\fR and only
60 applicable in related situations.
62 Default value: \fB200\fR.
68 \fBl2arc_feed_secs\fR (ulong)
71 Seconds between L2ARC writing
73 Default value: \fB1\fR.
79 \fBl2arc_headroom\fR (ulong)
82 How far through the ARC lists to search for L2ARC cacheable content, expressed
83 as a multiplier of \fBl2arc_write_max\fR
85 Default value: \fB2\fR.
91 \fBl2arc_headroom_boost\fR (ulong)
94 Scales \fBl2arc_headroom\fR by this percentage when L2ARC contents are being
95 successfully compressed before writing. A value of 100 disables this feature.
97 Default value: \fB200\fR%.
103 \fBl2arc_noprefetch\fR (int)
106 Do not write buffers to L2ARC if they were prefetched but not used by
109 Use \fB1\fR for yes (default) and \fB0\fR to disable.
115 \fBl2arc_norw\fR (int)
118 No reads during writes
120 Use \fB1\fR for yes and \fB0\fR for no (default).
126 \fBl2arc_write_boost\fR (ulong)
129 Cold L2ARC devices will have \fBl2arc_write_max\fR increased by this amount
130 while they remain cold.
132 Default value: \fB8,388,608\fR.
138 \fBl2arc_write_max\fR (ulong)
141 Max write bytes per interval
143 Default value: \fB8,388,608\fR.
149 \fBmetaslab_aliquot\fR (ulong)
152 Metaslab granularity, in bytes. This is roughly similar to what would be
153 referred to as the "stripe size" in traditional RAID arrays. In normal
154 operation, ZFS will try to write this amount of data to a top-level vdev
155 before moving on to the next one.
157 Default value: \fB524,288\fR.
163 \fBmetaslab_bias_enabled\fR (int)
166 Enable metaslab group biasing based on its vdev's over- or under-utilization
167 relative to the pool.
169 Use \fB1\fR for yes (default) and \fB0\fR for no.
175 \fBzfs_metaslab_segment_weight_enabled\fR (int)
178 Enable/disable segment-based metaslab selection.
180 Use \fB1\fR for yes (default) and \fB0\fR for no.
186 \fBzfs_metaslab_switch_threshold\fR (int)
189 When using segment-based metaslab selection, continue allocating
190 from the active metaslab until \fBzfs_metaslab_switch_threshold\fR
191 worth of buckets have been exhausted.
193 Default value: \fB2\fR.
199 \fBmetaslab_debug_load\fR (int)
202 Load all metaslabs during pool import.
204 Use \fB1\fR for yes and \fB0\fR for no (default).
210 \fBmetaslab_debug_unload\fR (int)
213 Prevent metaslabs from being unloaded.
215 Use \fB1\fR for yes and \fB0\fR for no (default).
221 \fBmetaslab_fragmentation_factor_enabled\fR (int)
224 Enable use of the fragmentation metric in computing metaslab weights.
226 Use \fB1\fR for yes (default) and \fB0\fR for no.
232 \fBmetaslabs_per_vdev\fR (int)
235 When a vdev is added, it will be divided into approximately (but no more than) this number of metaslabs.
237 Default value: \fB200\fR.
243 \fBmetaslab_preload_enabled\fR (int)
246 Enable metaslab group preloading.
248 Use \fB1\fR for yes (default) and \fB0\fR for no.
254 \fBmetaslab_lba_weighting_enabled\fR (int)
257 Give more weight to metaslabs with lower LBAs, assuming they have
258 greater bandwidth as is typically the case on a modern constant
259 angular velocity disk drive.
261 Use \fB1\fR for yes (default) and \fB0\fR for no.
267 \fBspa_config_path\fR (charp)
272 Default value: \fB/etc/zfs/zpool.cache\fR.
278 \fBspa_asize_inflation\fR (int)
281 Multiplication factor used to estimate actual disk consumption from the
282 size of data being written. The default value is a worst case estimate,
283 but lower values may be valid for a given pool depending on its
284 configuration. Pool administrators who understand the factors involved
285 may wish to specify a more realistic inflation factor, particularly if
286 they operate close to quota or capacity limits.
288 Default value: \fB24\fR.
294 \fBspa_load_verify_data\fR (int)
297 Whether to traverse data blocks during an "extreme rewind" (\fB-X\fR)
298 import. Use 0 to disable and 1 to enable.
300 An extreme rewind import normally performs a full traversal of all
301 blocks in the pool for verification. If this parameter is set to 0,
302 the traversal skips non-metadata blocks. It can be toggled once the
303 import has started to stop or start the traversal of non-metadata blocks.
305 Default value: \fB1\fR.
311 \fBspa_load_verify_metadata\fR (int)
314 Whether to traverse blocks during an "extreme rewind" (\fB-X\fR)
315 pool import. Use 0 to disable and 1 to enable.
317 An extreme rewind import normally performs a full traversal of all
318 blocks in the pool for verification. If this parameter is set to 0,
319 the traversal is not performed. It can be toggled once the import has
320 started to stop or start the traversal.
322 Default value: \fB1\fR.
328 \fBspa_load_verify_maxinflight\fR (int)
331 Maximum concurrent I/Os during the traversal performed during an "extreme
332 rewind" (\fB-X\fR) pool import.
334 Default value: \fB10000\fR.
340 \fBspa_slop_shift\fR (int)
343 Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space
344 in the pool to be consumed. This ensures that we don't run the pool
345 completely out of space, due to unaccounted changes (e.g. to the MOS).
346 It also limits the worst-case time to allocate space. If we have
347 less than this amount of free space, most ZPL operations (e.g. write,
348 create) will return ENOSPC.
350 Default value: \fB5\fR.
356 \fBzfetch_array_rd_sz\fR (ulong)
359 If prefetching is enabled, disable prefetching for reads larger than this size.
361 Default value: \fB1,048,576\fR.
367 \fBzfetch_max_distance\fR (uint)
370 Max bytes to prefetch per stream (default 8MB).
372 Default value: \fB8,388,608\fR.
378 \fBzfetch_max_streams\fR (uint)
381 Max number of streams per zfetch (prefetch streams per file).
383 Default value: \fB8\fR.
389 \fBzfetch_min_sec_reap\fR (uint)
392 Min time before an active prefetch stream can be reclaimed
394 Default value: \fB2\fR.
400 \fBzfs_arc_dnode_limit\fR (ulong)
403 When the number of bytes consumed by dnodes in the ARC exceeds this number of
404 bytes, try to unpin some of it in response to demand for non-metadata. This
405 value acts as a ceiling to the amount of dnode metadata, and defaults to 0 which
406 indicates that a percent which is based on \fBzfs_arc_dnode_limit_percent\fR of
407 the ARC meta buffers that may be used for dnodes.
409 See also \fBzfs_arc_meta_prune\fR which serves a similar purpose but is used
410 when the amount of metadata in the ARC exceeds \fBzfs_arc_meta_limit\fR rather
411 than in response to overall demand for non-metadata.
414 Default value: \fB0\fR.
420 \fBzfs_arc_dnode_limit_percent\fR (ulong)
423 Percentage that can be consumed by dnodes of ARC meta buffers.
425 See also \fBzfs_arc_dnode_limit\fR which serves a similar purpose but has a
426 higher priority if set to nonzero value.
428 Default value: \fB10\fR%.
434 \fBzfs_arc_dnode_reduce_percent\fR (ulong)
437 Percentage of ARC dnodes to try to scan in response to demand for non-metadata
438 when the number of bytes consumed by dnodes exceeds \fBzfs_arc_dnode_limit\fR.
441 Default value: \fB10\fR% of the number of dnodes in the ARC.
447 \fBzfs_arc_average_blocksize\fR (int)
450 The ARC's buffer hash table is sized based on the assumption of an average
451 block size of \fBzfs_arc_average_blocksize\fR (default 8K). This works out
452 to roughly 1MB of hash table per 1GB of physical memory with 8-byte pointers.
453 For configurations with a known larger average block size this value can be
454 increased to reduce the memory footprint.
457 Default value: \fB8192\fR.
463 \fBzfs_arc_evict_batch_limit\fR (int)
466 Number ARC headers to evict per sub-list before proceeding to another sub-list.
467 This batch-style operation prevents entire sub-lists from being evicted at once
468 but comes at a cost of additional unlocking and locking.
470 Default value: \fB10\fR.
476 \fBzfs_arc_grow_retry\fR (int)
479 If set to a non zero value, it will replace the arc_grow_retry value with this value.
480 The arc_grow_retry value (default 5) is the number of seconds the ARC will wait before
481 trying to resume growth after a memory pressure event.
483 Default value: \fB0\fR.
489 \fBzfs_arc_lotsfree_percent\fR (int)
492 Throttle I/O when free system memory drops below this percentage of total
493 system memory. Setting this value to 0 will disable the throttle.
495 Default value: \fB10\fR%.
501 \fBzfs_arc_max\fR (ulong)
504 Max arc size of ARC in bytes. If set to 0 then it will consume 1/2 of system
505 RAM. This value must be at least 67108864 (64 megabytes).
507 This value can be changed dynamically with some caveats. It cannot be set back
508 to 0 while running and reducing it below the current ARC size will not cause
509 the ARC to shrink without memory pressure to induce shrinking.
511 Default value: \fB0\fR.
517 \fBzfs_arc_meta_adjust_restarts\fR (ulong)
520 The number of restart passes to make while scanning the ARC attempting
521 the free buffers in order to stay below the \fBzfs_arc_meta_limit\fR.
522 This value should not need to be tuned but is available to facilitate
523 performance analysis.
525 Default value: \fB4096\fR.
531 \fBzfs_arc_meta_limit\fR (ulong)
534 The maximum allowed size in bytes that meta data buffers are allowed to
535 consume in the ARC. When this limit is reached meta data buffers will
536 be reclaimed even if the overall arc_c_max has not been reached. This
537 value defaults to 0 which indicates that a percent which is based on
538 \fBzfs_arc_meta_limit_percent\fR of the ARC may be used for meta data.
540 This value my be changed dynamically except that it cannot be set back to 0
541 for a specific percent of the ARC; it must be set to an explicit value.
543 Default value: \fB0\fR.
549 \fBzfs_arc_meta_limit_percent\fR (ulong)
552 Percentage of ARC buffers that can be used for meta data.
554 See also \fBzfs_arc_meta_limit\fR which serves a similar purpose but has a
555 higher priority if set to nonzero value.
558 Default value: \fB75\fR%.
564 \fBzfs_arc_meta_min\fR (ulong)
567 The minimum allowed size in bytes that meta data buffers may consume in
568 the ARC. This value defaults to 0 which disables a floor on the amount
569 of the ARC devoted meta data.
571 Default value: \fB0\fR.
577 \fBzfs_arc_meta_prune\fR (int)
580 The number of dentries and inodes to be scanned looking for entries
581 which can be dropped. This may be required when the ARC reaches the
582 \fBzfs_arc_meta_limit\fR because dentries and inodes can pin buffers
583 in the ARC. Increasing this value will cause to dentry and inode caches
584 to be pruned more aggressively. Setting this value to 0 will disable
585 pruning the inode and dentry caches.
587 Default value: \fB10,000\fR.
593 \fBzfs_arc_meta_strategy\fR (int)
596 Define the strategy for ARC meta data buffer eviction (meta reclaim strategy).
597 A value of 0 (META_ONLY) will evict only the ARC meta data buffers.
598 A value of 1 (BALANCED) indicates that additional data buffers may be evicted if
599 that is required to in order to evict the required number of meta data buffers.
601 Default value: \fB1\fR.
607 \fBzfs_arc_min\fR (ulong)
610 Min arc size of ARC in bytes. If set to 0 then arc_c_min will default to
611 consuming the larger of 32M or 1/32 of total system memory.
613 Default value: \fB0\fR.
619 \fBzfs_arc_min_prefetch_ms\fR (int)
622 Minimum time prefetched blocks are locked in the ARC, specified in ms.
623 A value of \fB0\fR will default to 1000 ms.
625 Default value: \fB0\fR.
631 \fBzfs_arc_min_prescient_prefetch_ms\fR (int)
634 Minimum time "prescient prefetched" blocks are locked in the ARC, specified
635 in ms. These blocks are meant to be prefetched fairly aggresively ahead of
636 the code that may use them. A value of \fB0\fR will default to 6000 ms.
638 Default value: \fB0\fR.
644 \fBzfs_multilist_num_sublists\fR (int)
647 To allow more fine-grained locking, each ARC state contains a series
648 of lists for both data and meta data objects. Locking is performed at
649 the level of these "sub-lists". This parameters controls the number of
650 sub-lists per ARC state, and also applies to other uses of the
651 multilist data structure.
653 Default value: \fB4\fR or the number of online CPUs, whichever is greater
659 \fBzfs_arc_overflow_shift\fR (int)
662 The ARC size is considered to be overflowing if it exceeds the current
663 ARC target size (arc_c) by a threshold determined by this parameter.
664 The threshold is calculated as a fraction of arc_c using the formula
665 "arc_c >> \fBzfs_arc_overflow_shift\fR".
667 The default value of 8 causes the ARC to be considered to be overflowing
668 if it exceeds the target size by 1/256th (0.3%) of the target size.
670 When the ARC is overflowing, new buffer allocations are stalled until
671 the reclaim thread catches up and the overflow condition no longer exists.
673 Default value: \fB8\fR.
680 \fBzfs_arc_p_min_shift\fR (int)
683 If set to a non zero value, this will update arc_p_min_shift (default 4)
685 arc_p_min_shift is used to shift of arc_c for calculating both min and max
688 Default value: \fB0\fR.
694 \fBzfs_arc_p_dampener_disable\fR (int)
697 Disable arc_p adapt dampener
699 Use \fB1\fR for yes (default) and \fB0\fR to disable.
705 \fBzfs_arc_shrink_shift\fR (int)
708 If set to a non zero value, this will update arc_shrink_shift (default 7)
711 Default value: \fB0\fR.
717 \fBzfs_arc_pc_percent\fR (uint)
720 Percent of pagecache to reclaim arc to
722 This tunable allows ZFS arc to play more nicely with the kernel's LRU
723 pagecache. It can guarantee that the arc size won't collapse under scanning
724 pressure on the pagecache, yet still allows arc to be reclaimed down to
725 zfs_arc_min if necessary. This value is specified as percent of pagecache
726 size (as measured by NR_FILE_PAGES) where that percent may exceed 100. This
727 only operates during memory pressure/reclaim.
729 Default value: \fB0\fR% (disabled).
735 \fBzfs_arc_sys_free\fR (ulong)
738 The target number of bytes the ARC should leave as free memory on the system.
739 Defaults to the larger of 1/64 of physical memory or 512K. Setting this
740 option to a non-zero value will override the default.
742 Default value: \fB0\fR.
748 \fBzfs_autoimport_disable\fR (int)
751 Disable pool import at module load by ignoring the cache file (typically \fB/etc/zfs/zpool.cache\fR).
753 Use \fB1\fR for yes (default) and \fB0\fR for no.
759 \fBzfs_commit_timeout_pct\fR (int)
762 This controls the amount of time that a ZIL block (lwb) will remain "open"
763 when it isn't "full", and it has a thread waiting for it to be committed to
764 stable storage. The timeout is scaled based on a percentage of the last lwb
765 latency to avoid significantly impacting the latency of each individual
766 transaction record (itx).
768 Default value: \fB5\fR%.
774 \fBzfs_dbgmsg_enable\fR (int)
777 Internally ZFS keeps a small log to facilitate debugging. By default the log
778 is disabled, to enable it set this option to 1. The contents of the log can
779 be accessed by reading the /proc/spl/kstat/zfs/dbgmsg file. Writing 0 to
780 this proc file clears the log.
782 Default value: \fB0\fR.
788 \fBzfs_dbgmsg_maxsize\fR (int)
791 The maximum size in bytes of the internal ZFS debug log.
793 Default value: \fB4M\fR.
799 \fBzfs_dbuf_state_index\fR (int)
802 This feature is currently unused. It is normally used for controlling what
803 reporting is available under /proc/spl/kstat/zfs.
805 Default value: \fB0\fR.
811 \fBzfs_deadman_enabled\fR (int)
814 When a pool sync operation takes longer than \fBzfs_deadman_synctime_ms\fR
815 milliseconds, or when an individual I/O takes longer than
816 \fBzfs_deadman_ziotime_ms\fR milliseconds, then the operation is considered to
817 be "hung". If \fBzfs_deadman_enabled\fR is set then the deadman behavior is
818 invoked as described by the \fBzfs_deadman_failmode\fR module option.
819 By default the deadman is enabled and configured to \fBwait\fR which results
820 in "hung" I/Os only being logged. The deadman is automatically disabled
821 when a pool gets suspended.
823 Default value: \fB1\fR.
829 \fBzfs_deadman_failmode\fR (charp)
832 Controls the failure behavior when the deadman detects a "hung" I/O. Valid
833 values are \fBwait\fR, \fBcontinue\fR, and \fBpanic\fR.
835 \fBwait\fR - Wait for a "hung" I/O to complete. For each "hung" I/O a
836 "deadman" event will be posted describing that I/O.
838 \fBcontinue\fR - Attempt to recover from a "hung" I/O by re-dispatching it
839 to the I/O pipeline if possible.
841 \fBpanic\fR - Panic the system. This can be used to facilitate an automatic
842 fail-over to a properly configured fail-over partner.
844 Default value: \fBwait\fR.
850 \fBzfs_deadman_checktime_ms\fR (int)
853 Check time in milliseconds. This defines the frequency at which we check
854 for hung I/O and potentially invoke the \fBzfs_deadman_failmode\fR behavior.
856 Default value: \fB60,000\fR.
862 \fBzfs_deadman_synctime_ms\fR (ulong)
865 Interval in milliseconds after which the deadman is triggered and also
866 the interval after which a pool sync operation is considered to be "hung".
867 Once this limit is exceeded the deadman will be invoked every
868 \fBzfs_deadman_checktime_ms\fR milliseconds until the pool sync completes.
870 Default value: \fB600,000\fR.
876 \fBzfs_deadman_ziotime_ms\fR (ulong)
879 Interval in milliseconds after which the deadman is triggered and an
880 individual IO operation is considered to be "hung". As long as the I/O
881 remains "hung" the deadman will be invoked every \fBzfs_deadman_checktime_ms\fR
882 milliseconds until the I/O completes.
884 Default value: \fB300,000\fR.
890 \fBzfs_dedup_prefetch\fR (int)
893 Enable prefetching dedup-ed blks
895 Use \fB1\fR for yes and \fB0\fR to disable (default).
901 \fBzfs_delay_min_dirty_percent\fR (int)
904 Start to delay each transaction once there is this amount of dirty data,
905 expressed as a percentage of \fBzfs_dirty_data_max\fR.
906 This value should be >= zfs_vdev_async_write_active_max_dirty_percent.
907 See the section "ZFS TRANSACTION DELAY".
909 Default value: \fB60\fR%.
915 \fBzfs_delay_scale\fR (int)
918 This controls how quickly the transaction delay approaches infinity.
919 Larger values cause longer delays for a given amount of dirty data.
921 For the smoothest delay, this value should be about 1 billion divided
922 by the maximum number of operations per second. This will smoothly
923 handle between 10x and 1/10th this number.
925 See the section "ZFS TRANSACTION DELAY".
927 Note: \fBzfs_delay_scale\fR * \fBzfs_dirty_data_max\fR must be < 2^64.
929 Default value: \fB500,000\fR.
935 \fBzfs_delete_blocks\fR (ulong)
938 This is the used to define a large file for the purposes of delete. Files
939 containing more than \fBzfs_delete_blocks\fR will be deleted asynchronously
940 while smaller files are deleted synchronously. Decreasing this value will
941 reduce the time spent in an unlink(2) system call at the expense of a longer
942 delay before the freed space is available.
944 Default value: \fB20,480\fR.
950 \fBzfs_dirty_data_max\fR (int)
953 Determines the dirty space limit in bytes. Once this limit is exceeded, new
954 writes are halted until space frees up. This parameter takes precedence
955 over \fBzfs_dirty_data_max_percent\fR.
956 See the section "ZFS TRANSACTION DELAY".
958 Default value: \fB10\fR% of physical RAM, capped at \fBzfs_dirty_data_max_max\fR.
964 \fBzfs_dirty_data_max_max\fR (int)
967 Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed in bytes.
968 This limit is only enforced at module load time, and will be ignored if
969 \fBzfs_dirty_data_max\fR is later changed. This parameter takes
970 precedence over \fBzfs_dirty_data_max_max_percent\fR. See the section
971 "ZFS TRANSACTION DELAY".
973 Default value: \fB25\fR% of physical RAM.
979 \fBzfs_dirty_data_max_max_percent\fR (int)
982 Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed as a
983 percentage of physical RAM. This limit is only enforced at module load
984 time, and will be ignored if \fBzfs_dirty_data_max\fR is later changed.
985 The parameter \fBzfs_dirty_data_max_max\fR takes precedence over this
986 one. See the section "ZFS TRANSACTION DELAY".
988 Default value: \fB25\fR%.
994 \fBzfs_dirty_data_max_percent\fR (int)
997 Determines the dirty space limit, expressed as a percentage of all
998 memory. Once this limit is exceeded, new writes are halted until space frees
999 up. The parameter \fBzfs_dirty_data_max\fR takes precedence over this
1000 one. See the section "ZFS TRANSACTION DELAY".
1002 Default value: \fB10\fR%, subject to \fBzfs_dirty_data_max_max\fR.
1008 \fBzfs_dirty_data_sync\fR (int)
1011 Start syncing out a transaction group if there is at least this much dirty data.
1013 Default value: \fB67,108,864\fR.
1019 \fBzfs_fletcher_4_impl\fR (string)
1022 Select a fletcher 4 implementation.
1024 Supported selectors are: \fBfastest\fR, \fBscalar\fR, \fBsse2\fR, \fBssse3\fR,
1025 \fBavx2\fR, \fBavx512f\fR, and \fBaarch64_neon\fR.
1026 All of the selectors except \fBfastest\fR and \fBscalar\fR require instruction
1027 set extensions to be available and will only appear if ZFS detects that they are
1028 present at runtime. If multiple implementations of fletcher 4 are available,
1029 the \fBfastest\fR will be chosen using a micro benchmark. Selecting \fBscalar\fR
1030 results in the original, CPU based calculation, being used. Selecting any option
1031 other than \fBfastest\fR and \fBscalar\fR results in vector instructions from
1032 the respective CPU instruction set being used.
1034 Default value: \fBfastest\fR.
1040 \fBzfs_free_bpobj_enabled\fR (int)
1043 Enable/disable the processing of the free_bpobj object.
1045 Default value: \fB1\fR.
1051 \fBzfs_free_max_blocks\fR (ulong)
1054 Maximum number of blocks freed in a single txg.
1056 Default value: \fB100,000\fR.
1062 \fBzfs_vdev_async_read_max_active\fR (int)
1065 Maximum asynchronous read I/Os active to each device.
1066 See the section "ZFS I/O SCHEDULER".
1068 Default value: \fB3\fR.
1074 \fBzfs_vdev_async_read_min_active\fR (int)
1077 Minimum asynchronous read I/Os active to each device.
1078 See the section "ZFS I/O SCHEDULER".
1080 Default value: \fB1\fR.
1086 \fBzfs_vdev_async_write_active_max_dirty_percent\fR (int)
1089 When the pool has more than
1090 \fBzfs_vdev_async_write_active_max_dirty_percent\fR dirty data, use
1091 \fBzfs_vdev_async_write_max_active\fR to limit active async writes. If
1092 the dirty data is between min and max, the active I/O limit is linearly
1093 interpolated. See the section "ZFS I/O SCHEDULER".
1095 Default value: \fB60\fR%.
1101 \fBzfs_vdev_async_write_active_min_dirty_percent\fR (int)
1104 When the pool has less than
1105 \fBzfs_vdev_async_write_active_min_dirty_percent\fR dirty data, use
1106 \fBzfs_vdev_async_write_min_active\fR to limit active async writes. If
1107 the dirty data is between min and max, the active I/O limit is linearly
1108 interpolated. See the section "ZFS I/O SCHEDULER".
1110 Default value: \fB30\fR%.
1116 \fBzfs_vdev_async_write_max_active\fR (int)
1119 Maximum asynchronous write I/Os active to each device.
1120 See the section "ZFS I/O SCHEDULER".
1122 Default value: \fB10\fR.
1128 \fBzfs_vdev_async_write_min_active\fR (int)
1131 Minimum asynchronous write I/Os active to each device.
1132 See the section "ZFS I/O SCHEDULER".
1134 Lower values are associated with better latency on rotational media but poorer
1135 resilver performance. The default value of 2 was chosen as a compromise. A
1136 value of 3 has been shown to improve resilver performance further at a cost of
1137 further increasing latency.
1139 Default value: \fB2\fR.
1145 \fBzfs_vdev_max_active\fR (int)
1148 The maximum number of I/Os active to each device. Ideally, this will be >=
1149 the sum of each queue's max_active. It must be at least the sum of each
1150 queue's min_active. See the section "ZFS I/O SCHEDULER".
1152 Default value: \fB1,000\fR.
1158 \fBzfs_vdev_scrub_max_active\fR (int)
1161 Maximum scrub I/Os active to each device.
1162 See the section "ZFS I/O SCHEDULER".
1164 Default value: \fB2\fR.
1170 \fBzfs_vdev_scrub_min_active\fR (int)
1173 Minimum scrub I/Os active to each device.
1174 See the section "ZFS I/O SCHEDULER".
1176 Default value: \fB1\fR.
1182 \fBzfs_vdev_sync_read_max_active\fR (int)
1185 Maximum synchronous read I/Os active to each device.
1186 See the section "ZFS I/O SCHEDULER".
1188 Default value: \fB10\fR.
1194 \fBzfs_vdev_sync_read_min_active\fR (int)
1197 Minimum synchronous read I/Os active to each device.
1198 See the section "ZFS I/O SCHEDULER".
1200 Default value: \fB10\fR.
1206 \fBzfs_vdev_sync_write_max_active\fR (int)
1209 Maximum synchronous write I/Os active to each device.
1210 See the section "ZFS I/O SCHEDULER".
1212 Default value: \fB10\fR.
1218 \fBzfs_vdev_sync_write_min_active\fR (int)
1221 Minimum synchronous write I/Os active to each device.
1222 See the section "ZFS I/O SCHEDULER".
1224 Default value: \fB10\fR.
1230 \fBzfs_vdev_queue_depth_pct\fR (int)
1233 Maximum number of queued allocations per top-level vdev expressed as
1234 a percentage of \fBzfs_vdev_async_write_max_active\fR which allows the
1235 system to detect devices that are more capable of handling allocations
1236 and to allocate more blocks to those devices. It allows for dynamic
1237 allocation distribution when devices are imbalanced as fuller devices
1238 will tend to be slower than empty devices.
1240 See also \fBzio_dva_throttle_enabled\fR.
1242 Default value: \fB1000\fR%.
1248 \fBzfs_expire_snapshot\fR (int)
1251 Seconds to expire .zfs/snapshot
1253 Default value: \fB300\fR.
1259 \fBzfs_admin_snapshot\fR (int)
1262 Allow the creation, removal, or renaming of entries in the .zfs/snapshot
1263 directory to cause the creation, destruction, or renaming of snapshots.
1264 When enabled this functionality works both locally and over NFS exports
1265 which have the 'no_root_squash' option set. This functionality is disabled
1268 Use \fB1\fR for yes and \fB0\fR for no (default).
1274 \fBzfs_flags\fR (int)
1277 Set additional debugging flags. The following flags may be bitwise-or'd
1289 Enable dprintf entries in the debug log.
1291 2 ZFS_DEBUG_DBUF_VERIFY *
1292 Enable extra dbuf verifications.
1294 4 ZFS_DEBUG_DNODE_VERIFY *
1295 Enable extra dnode verifications.
1297 8 ZFS_DEBUG_SNAPNAMES
1298 Enable snapshot name verification.
1301 Check for illegally modified ARC buffers.
1304 Enable spa_dbgmsg entries in the debug log.
1306 64 ZFS_DEBUG_ZIO_FREE
1307 Enable verification of block frees.
1309 128 ZFS_DEBUG_HISTOGRAM_VERIFY
1310 Enable extra spacemap histogram verifications.
1312 256 ZFS_DEBUG_METASLAB_VERIFY
1313 Verify space accounting on disk matches in-core range_trees.
1315 512 ZFS_DEBUG_SET_ERROR
1316 Enable SET_ERROR and dprintf entries in the debug log.
1319 * Requires debug build.
1321 Default value: \fB0\fR.
1327 \fBzfs_free_leak_on_eio\fR (int)
1330 If destroy encounters an EIO while reading metadata (e.g. indirect
1331 blocks), space referenced by the missing metadata can not be freed.
1332 Normally this causes the background destroy to become "stalled", as
1333 it is unable to make forward progress. While in this stalled state,
1334 all remaining space to free from the error-encountering filesystem is
1335 "temporarily leaked". Set this flag to cause it to ignore the EIO,
1336 permanently leak the space from indirect blocks that can not be read,
1337 and continue to free everything else that it can.
1339 The default, "stalling" behavior is useful if the storage partially
1340 fails (i.e. some but not all i/os fail), and then later recovers. In
1341 this case, we will be able to continue pool operations while it is
1342 partially failed, and when it recovers, we can continue to free the
1343 space, with no leaks. However, note that this case is actually
1346 Typically pools either (a) fail completely (but perhaps temporarily,
1347 e.g. a top-level vdev going offline), or (b) have localized,
1348 permanent errors (e.g. disk returns the wrong data due to bit flip or
1349 firmware bug). In case (a), this setting does not matter because the
1350 pool will be suspended and the sync thread will not be able to make
1351 forward progress regardless. In case (b), because the error is
1352 permanent, the best we can do is leak the minimum amount of space,
1353 which is what setting this flag will do. Therefore, it is reasonable
1354 for this flag to normally be set, but we chose the more conservative
1355 approach of not setting it, so that there is no possibility of
1356 leaking space in the "partial temporary" failure case.
1358 Default value: \fB0\fR.
1364 \fBzfs_free_min_time_ms\fR (int)
1367 During a \fBzfs destroy\fR operation using \fBfeature@async_destroy\fR a minimum
1368 of this much time will be spent working on freeing blocks per txg.
1370 Default value: \fB1,000\fR.
1376 \fBzfs_immediate_write_sz\fR (long)
1379 Largest data block to write to zil. Larger blocks will be treated as if the
1380 dataset being written to had the property setting \fBlogbias=throughput\fR.
1382 Default value: \fB32,768\fR.
1388 \fBzfs_max_recordsize\fR (int)
1391 We currently support block sizes from 512 bytes to 16MB. The benefits of
1392 larger blocks, and thus larger IO, need to be weighed against the cost of
1393 COWing a giant block to modify one byte. Additionally, very large blocks
1394 can have an impact on i/o latency, and also potentially on the memory
1395 allocator. Therefore, we do not allow the recordsize to be set larger than
1396 zfs_max_recordsize (default 1MB). Larger blocks can be created by changing
1397 this tunable, and pools with larger blocks can always be imported and used,
1398 regardless of this setting.
1400 Default value: \fB1,048,576\fR.
1406 \fBzfs_mdcomp_disable\fR (int)
1409 Disable meta data compression
1411 Use \fB1\fR for yes and \fB0\fR for no (default).
1417 \fBzfs_metaslab_fragmentation_threshold\fR (int)
1420 Allow metaslabs to keep their active state as long as their fragmentation
1421 percentage is less than or equal to this value. An active metaslab that
1422 exceeds this threshold will no longer keep its active status allowing
1423 better metaslabs to be selected.
1425 Default value: \fB70\fR.
1431 \fBzfs_mg_fragmentation_threshold\fR (int)
1434 Metaslab groups are considered eligible for allocations if their
1435 fragmentation metric (measured as a percentage) is less than or equal to
1436 this value. If a metaslab group exceeds this threshold then it will be
1437 skipped unless all metaslab groups within the metaslab class have also
1438 crossed this threshold.
1440 Default value: \fB85\fR.
1446 \fBzfs_mg_noalloc_threshold\fR (int)
1449 Defines a threshold at which metaslab groups should be eligible for
1450 allocations. The value is expressed as a percentage of free space
1451 beyond which a metaslab group is always eligible for allocations.
1452 If a metaslab group's free space is less than or equal to the
1453 threshold, the allocator will avoid allocating to that group
1454 unless all groups in the pool have reached the threshold. Once all
1455 groups have reached the threshold, all groups are allowed to accept
1456 allocations. The default value of 0 disables the feature and causes
1457 all metaslab groups to be eligible for allocations.
1459 This parameter allows one to deal with pools having heavily imbalanced
1460 vdevs such as would be the case when a new vdev has been added.
1461 Setting the threshold to a non-zero percentage will stop allocations
1462 from being made to vdevs that aren't filled to the specified percentage
1463 and allow lesser filled vdevs to acquire more allocations than they
1464 otherwise would under the old \fBzfs_mg_alloc_failures\fR facility.
1466 Default value: \fB0\fR.
1472 \fBzfs_multihost_history\fR (int)
1475 Historical statistics for the last N multihost updates will be available in
1476 \fB/proc/spl/kstat/zfs/<pool>/multihost\fR
1478 Default value: \fB0\fR.
1484 \fBzfs_multihost_interval\fR (ulong)
1487 Used to control the frequency of multihost writes which are performed when the
1488 \fBmultihost\fR pool property is on. This is one factor used to determine
1489 the length of the activity check during import.
1491 The multihost write period is \fBzfs_multihost_interval / leaf-vdevs\fR milliseconds.
1492 This means that on average a multihost write will be issued for each leaf vdev every
1493 \fBzfs_multihost_interval\fR milliseconds. In practice, the observed period can
1494 vary with the I/O load and this observed value is the delay which is stored in
1497 On import the activity check waits a minimum amount of time determined by
1498 \fBzfs_multihost_interval * zfs_multihost_import_intervals\fR. The activity
1499 check time may be further extended if the value of mmp delay found in the best
1500 uberblock indicates actual multihost updates happened at longer intervals than
1501 \fBzfs_multihost_interval\fR. A minimum value of \fB100ms\fR is enforced.
1503 Default value: \fB1000\fR.
1509 \fBzfs_multihost_import_intervals\fR (uint)
1512 Used to control the duration of the activity test on import. Smaller values of
1513 \fBzfs_multihost_import_intervals\fR will reduce the import time but increase
1514 the risk of failing to detect an active pool. The total activity check time is
1515 never allowed to drop below one second. A value of 0 is ignored and treated as
1518 Default value: \fB10\fR.
1524 \fBzfs_multihost_fail_intervals\fR (uint)
1527 Controls the behavior of the pool when multihost write failures are detected.
1529 When \fBzfs_multihost_fail_intervals = 0\fR then multihost write failures are ignored.
1530 The failures will still be reported to the ZED which depending on its
1531 configuration may take action such as suspending the pool or offlining a device.
1533 When \fBzfs_multihost_fail_intervals > 0\fR then sequential multihost write failures
1534 will cause the pool to be suspended. This occurs when
1535 \fBzfs_multihost_fail_intervals * zfs_multihost_interval\fR milliseconds have
1536 passed since the last successful multihost write. This guarantees the activity test
1537 will see multihost writes if the pool is imported.
1539 Default value: \fB5\fR.
1545 \fBzfs_no_scrub_io\fR (int)
1548 Set for no scrub I/O. This results in scrubs not actually scrubbing data and
1549 simply doing a metadata crawl of the pool instead.
1551 Use \fB1\fR for yes and \fB0\fR for no (default).
1557 \fBzfs_no_scrub_prefetch\fR (int)
1560 Set to disable block prefetching for scrubs.
1562 Use \fB1\fR for yes and \fB0\fR for no (default).
1568 \fBzfs_nocacheflush\fR (int)
1571 Disable cache flush operations on disks when writing. Beware, this may cause
1572 corruption if disks re-order writes.
1574 Use \fB1\fR for yes and \fB0\fR for no (default).
1580 \fBzfs_nopwrite_enabled\fR (int)
1585 Use \fB1\fR for yes (default) and \fB0\fR to disable.
1591 \fBzfs_dmu_offset_next_sync\fR (int)
1594 Enable forcing txg sync to find holes. When enabled forces ZFS to act
1595 like prior versions when SEEK_HOLE or SEEK_DATA flags are used, which
1596 when a dnode is dirty causes txg's to be synced so that this data can be
1599 Use \fB1\fR for yes and \fB0\fR to disable (default).
1605 \fBzfs_pd_bytes_max\fR (int)
1608 The number of bytes which should be prefetched during a pool traversal
1609 (eg: \fBzfs send\fR or other data crawling operations)
1611 Default value: \fB52,428,800\fR.
1617 \fBzfs_per_txg_dirty_frees_percent \fR (ulong)
1620 Tunable to control percentage of dirtied blocks from frees in one TXG.
1621 After this threshold is crossed, additional dirty blocks from frees
1622 wait until the next TXG.
1623 A value of zero will disable this throttle.
1625 Default value: \fB30\fR and \fB0\fR to disable.
1633 \fBzfs_prefetch_disable\fR (int)
1636 This tunable disables predictive prefetch. Note that it leaves "prescient"
1637 prefetch (e.g. prefetch for zfs send) intact. Unlike predictive prefetch,
1638 prescient prefetch never issues i/os that end up not being needed, so it
1639 can't hurt performance.
1641 Use \fB1\fR for yes and \fB0\fR for no (default).
1647 \fBzfs_read_chunk_size\fR (long)
1650 Bytes to read per chunk
1652 Default value: \fB1,048,576\fR.
1658 \fBzfs_read_history\fR (int)
1661 Historical statistics for the last N reads will be available in
1662 \fB/proc/spl/kstat/zfs/<pool>/reads\fR
1664 Default value: \fB0\fR (no data is kept).
1670 \fBzfs_read_history_hits\fR (int)
1673 Include cache hits in read history
1675 Use \fB1\fR for yes and \fB0\fR for no (default).
1681 \fBzfs_recover\fR (int)
1684 Set to attempt to recover from fatal errors. This should only be used as a
1685 last resort, as it typically results in leaked space, or worse.
1687 Use \fB1\fR for yes and \fB0\fR for no (default).
1693 \fBzfs_resilver_min_time_ms\fR (int)
1696 Resilvers are processed by the sync thread. While resilvering it will spend
1697 at least this much time working on a resilver between txg flushes.
1699 Default value: \fB3,000\fR.
1705 \fBzfs_scrub_min_time_ms\fR (int)
1708 Scrubs are processed by the sync thread. While scrubbing it will spend
1709 at least this much time working on a scrub between txg flushes.
1711 Default value: \fB1,000\fR.
1717 \fBzfs_scan_checkpoint_intval\fR (int)
1720 To preserve progress across reboots the sequential scan algorithm periodically
1721 needs to stop metadata scanning and issue all the verifications I/Os to disk.
1722 The frequency of this flushing is determined by the
1723 \fBfBzfs_scan_checkpoint_intval\fR tunable.
1725 Default value: \fB7200\fR seconds (every 2 hours).
1731 \fBzfs_scan_fill_weight\fR (int)
1734 This tunable affects how scrub and resilver I/O segments are ordered. A higher
1735 number indicates that we care more about how filled in a segment is, while a
1736 lower number indicates we care more about the size of the extent without
1737 considering the gaps within a segment. This value is only tunable upon module
1738 insertion. Changing the value afterwards will have no affect on scrub or
1739 resilver performance.
1741 Default value: \fB3\fR.
1747 \fBzfs_scan_issue_strategy\fR (int)
1750 Determines the order that data will be verified while scrubbing or resilvering.
1751 If set to \fB1\fR, data will be verified as sequentially as possible, given the
1752 amount of memory reserved for scrubbing (see \fBzfs_scan_mem_lim_fact\fR). This
1753 may improve scrub performance if the pool's data is very fragmented. If set to
1754 \fB2\fR, the largest mostly-contiguous chunk of found data will be verified
1755 first. By deferring scrubbing of small segments, we may later find adjacent data
1756 to coalesce and increase the segment size. If set to \fB0\fR, zfs will use
1757 strategy \fB1\fR during normal verification and strategy \fB2\fR while taking a
1760 Default value: \fB0\fR.
1766 \fBzfs_scan_legacy\fR (int)
1769 A value of 0 indicates that scrubs and resilvers will gather metadata in
1770 memory before issuing sequential I/O. A value of 1 indicates that the legacy
1771 algorithm will be used where I/O is initiated as soon as it is discovered.
1772 Changing this value to 0 will not affect scrubs or resilvers that are already
1775 Default value: \fB0\fR.
1781 \fBzfs_scan_max_ext_gap\fR (int)
1784 Indicates the largest gap in bytes between scrub / resilver I/Os that will still
1785 be considered sequential for sorting purposes. Changing this value will not
1786 affect scrubs or resilvers that are already in progress.
1788 Default value: \fB2097152 (2 MB)\fR.
1794 \fBzfs_scan_mem_lim_fact\fR (int)
1797 Maximum fraction of RAM used for I/O sorting by sequential scan algorithm.
1798 This tunable determines the hard limit for I/O sorting memory usage.
1799 When the hard limit is reached we stop scanning metadata and start issuing
1800 data verification I/O. This is done until we get below the soft limit.
1802 Default value: \fB20\fR which is 5% of RAM (1/20).
1808 \fBzfs_scan_mem_lim_soft_fact\fR (int)
1811 The fraction of the hard limit used to determined the soft limit for I/O sorting
1812 by the sequential scan algorithm. When we cross this limit from bellow no action
1813 is taken. When we cross this limit from above it is because we are issuing
1814 verification I/O. In this case (unless the metadata scan is done) we stop
1815 issuing verification I/O and start scanning metadata again until we get to the
1818 Default value: \fB20\fR which is 5% of the hard limit (1/20).
1824 \fBzfs_scan_vdev_limit\fR (int)
1827 Maximum amount of data that can be concurrently issued at once for scrubs and
1828 resilvers per leaf device, given in bytes.
1830 Default value: \fB41943040\fR.
1836 \fBzfs_send_corrupt_data\fR (int)
1839 Allow sending of corrupt data (ignore read/checksum errors when sending data)
1841 Use \fB1\fR for yes and \fB0\fR for no (default).
1847 \fBzfs_sync_pass_deferred_free\fR (int)
1850 Flushing of data to disk is done in passes. Defer frees starting in this pass
1852 Default value: \fB2\fR.
1858 \fBzfs_sync_pass_dont_compress\fR (int)
1861 Don't compress starting in this pass
1863 Default value: \fB5\fR.
1869 \fBzfs_sync_pass_rewrite\fR (int)
1872 Rewrite new block pointers starting in this pass
1874 Default value: \fB2\fR.
1880 \fBzfs_sync_taskq_batch_pct\fR (int)
1883 This controls the number of threads used by the dp_sync_taskq. The default
1884 value of 75% will create a maximum of one thread per cpu.
1886 Default value: \fB75\fR%.
1892 \fBzfs_txg_history\fR (int)
1895 Historical statistics for the last N txgs will be available in
1896 \fB/proc/spl/kstat/zfs/<pool>/txgs\fR
1898 Default value: \fB0\fR.
1904 \fBzfs_txg_timeout\fR (int)
1907 Flush dirty data to disk at least every N seconds (maximum txg duration)
1909 Default value: \fB5\fR.
1915 \fBzfs_vdev_aggregation_limit\fR (int)
1918 Max vdev I/O aggregation size
1920 Default value: \fB131,072\fR.
1926 \fBzfs_vdev_cache_bshift\fR (int)
1929 Shift size to inflate reads too
1931 Default value: \fB16\fR (effectively 65536).
1937 \fBzfs_vdev_cache_max\fR (int)
1940 Inflate reads smaller than this value to meet the \fBzfs_vdev_cache_bshift\fR
1943 Default value: \fB16384\fR.
1949 \fBzfs_vdev_cache_size\fR (int)
1952 Total size of the per-disk cache in bytes.
1954 Currently this feature is disabled as it has been found to not be helpful
1955 for performance and in some cases harmful.
1957 Default value: \fB0\fR.
1963 \fBzfs_vdev_mirror_rotating_inc\fR (int)
1966 A number by which the balancing algorithm increments the load calculation for
1967 the purpose of selecting the least busy mirror member when an I/O immediately
1968 follows its predecessor on rotational vdevs for the purpose of making decisions
1971 Default value: \fB0\fR.
1977 \fBzfs_vdev_mirror_rotating_seek_inc\fR (int)
1980 A number by which the balancing algorithm increments the load calculation for
1981 the purpose of selecting the least busy mirror member when an I/O lacks
1982 locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within
1983 this that are not immediately following the previous I/O are incremented by
1986 Default value: \fB5\fR.
1992 \fBzfs_vdev_mirror_rotating_seek_offset\fR (int)
1995 The maximum distance for the last queued I/O in which the balancing algorithm
1996 considers an I/O to have locality.
1997 See the section "ZFS I/O SCHEDULER".
1999 Default value: \fB1048576\fR.
2005 \fBzfs_vdev_mirror_non_rotating_inc\fR (int)
2008 A number by which the balancing algorithm increments the load calculation for
2009 the purpose of selecting the least busy mirror member on non-rotational vdevs
2010 when I/Os do not immediately follow one another.
2012 Default value: \fB0\fR.
2018 \fBzfs_vdev_mirror_non_rotating_seek_inc\fR (int)
2021 A number by which the balancing algorithm increments the load calculation for
2022 the purpose of selecting the least busy mirror member when an I/O lacks
2023 locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within
2024 this that are not immediately following the previous I/O are incremented by
2027 Default value: \fB1\fR.
2033 \fBzfs_vdev_read_gap_limit\fR (int)
2036 Aggregate read I/O operations if the gap on-disk between them is within this
2039 Default value: \fB32,768\fR.
2045 \fBzfs_vdev_scheduler\fR (charp)
2048 Set the Linux I/O scheduler on whole disk vdevs to this scheduler. Valid options
2049 are noop, cfq, bfq & deadline
2051 Default value: \fBnoop\fR.
2057 \fBzfs_vdev_write_gap_limit\fR (int)
2060 Aggregate write I/O over gap
2062 Default value: \fB4,096\fR.
2068 \fBzfs_vdev_raidz_impl\fR (string)
2071 Parameter for selecting raidz parity implementation to use.
2073 Options marked (always) below may be selected on module load as they are
2074 supported on all systems.
2075 The remaining options may only be set after the module is loaded, as they
2076 are available only if the implementations are compiled in and supported
2077 on the running system.
2079 Once the module is loaded, the content of
2080 /sys/module/zfs/parameters/zfs_vdev_raidz_impl will show available options
2081 with the currently selected one enclosed in [].
2082 Possible options are:
2083 fastest - (always) implementation selected using built-in benchmark
2084 original - (always) original raidz implementation
2085 scalar - (always) scalar raidz implementation
2086 sse2 - implementation using SSE2 instruction set (64bit x86 only)
2087 ssse3 - implementation using SSSE3 instruction set (64bit x86 only)
2088 avx2 - implementation using AVX2 instruction set (64bit x86 only)
2089 avx512f - implementation using AVX512F instruction set (64bit x86 only)
2090 avx512bw - implementation using AVX512F & AVX512BW instruction sets (64bit x86 only)
2091 aarch64_neon - implementation using NEON (Aarch64/64 bit ARMv8 only)
2092 aarch64_neonx2 - implementation using NEON with more unrolling (Aarch64/64 bit ARMv8 only)
2094 Default value: \fBfastest\fR.
2100 \fBzfs_zevent_cols\fR (int)
2103 When zevents are logged to the console use this as the word wrap width.
2105 Default value: \fB80\fR.
2111 \fBzfs_zevent_console\fR (int)
2114 Log events to the console
2116 Use \fB1\fR for yes and \fB0\fR for no (default).
2122 \fBzfs_zevent_len_max\fR (int)
2125 Max event queue length. A value of 0 will result in a calculated value which
2126 increases with the number of CPUs in the system (minimum 64 events). Events
2127 in the queue can be viewed with the \fBzpool events\fR command.
2129 Default value: \fB0\fR.
2135 \fBzfs_zil_clean_taskq_maxalloc\fR (int)
2138 The maximum number of taskq entries that are allowed to be cached. When this
2139 limit is exceeded transaction records (itxs) will be cleaned synchronously.
2141 Default value: \fB1048576\fR.
2147 \fBzfs_zil_clean_taskq_minalloc\fR (int)
2150 The number of taskq entries that are pre-populated when the taskq is first
2151 created and are immediately available for use.
2153 Default value: \fB1024\fR.
2159 \fBzfs_zil_clean_taskq_nthr_pct\fR (int)
2162 This controls the number of threads used by the dp_zil_clean_taskq. The default
2163 value of 100% will create a maximum of one thread per cpu.
2165 Default value: \fB100\fR%.
2171 \fBzil_replay_disable\fR (int)
2174 Disable intent logging replay. Can be disabled for recovery from corrupted
2177 Use \fB1\fR for yes and \fB0\fR for no (default).
2183 \fBzil_slog_bulk\fR (ulong)
2186 Limit SLOG write size per commit executed with synchronous priority.
2187 Any writes above that will be executed with lower (asynchronous) priority
2188 to limit potential SLOG device abuse by single active ZIL writer.
2190 Default value: \fB786,432\fR.
2196 \fBzio_delay_max\fR (int)
2199 A zevent will be logged if a ZIO operation takes more than N milliseconds to
2200 complete. Note that this is only a logging facility, not a timeout on
2203 Default value: \fB30,000\fR.
2209 \fBzio_dva_throttle_enabled\fR (int)
2212 Throttle block allocations in the ZIO pipeline. This allows for
2213 dynamic allocation distribution when devices are imbalanced.
2214 When enabled, the maximum number of pending allocations per top-level vdev
2215 is limited by \fBzfs_vdev_queue_depth_pct\fR.
2217 Default value: \fB1\fR.
2223 \fBzio_requeue_io_start_cut_in_line\fR (int)
2226 Prioritize requeued I/O
2228 Default value: \fB0\fR.
2234 \fBzio_taskq_batch_pct\fR (uint)
2237 Percentage of online CPUs (or CPU cores, etc) which will run a worker thread
2238 for IO. These workers are responsible for IO work such as compression and
2239 checksum calculations. Fractional number of CPUs will be rounded down.
2241 The default value of 75 was chosen to avoid using all CPUs which can result in
2242 latency issues and inconsistent application performance, especially when high
2243 compression is enabled.
2245 Default value: \fB75\fR.
2251 \fBzvol_inhibit_dev\fR (uint)
2254 Do not create zvol device nodes. This may slightly improve startup time on
2255 systems with a very large number of zvols.
2257 Use \fB1\fR for yes and \fB0\fR for no (default).
2263 \fBzvol_major\fR (uint)
2266 Major number for zvol block devices
2268 Default value: \fB230\fR.
2274 \fBzvol_max_discard_blocks\fR (ulong)
2277 Discard (aka TRIM) operations done on zvols will be done in batches of this
2278 many blocks, where block size is determined by the \fBvolblocksize\fR property
2281 Default value: \fB16,384\fR.
2287 \fBzvol_prefetch_bytes\fR (uint)
2290 When adding a zvol to the system prefetch \fBzvol_prefetch_bytes\fR
2291 from the start and end of the volume. Prefetching these regions
2292 of the volume is desirable because they are likely to be accessed
2293 immediately by \fBblkid(8)\fR or by the kernel scanning for a partition
2296 Default value: \fB131,072\fR.
2302 \fBzvol_request_sync\fR (uint)
2305 When processing I/O requests for a zvol submit them synchronously. This
2306 effectively limits the queue depth to 1 for each I/O submitter. When set
2307 to 0 requests are handled asynchronously by a thread pool. The number of
2308 requests which can be handled concurrently is controller by \fBzvol_threads\fR.
2310 Default value: \fB0\fR.
2316 \fBzvol_threads\fR (uint)
2319 Max number of threads which can handle zvol I/O requests concurrently.
2321 Default value: \fB32\fR.
2327 \fBzvol_volmode\fR (uint)
2330 Defines zvol block devices behaviour when \fBvolmode\fR is set to \fBdefault\fR.
2331 Valid values are \fB1\fR (full), \fB2\fR (dev) and \fB3\fR (none).
2333 Default value: \fB1\fR.
2339 \fBzfs_qat_disable\fR (int)
2342 This tunable disables qat hardware acceleration for gzip compression.
2343 It is available only if qat acceleration is compiled in and qat driver
2346 Use \fB1\fR for yes and \fB0\fR for no (default).
2349 .SH ZFS I/O SCHEDULER
2350 ZFS issues I/O operations to leaf vdevs to satisfy and complete I/Os.
2351 The I/O scheduler determines when and in what order those operations are
2352 issued. The I/O scheduler divides operations into five I/O classes
2353 prioritized in the following order: sync read, sync write, async read,
2354 async write, and scrub/resilver. Each queue defines the minimum and
2355 maximum number of concurrent operations that may be issued to the
2356 device. In addition, the device has an aggregate maximum,
2357 \fBzfs_vdev_max_active\fR. Note that the sum of the per-queue minimums
2358 must not exceed the aggregate maximum. If the sum of the per-queue
2359 maximums exceeds the aggregate maximum, then the number of active I/Os
2360 may reach \fBzfs_vdev_max_active\fR, in which case no further I/Os will
2361 be issued regardless of whether all per-queue minimums have been met.
2363 For many physical devices, throughput increases with the number of
2364 concurrent operations, but latency typically suffers. Further, physical
2365 devices typically have a limit at which more concurrent operations have no
2366 effect on throughput or can actually cause it to decrease.
2368 The scheduler selects the next operation to issue by first looking for an
2369 I/O class whose minimum has not been satisfied. Once all are satisfied and
2370 the aggregate maximum has not been hit, the scheduler looks for classes
2371 whose maximum has not been satisfied. Iteration through the I/O classes is
2372 done in the order specified above. No further operations are issued if the
2373 aggregate maximum number of concurrent operations has been hit or if there
2374 are no operations queued for an I/O class that has not hit its maximum.
2375 Every time an I/O is queued or an operation completes, the I/O scheduler
2376 looks for new operations to issue.
2378 In general, smaller max_active's will lead to lower latency of synchronous
2379 operations. Larger max_active's may lead to higher overall throughput,
2380 depending on underlying storage.
2382 The ratio of the queues' max_actives determines the balance of performance
2383 between reads, writes, and scrubs. E.g., increasing
2384 \fBzfs_vdev_scrub_max_active\fR will cause the scrub or resilver to complete
2385 more quickly, but reads and writes to have higher latency and lower throughput.
2387 All I/O classes have a fixed maximum number of outstanding operations
2388 except for the async write class. Asynchronous writes represent the data
2389 that is committed to stable storage during the syncing stage for
2390 transaction groups. Transaction groups enter the syncing state
2391 periodically so the number of queued async writes will quickly burst up
2392 and then bleed down to zero. Rather than servicing them as quickly as
2393 possible, the I/O scheduler changes the maximum number of active async
2394 write I/Os according to the amount of dirty data in the pool. Since
2395 both throughput and latency typically increase with the number of
2396 concurrent operations issued to physical devices, reducing the
2397 burstiness in the number of concurrent operations also stabilizes the
2398 response time of operations from other -- and in particular synchronous
2399 -- queues. In broad strokes, the I/O scheduler will issue more
2400 concurrent operations from the async write queue as there's more dirty
2405 The number of concurrent operations issued for the async write I/O class
2406 follows a piece-wise linear function defined by a few adjustable points.
2409 | o---------| <-- zfs_vdev_async_write_max_active
2416 |-------o | | <-- zfs_vdev_async_write_min_active
2417 0|_______^______|_________|
2418 0% | | 100% of zfs_dirty_data_max
2420 | `-- zfs_vdev_async_write_active_max_dirty_percent
2421 `--------- zfs_vdev_async_write_active_min_dirty_percent
2424 Until the amount of dirty data exceeds a minimum percentage of the dirty
2425 data allowed in the pool, the I/O scheduler will limit the number of
2426 concurrent operations to the minimum. As that threshold is crossed, the
2427 number of concurrent operations issued increases linearly to the maximum at
2428 the specified maximum percentage of the dirty data allowed in the pool.
2430 Ideally, the amount of dirty data on a busy pool will stay in the sloped
2431 part of the function between \fBzfs_vdev_async_write_active_min_dirty_percent\fR
2432 and \fBzfs_vdev_async_write_active_max_dirty_percent\fR. If it exceeds the
2433 maximum percentage, this indicates that the rate of incoming data is
2434 greater than the rate that the backend storage can handle. In this case, we
2435 must further throttle incoming writes, as described in the next section.
2437 .SH ZFS TRANSACTION DELAY
2438 We delay transactions when we've determined that the backend storage
2439 isn't able to accommodate the rate of incoming writes.
2441 If there is already a transaction waiting, we delay relative to when
2442 that transaction will finish waiting. This way the calculated delay time
2443 is independent of the number of threads concurrently executing
2446 If we are the only waiter, wait relative to when the transaction
2447 started, rather than the current time. This credits the transaction for
2448 "time already served", e.g. reading indirect blocks.
2450 The minimum time for a transaction to take is calculated as:
2452 min_time = zfs_delay_scale * (dirty - min) / (max - dirty)
2453 min_time is then capped at 100 milliseconds.
2456 The delay has two degrees of freedom that can be adjusted via tunables. The
2457 percentage of dirty data at which we start to delay is defined by
2458 \fBzfs_delay_min_dirty_percent\fR. This should typically be at or above
2459 \fBzfs_vdev_async_write_active_max_dirty_percent\fR so that we only start to
2460 delay after writing at full speed has failed to keep up with the incoming write
2461 rate. The scale of the curve is defined by \fBzfs_delay_scale\fR. Roughly speaking,
2462 this variable determines the amount of delay at the midpoint of the curve.
2466 10ms +-------------------------------------------------------------*+
2482 2ms + (midpoint) * +
2485 | zfs_delay_scale ----------> ******** |
2486 0 +-------------------------------------*********----------------+
2487 0% <- zfs_dirty_data_max -> 100%
2490 Note that since the delay is added to the outstanding time remaining on the
2491 most recent transaction, the delay is effectively the inverse of IOPS.
2492 Here the midpoint of 500us translates to 2000 IOPS. The shape of the curve
2493 was chosen such that small changes in the amount of accumulated dirty data
2494 in the first 3/4 of the curve yield relatively small differences in the
2497 The effects can be easier to understand when the amount of delay is
2498 represented on a log scale:
2502 100ms +-------------------------------------------------------------++
2511 + zfs_delay_scale ----------> ***** +
2522 +--------------------------------------------------------------+
2523 0% <- zfs_dirty_data_max -> 100%
2526 Note here that only as the amount of dirty data approaches its limit does
2527 the delay start to increase rapidly. The goal of a properly tuned system
2528 should be to keep the amount of dirty data out of that range by first
2529 ensuring that the appropriate limits are set for the I/O scheduler to reach
2530 optimal throughput on the backend storage, and then by changing the value
2531 of \fBzfs_delay_scale\fR to increase the steepness of the curve.
projects / zfs / blob