On Feb 17, 2010, at 7:54 AM, Mark Heily wrote:
> I have noticed a strange issue on two idle Solaris 10 U6 systems where the
reported size of the ARC cache is less than c_min. I was under the impression
that c_min is a hard limit on the minimum size of the ARC cache [1]. As an
example, I have included the output of arc_summary[2] and kstat[3] for one of
the systems.
>
> Is the formula "arcstats:size >= arcstats:c_min" always true?
No. c is the target size, not the actual size.
> Or are there legitimate cases when the ARC cache will shrink below the
c_min limit.
At boot, for instance.
-- richard
>
> Thanks,
>
> - Mark
>
>
> [1]
>
> http://mail.opensolaris.org/pipermail/perf-discuss/2009-June/002312.html
>
>
> [1] Excerpt from Ben Rockwood''s arc_summary:
>
> System Memory:
> Physical RAM: 32046 MB
> Free Memory : 4255 MB
> LotsFree: 496 MB
>
> ZFS Tunables (/etc/system):
>
> ARC Size:
> Current Size: 288 MB (arcsize)
> Target Size (Adaptive): 25728 MB (c)
> Min Size (Hard Limit): 3877 MB (zfs_arc_min)
> Max Size (Hard Limit): 31022 MB (zfs_arc_max)
>
>
> [2] Output of "kstat zfs:0:arcstats"
>
> module: zfs instance: 0
> name: arcstats class: misc
> c 26978509172
> c_max 32529489920
> c_min 4066186240
> crtime 2251383.06747617
> data_size 166427648
> deleted 20710
> demand_data_hits 1708848
> demand_data_misses 161416
> demand_metadata_hits 2532283
> demand_metadata_misses 129957
> evict_skip 455
> hash_chain_max 7
> hash_chains 66762
> hash_collisions 221681
> hash_elements 323508
> hash_elements_max 330316
> hdr_size 67452840
> hits 4582847
> l2_abort_lowmem 0
> l2_cksum_bad 0
> l2_evict_lock_retry 0
> l2_evict_reading 0
> l2_feeds 0
> l2_free_on_write 0
> l2_hdr_size 0
> l2_hits 0
> l2_io_error 0
> l2_misses 0
> l2_read_bytes 0
> l2_rw_clash 0
> l2_size 0
> l2_write_bytes 0
> l2_writes_done 0
> l2_writes_error 0
> l2_writes_hdr_miss 0
> l2_writes_sent 0
> memory_throttle_count 0
> mfu_ghost_hits 630167
> mfu_hits 3166923
> misses 1090660
> mru_ghost_hits 137293
> mru_hits 1076653
> mutex_miss 2290
> other_size 69767200
> p 3228355146
> prefetch_data_hits 5940
> prefetch_data_misses 761634
> prefetch_metadata_hits 335776
> prefetch_metadata_misses 37653
> recycle_miss 3549
> size 303647688
> snaptime 3616598.76049297
>
> _______________________________________________
> zfs-code mailing list
> zfs-code at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-code
ZFS storage and performance consulting at http://www.RichardElling.com
ZFS training on deduplication, NexentaStor, and NAS performance
http://nexenta-atlanta.eventbrite.com (March 15-17, 2010)