# echo zfs_prefetch_disable/W0t1 | mdb -kw This mechanism looks at the patterns of reads to files and anticipates on some reads, thereby reducing application wait times. This parameter determines a file-level prefetching mechanism called zfetch. ZFS data block compression is controlled by the ZFS compression property that can be set per file system.

This parameter controls compression of ZFS metadata (indirect blocks only). Change depending on amount of RAM in your system). When you have 196GB of memory 1/32nd is 6.3GB, which can cause up to 20-30 seconds of unresponsiveness (depending on the record size). Back when 8GB was a lot of memory 1/32nd meant 256MB max at a time. This is limited because evicting large amounts of data from ARC stalls all other processes. Default max 1/32nd of the ARC can be evicted at a time. Based on that, we would be shrinking the arc by about 100MB per shrink event, rather than 6GB of RAM.Įvery second a process runs which checks if data can be removed from the ARC and evicts it. We tuned this to 11, which is 1/2048 of arc_max. By default this is set to 5, which equates to shrinking by 1/32 of arc_max. This variable controls the amount of RAM that arc_shrinks will try to reclaim. Zfs_arc_max, zfs_arc_min (deprecated in 11.2)ĭetermines the maximum/minimum size of the ZFS Adjustable Replacement Cache (ARC). Solaris 11.2 deprecates the zfs_arc_max kernel parameter in favor of user_reserve_hint_pct and that’s cool. # echo arc_reduce_dnlc_percent/W0t2 | mdb -kw I’m fixing this at the moment and I’ll send around a new binary this afternoon.

The version of statit that we’re using is still attached to ancient 32-bit counters that /are/ overflowing on our runs. So yeah, dnlc_nentries would be really interesting to see (especially if its << ncsize). If the ARC detects low memory (via arc_reclaim_needed()), then we call arc_kmem_reap_now() and subsequently dnlc_reduce_cache() – which reduces the # of dnlc entries by 3% (ARC_REDUCE_DNLC_PERCENT).

Vdev eviction openzfs