Monitoring zfs arc size (for tuning)
jtotz at imperial.ac.uk
Fri Aug 26 14:19:12 UTC 2011
I've been monitoring my arc size, hit and miss rates etc to determine
what a minimum arc size would be without performance going down the
drain, for my workload.
The only interesting point so far is that arc size still exceeds its
That is, kstat.zfs.misc.arcstats.size is larger than
kstat.zfs.misc.arcstats.c_max (which is the same as vfs.zfs.arc_max) by
a few hundred megabytes.
Also, some of the periodic tasks that this machine does makes
kstat.zfs.misc.arcstats.size go down to almost zero, and then shoot up
to the max again in a short period of time.
This is on:
FreeBSD XXX 8.2-STABLE FreeBSD 8.2-STABLE #0 r224227: Wed Jul 20
16:55:23 BST 2011 root at XXX:/usr/obj/usr/src/sys/GENERIC amd64
More information about the freebsd-fs