Memory leak in ZFS?
gldisater at gmail.com
Wed Feb 9 00:49:43 UTC 2011
On 02/08/11 16:36, Bartosz Stec wrote:
>> gldisater at constans:~% zpool upgrade
>> This system is currently running ZFS pool version 28.
>> All pools are formatted using this version.
>> gldisater at constans:~% sh memleak-detect.sh
>> PHYSMEM: 12268.94 MB
>> ACTIVE: 427.57 MB
>> INACTIVE: 724.98 MB
>> WIRED: 10155.28 MB
>> CACHED: 28.32 MB
>> FREE: 556.16 MB
>> SUM: 11892.32 MB
>> MISSING: 376.61 MB
>> gldisater at constans:~% uname -a
>> FreeBSD constans 8.2-PRERELEASE FreeBSD 8.2-PRERELEASE #14: Tue Jan 18
>> 15:26:02 EST 2011 root at constans:/usr/obj/usr/src/sys/CONSTANS amd64
>> gldisater at constans:~% uptime
>> 3:22PM up 8 days, 5:07, 2 users, load averages: 0.21, 0.22, 0.24
> Have you noticed MISSING count increasing over time or you just looking
> at this exact calculation?
> This script is far from being perfect and I don't know as much about
> FreeBSD internals to calculate it to show exactly zero at this point (if
> it is possible at all).
> It only calculates physical memory minus memory pools visible in top(1),
> and on my system it gives MISSING count about 36MB just after reboot.
> What's important - check if this variable is stable over time. You can
> confirm memory leakage if MISSING is increasing over time or while using
> sendfile. You can try for instance 'sh memleak-detect.sh -m -r 60' to
> check it every one minute.
My MISSING has increased to 376.62 MB. My usage of this server is
primary reading not a lot of writing so a minutely check won't show
anything. I'll try to write a lot of data to it later.
More information about the freebsd-fs