zfs arc and amount of wired memory

Artem Belevich art at freebsd.org
Thu Feb 9 01:08:05 UTC 2012

On Wed, Feb 8, 2012 at 4:28 PM, Jeremy Chadwick
<freebsd at jdc.parodius.com> wrote:
> On Thu, Feb 09, 2012 at 01:11:36AM +0100, Miroslav Lachman wrote:
>> ARC Size:
>>          Current Size:             1769 MB (arcsize)
>>          Target Size (Adaptive):   512 MB (c)
>>          Min Size (Hard Limit):    512 MB (zfs_arc_min)
>>          Max Size (Hard Limit):    3584 MB (zfs_arc_max)
>> The target size is going down to the min size and after few more
>> days, the system is so slow, that I must reboot the machine. Then it
>> is running fine for about 107 days and then it all repeat again.
>> You can see more on MRTG graphs
>> http://freebsd.quip.cz/ext/2012/2012-02-08-kiwi-mrtg-12-15/
>> You can see links to other useful informations on top of the page
>> (arc_summary, top, dmesg, fs usage, loader.conf)
>> There you can see nightly backups (higher CPU load started at
>> 01:13), otherwise the machine is idle.
>> It coresponds with ARC target size lowering in last 5 days
>> http://freebsd.quip.cz/ext/2012/2012-02-08-kiwi-mrtg-12-15/local_zfs_arcstats_size.html
>> And with ARC metadata cache overflowing the limit in last 5 days
>> http://freebsd.quip.cz/ext/2012/2012-02-08-kiwi-mrtg-12-15/local_zfs_vfs_meta.html
>> I don't know what's going on and I don't know if it is something
>> know / fixed in newer releases. We are running a few more ZFS
>> systems on 8.2 without this issue. But those systems are in
>> different roles.
> This sounds like the... damn, what is it called... some kind of internal
> "counter" or "ticks" thing within the ZFS code that was discovered to
> only begin happening after a certain period of time (which correlated to
> some number of days, possibly 107).  I'm sorry that I can't be more
> specific, but it's been discussed heavily on the lists in the past, and
> fixes for all of that were committed to RELENG_8.  I wish I could
> remember the name of the function or macro or variable name it pertained
> to, something like LTHAW or TLOCK or something like that.  I would say
> "I don't know why I can't remember", but I do know why I can't remember:
> because I gave up trying to track all of these problems.
> Does someone else remember this issue?  CC'ing Martin who might remember
> for certain.

It's LBOLT. :-)

And there was more than one related integer overflow. One of them
manifested itself as L2ARC feeding thread hogging CPU time after about
a month of uptime. Another one caused issue with ARC reclaim after 107
days. See more details in this thread:



More information about the freebsd-stable mailing list