zfs arc - just take it all and be good to me

[Artem Belevich]

I've tried the change on my ZFS-only 8-stable box which has
v15+metaslab patches applied and so far it seems to work OK under
memory shortage. As intended ZFS does not start giving up memory until
pagedaemon wakes up. I'll bang on the box some more through the week.

--Artem



2010/8/24 Andriy Gapon <avg at icyb.net.ua>:
> on 24/08/2010 23:05 Peter Jeremy said the following:
>> On 2010-Aug-24 17:00:36 +0200, Ollivier Robert <roberto at keltia.freenix.fr> wrote:
>>> According to Peter Jeremy:
>>>> I suspect Artem is referring to his patch at http://pastebin.com/ZCkzkWcs
>>>> which I have tweaked somewhat (see the last patch in
>>>> http://www.freebsd.org/cgi/query-pr.cgi?pr=146410 ).
>>>
>>> Thanks, cou you please send it in a non-QP-encoded form please?
>>
>> See attached.
>>>> Whilst these patches _are_ hacks, they seem to do a good job of
>>>> making ZFS and UFS play together.
>>>
>>> Is the patch only useful in these mixed situations or could it be also interestng for those of use running full-zfs (cf. http://www.keltia.net/howtos/zfsboot)?
>>
>> I think it will be useful.  As well as the trivial fix to count
>> "cache" as "free" space (which is now in -stable), the intent of the
>> patch is to improve the ability of ZFS to apply pressure to the VM
>> subsystem.  In theory, this should improve overall system performance
>> even in a ZFS-only environment where there is memory pressure due to
>> large, long-running processes
>
> Peter, Ollivier,
>
> would it be possible for you to test patch informally specified in the following
> post?
> http://thread.gmane.org/gmane.os.freebsd.devel.hackers/40788
>
> --
> Andriy Gapon
> _______________________________________________
> freebsd-fs at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-fs
> To unsubscribe, send any mail to "freebsd-fs-unsubscribe at freebsd.org"
>