[ZFS] ARC accounting bug ?

[Linda Kateley]

Metadata updates when you access a file. Atime is most often updated and 
you will miss.

If you are running vm's on zfs I always recommend to turn off atime updates.

Metadata is usually tiny so you probably won't see disk activity


On 8/26/16 10:27 AM, Ben RUBSON wrote:
> I'm sorry but I still can't find out where do these ARC demand metadata misses come from.
> In addition, why they are not cached, as my ARC is 1% used.
> And why I don't see any disk activity to honour them.
>
> I would really be glad if you could enlighten me :)
>
> Many thanks again,
>
> Ben
>
>
>
>> On 26 Aug 2016, at 14:28, Ben RUBSON <ben.rubson at gmail.com> wrote:
>>
>> Yes Juergen, metadata only stores in ARC files' info (location in pool and some other useful info), not the user data.
>>
>> I think "find" only needs metadata, as it does not read the files' content.
>> Which can be "verified" because :
>> - there is no IO at all on the pool (after the first find loop, once ARC has the needed metadata) ;
>> - there is no ARC demand data at all.
>> But I may be wrong, so of course feel free to correct me. Thank you.
>>
>> Thank you for the links, I already read them before posting here, but I may have missed something.
>>
>> Ben
>>
>>> On 26 Aug 2016, at 14:16, InterNetX - Juergen Gotteswinter <juergen.gotteswinter at internetx.com> wrote:
>>>
>>> so, you know the difference between
>>>
>>> - metadata
>>> - data
>>>
>>> ?
>>>
>>> if yes, what is your find actually doing in this case?
>>>
>>> regarding arc
>>>
>>> http://dtrace.org/blogs/brendan/2012/01/09/activity-of-the-zfs-arc/
>>> https://www.patpro.net/blog/index.php/2014/03/19/2628-zfs-primarycache-all-versus-metadata/
>>>
>>>
>>>
>>> Am 26.08.2016 um 14:06 schrieb Ben RUBSON:
>>>> Juergen, thank you for your answer.
>>>> So where do you think these ARC misses come from ? I can't find out :S
>>>> Thank you !
>>>>
>>>>> On 26 Aug 2016, at 14:01, InterNetX - Juergen Gotteswinter <juergen.gotteswinter at internetx.com> wrote:
>>>>>
>>>>> no bug, everything like i whould expect
>>>>>
>>>>> Am 26.08.2016 um 11:40 schrieb Ben RUBSON:
>>>>>> uname -v
>>>>>> FreeBSD 11.0-RC2 #0 r304729: Wed Aug 24 06:59:03 UTC 2016 root at releng2.nyi.freebsd.org:/usr/obj/usr/src/sys/GENERIC
> _______________________________________________
> freebsd-fs at freebsd.org mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-fs
> To unsubscribe, send any mail to "freebsd-fs-unsubscribe at freebsd.org"