[zfs-discuss] ZFS monitoring on Linux

Bogdan bogdar at gmail.com
Mon Jan 4 11:23:54 EST 2016


Gregor, you probably meant metaslab_df_free_pct option.
The question is: how to determine if ZFS fill/fragmentation is harmfull for
my current workload?

On Mon, Jan 4, 2016 at 7:06 PM, Gregor Kopka via zfs-discuss <
zfs-discuss at list.zfsonlinux.org> wrote:

>
>
> Am 04.01.2016 um 15:22 schrieb Daniel Letai via zfs-discuss:
>
> IIRC it has mostly to do with fragmentation - when capacity is over 80%,
> usually there isn't enough space to allocate contiguous blocks, and files
> get highly fragmented, resulting in very degraded performance of the FS. Of
> course, 80% of 1GB disks, is not the same as 80% of 10TB disks - you might
> not witness fragmentation until 99% of a 10TB disk, depending on your usage
> pattern (file sizes, and how much growth/shrinkage different files
> experience simultaneously).
>
> There might be other factors, but I think this used to be the major issue.
>
>
> IIRC the fill grade at which ZFS switches to another allocation scheme is
> hardcoded (or maybe a parameter?).
>
> Gregor
>
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at list.zfsonlinux.org
> http://list.zfsonlinux.org/cgi-bin/mailman/listinfo/zfs-discuss
>
>


-- 
WBR,  Bogdan B. Rudas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20160104/d02ed4ab/attachment.html>


More information about the zfs-discuss mailing list