[zfs-discuss] ZFS, DRBD, GFS and Clustering

Gordan Bobic gordan.bobic at gmail.com
Wed Oct 16 05:10:33 EDT 2013


In fairness, all cluster file systems suffer from it due to lock bouncing.
Lock caching in GFS/GFS2 helps a lot, but it takes a lot of care in design
higher up to avoid nodes from stepping all over each other's locks and
causing massive thrashing.



On Wed, Oct 16, 2013 at 9:58 AM, Reinis Rozitis <r at roze.lv> wrote:

>  Performance is was also very poor on metadata-heavy workloads.
>>
>
> It is still the case (well in general they can't work (easily) arround it
> since there is no dedicated service for metadata and everything is read
> from file attributes).
>
> There are also some issues with ZFS regarding xattrs:
>
> https://github.com/zfsonlinux/**zfs/issues/457<https://github.com/zfsonlinux/zfs/issues/457>
> https://github.com/zfsonlinux/**zfs/issues/1469<https://github.com/zfsonlinux/zfs/issues/1469>
> https://github.com/zfsonlinux/**zfs/issues/1176<https://github.com/zfsonlinux/zfs/issues/1176>
>
> In short zfs might hiccup on some operations (mount/unmount) if large
> amount of files with xatts are deleted.
>
>
> rr
> To unsubscribe from this group and stop receiving emails from it, send an
> email to zfs-discuss+unsubscribe@**zfsonlinux.org<zfs-discuss%2Bunsubscribe at zfsonlinux.org>
> .
>

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe at zfsonlinux.org.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20131016/a96354bb/attachment.html>


More information about the zfs-discuss mailing list