[Gluster-users] Fwd: Re: Use of xfs_fsr
trtracer at gmail.com
Sat Aug 31 17:00:54 UTC 2013
Am 30.08.2013 21:40, schrieb Robert Hajime Lanning:
> On 08/29/13 10:50, Tobias Reinhard wrote:
>> I would like to defrag the XFS partition on which my glusterfs brick is.
>> Normally no program cares about xfs_fsr.
>> Just to make sure: is it "safe" to run xfs_fsr on a glusterfs brick
>> while everything is online or does any part of glusterfs gets confused?
>> I have a two node replicated volume (if that is relevant) running
>> glusterfs 3.4
> The GlusterFS server is all userspace. So, as long as you are not
> taking the underlying (XFS) filesystem offline, it won't care one bit.
> Though you might see a performance hit for anything doing I/O to the
> filesystem in question.
> Do you run your filesystems near full capacity? (wondering why you
> need a defrag.)
The FS is 70% full (on a 3.5TB parition).
I placed some KVM images on the cluster in QED format. It seems they get
fragmented as they grow.
Whenever a node goes offline and gets online again the self-healing is
nearly killing the cluster. The images are only read at 15mb/s.
Even throtteling down the self-heal only helps a little.
I'll try if a defrag solves this issue.
More information about the Gluster-users