Gluster locking the DC
Hi, 1) FYI: https://bugzilla.redhat.com/show_bug.cgi?id=1217576 I will need some help from a Gluster expert to debug this issue, since I don't know Gluster internals. On a replica-3 cluster with best practices, it's very frustrating to see the whole DC going down because of a FS lock. 2) Should we at some point freeze gluster updates and stick to a stable version, or are we following new features required by oVirt? I feel Gluster very unstable yet, but maybe it's just me. I guess that if a "stat" operation (mount, lsof, etc) of a gluster volume is locking, it's the layer below sanlock, right? I don't think that sanlock should be able to lock Gluster, does it? 3) My feature request: Sanlock+Gluster unit-tests written by the sanlock authors to stress and exploit all Gluster weaknesses. PS: The good news is that [1] and other related issues were finally fixed. [1] https://bugzilla.redhat.com/show_bug.cgi?id=1201355 Best regards, Christopher.
Il 30/04/2015 20:17, Christopher Pereira ha scritto:
Hi,
1) FYI: https://bugzilla.redhat.com/show_bug.cgi?id=1217576 I will need some help from a Gluster expert to debug this issue, since I don't know Gluster internals.
Adding Sahina
On a replica-3 cluster with best practices, it's very frustrating to see the whole DC going down because of a FS lock.
2) Should we at some point freeze gluster updates and stick to a stable version, or are we following new features required by oVirt? I feel Gluster very unstable yet, but maybe it's just me. I guess that if a "stat" operation (mount, lsof, etc) of a gluster volume is locking, it's the layer below sanlock, right? I don't think that sanlock should be able to lock Gluster, does it?
3) My feature request: Sanlock+Gluster unit-tests written by the sanlock authors to stress and exploit all Gluster weaknesses.
PS: The good news is that [1] and other related issues were finally fixed.
[1] https://bugzilla.redhat.com/show_bug.cgi?id=1201355
Best regards, Christopher. _______________________________________________ Devel mailing list Devel@ovirt.org http://lists.ovirt.org/mailman/listinfo/devel
-- Sandro Bonazzola Better technology. Faster innovation. Powered by community collaboration. See how it works at redhat.com
Il 30/04/2015 20:17, Christopher Pereira ha scritto:
Hi,
1) FYI: https://bugzilla.redhat.com/show_bug.cgi?id=1217576 I will need some help from a Gluster expert to debug this issue, since I don't know Gluster internals. On a replica-3 cluster with best practices, it's very frustrating to see the whole DC going down because of a FS lock.
2) Should we at some point freeze gluster updates and stick to a stable version, or are we following new features required by oVirt? I feel Gluster very unstable yet, but maybe it's just me. I guess that if a "stat" operation (mount, lsof, etc) of a gluster volume is locking, it's the layer below sanlock, right? I don't think that sanlock should be able to lock Gluster, does it?
Gluster branched 3.7 from master, I've updated the ovirt-release rpms to follow the 3.7 nightly branch instead of master. We should move to the 3.7 stable branch after 3.7 will be released.
3) My feature request: Sanlock+Gluster unit-tests written by the sanlock authors to stress and exploit all Gluster weaknesses.
PS: The good news is that [1] and other related issues were finally fixed.
[1] https://bugzilla.redhat.com/show_bug.cgi?id=1201355
Best regards, Christopher. _______________________________________________ Devel mailing list Devel@ovirt.org http://lists.ovirt.org/mailman/listinfo/devel
-- Sandro Bonazzola Better technology. Faster innovation. Powered by community collaboration. See how it works at redhat.com
participants (2)
-
Christopher Pereira -
Sandro Bonazzola