We use XFS on EC2 EBS running on Arch Linux (kernel 3.6.7-1-ARCH). Before we take an EC2 snapshot, we call xfs_freeze on the mount so that we can have a consistent snapshot.
On instances running for around 30 days, we see the XFS freeze hang, resulting in a skyrocketing load (but not CPU). We can't kill the xfs_freeze process. The only route available is usually to force a reboot of the server without cleaning unmounting.
There are backtraces in dmesg complaining both of hung xfs processes, and hung java processes. https://gist.github.com/3d46830eac52df44d30f
Is this a known issue? We call xfs_freeze quite often on thousands of XFS volumes, so I wouldn't be surprised if we have encountered an edge case considering how often we do this.
Any help appreciated. I'm recampbell on #xfs/freenode.
Thanks,
Ryan
_______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
