[Bugs] [Bug 1532868] gluster upgrade causes vm disk errors

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1532868

Krutika Dhananjay <kdhananj at redhat.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |needinfo?(bugs at josemaas.net
                   |                            |)



--- Comment #8 from Krutika Dhananjay <kdhananj at redhat.com> ---
Thanks David.

This issue is caused due to a bug in bit-rot. The patch at
https://review.gluster.org/#/c/19021/ should fix this.

There is evidence in the logs to suggest everytime there was a log from shard
about missing trusted.glusterfs.shard.file-size xattr, there was a metadata
heal on the same file preceding it.

There is also evidence in tcpdump shared by David that suggests in one rare
instance of STAT, the xattr trusted.glusterfs.shard.file-size xattr was not
returned by the brick and right then the pause happened.

In upstream this fix is available since glusterfs-3.12.5.

You can work around the issue by disabling metadata-self-heal:

# gluster volume set <VOL> cluster.metadata-self-heal off

Could you please confirm if this fixed the issue for you?

-Krutika

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are on the CC list for the bug.