[Gluster-users] Update on GlusterFS-3.7.10

Kaushal M kshlmster at gmail.com
Thu Apr 7 05:42:51 UTC 2016


Some of you might have seen GlusterFS-3.7.10 packages appear in the
repositories, and wondered why an announcement hasn't been made.

GlusterFS-3.7.10 contains a regression (also present in 3.7.9), which
could prevent GlusterD from starting after a reboot.

This regression was decided as a blocker, and a decision was made to
do a quick GlusterFS-3.7.11 release solving it. The 3.7.11 release
should be available in within the next 2 days.


## The regression

GlusterD would fail to start after a system reboot, in case
snapshotted volumes were present. As a fix for another bug, GlusterD
began resolving and saving the resolved brick paths on startup. The
same also needs to be done for snapshotted bricks paths. But the path
resolution can only happen if the logical volume (LV) snapshot is
mounted, which is handled by GlusterD.

On a restart, the snapshot brick paths aren't mounted. GlusterD would
try to resolve this paths, before it had actually mounted these paths,
and fail to start.

There is a workaround present though. This script [1] will create the
paths for the snapshot bricks and allow GlusterD to start.

## Who it affects

This regression affects users running GlusterFS servers on Linux, and
who are using GlusterFS volume snapshots. If you are, it is
recommended to skip 3.7.10 and upgrade directly to 3.7.11 when
available.

BSDs should not be affected as snapshots don't work on them. Users not
using volume snapshots are also safe.


[1]:  https://gist.github.com/atinmu/a3682ba6782e1d79cf4362d040a89bd1


More information about the Gluster-users mailing list