Red Hat Bugzilla – Bug 994549
If glusterd operations fails on the commit stage, we will end up having inconsistent state over the cluster.
Last modified: 2016-02-17 19:21:00 EST
Description of problem: If the glusterd operations are failed in commit stage, the whole cluster will have inconsistent data. This failure at commite stage could happen if disk is full on one or more machines in the cluster.
Version-Release number of selected component (if applicable): glusterfs-18.104.22.168rhs-1.el6rhs.x86_64
How reproducible: Happens everytime
Steps to Reproduce:
1.Create and start a volume in a cluster having 2 or more machines.
2.Copy data from /var/lib/glusterd/ to a temp directory
3.Mount a loopback device of some 50M on /var/lib/glusterd/ and copy all those data from temp dir to /var/lib/glusterd/
4.fill up the mount point /var/lib/glusterd using the command "dd if=/dev/zero of=/var/lib/glusterd/file bs=1M"
5.Now do any glusterd operation from different machine in the cluster, like setting a volume option.
6.The command fails at commit stage , but you will end up having inconsistent data over the cluster.
Actual results: failure at commit stage results in inconsistent data over the cluster
Expected results: Cluster should never end up in inconsistent data.
Thank you for submitting this issue for consideration in Red Hat Gluster Storage. The release for which you requested us to review, is now End of Life. Please See https://access.redhat.com/support/policy/updates/rhs/
If you can reproduce this bug against a currently maintained version of Red Hat Gluster Storage, please feel free to file a new report against the current release.