Bug 1327366

glusterd volfile as seen in rhssp4 was :

<snip>
volume management
    type mgmt/glusterd
    option working-directory /var/lib/glusterd
    option transport-type socket,rdma
    option transport.socket.keepalive-time 10
    option transport.socket.keepalive-interval 2
    option transport.socket.read-fail-log off
    option ping-timeout 0
#   option base-port 49152
# 2014-12-19 b.goodfellow added for SMB
    option rpc-auth-allow-insecure on
end-volume
</snip>

The glusterd volfile has no option for event-threads, which means that glusterd will start with the default ( which is 2 ) number of epoll threads.

After upgrading to RHGS 3.1.1, the glusterd volfile should have got updated with this entry "option event-threads 1".

We already know of glusterd issues with multiple epoll threads and that's the reason why glusterd is started with single epoll thread.

A big thank you Satheesaran for identifying this. 

Raghavendra,

Question for the customer here is how did this happen?

1. Did the accidentally remove the event-threads entry from glusterd.vol ?
2. Or there was an upgrade issue where the binaries were successfully upgraded but not the configuration including glusterd.vol? We have seen a similar case in BZ 1305884.

As Satheesaran pointed out, that we have some known issues if multi threaded e-poll is enabled in glusterd.

Considering it, can we close this bug?

Oonkwee, Raghavendra,

There is no point of looking at the core since the set up was incorrectly configured and multi threaded e-poll got enabled by default in glusterd accidentally due to a manual change in glusterd.vol file. I'd request you to close this bug.

~Atin

Closing