[Gluster-users] libgfapi failover problem on replica bricks

[Paul Penev]

I did some more testing. This time with network.ping-timeout set to 2 seconds.

The results are the same. A reset (killall -KILL glusterfsd) on the
"bad" brick will bring the disk io down on the machine and ultimately
require a re-initialization of kvm+libgfapi.

2 seconds is well inside the normal tollerance for the root
filesystem, so it must be some other type of problem. Perhaps libgfapi
returns an error when it should not ?

I am sure that this is not intended behavior for gluster and libgfapi:
it defeats the whole purpose of having a replicated cluster and
running kvm on top of that.

Any comments are welcome.

Paul