[Gluster-users] Shutting down a server with no service disruption and Xen instance errors

[Torbjørn Thorsen]

Hello.

We've got a 2x2 volume with two servers, holding Xen images for about
200 instances.
Yesterday we wanted to reboot one of the servers into a new kernel and
a new version of Gluster, and simply rebooted the server in question.

This caused problems for some of the Xen instances, giving IO errors
in their "dmesg" output.
Some instances had to be rebooted and fsck-ed to come up again.

Is there any way to inform the Gluster FUSE client to stop using a
certain server,
so that a reboot of the server can be performed gracefully ?

I'm thinking maybe reducing network.ping-timeout to something very low,
so that the kernel of the Xen instance will not complain about IO errors.