[DRBD-user] linstor-proxmox controller toggle tests

[Robert Altnoeder]

On 11/14/18 1:09 PM, opengreb at free.fr wrote:
> The started vms that we stopped can not restart and the vms that were already stopped can not start.
> The following error is then handled: drbdsetup wait-connect-resource vm-xxx-disk-x
> As long as a peer resource is not available to respond, operations fail. A timeout is then generated.

I am pretty sure that this behavior is caused by some other component in
the system. DRBD does not decide on its own that now is a good time to
run a drbdsetup wait-connect-resource command for no apparent reason.
Obviously, if the other node is not online, then wait-connect-resource
will fail.

> The solution Yannis has found is simply to disconnect the resource before starting the vm. (drbdadm disconnect vm-xxx-disk-x or with drbdtop)

That's a workaround. The real question is why your system is
intentionally running a command that is completely unnecessary and that
is guaranteed to fail if one node is not online. The solution would be
not running the command when you actually want to access the storage
device while its replication link is down.

br,
Robert