[Linux-cluster] GFS hangs, nodes die

[Sebastian Walter]

Hi Maciej,

Maciej Bogucki wrote:
> 1. Do You have Fibre-Channel SAN storage or use GNDB/iSCSI?
This is FC SAN on QLogic HBAs.

> 2. The other nodes can't access GFS fs because cluster is in recover
> state. Do You have fencing properly configured?
>   
Fencing normally works (e.g. when I manually reboot a node, it gets
restarted again after 10 seconds). With this problem showing up, fenced
is not interacting as far as I see it.

The cluster generally works well. Only if I increase the I/O load to a
certain level, the problems appear.

Regards,
Sebastian