Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Hello, After experiencing several dual-node failures due to faulty RAID controller on one of nodes (it is famous 3ware 9650-16ML, which may reset and do not respond for 30 seconds at all due to command sent to its control channel) I noticed 'disk-timeout' option which may help with such issue. Visible symptoms are that monitor operations of iscsi target and luns (ietd) timeout due to block device (drbd in 'Primary') stops to respond because 'Secondary' replica does not respond (protocol C) due to raid reset. Target is then restarted. Of course, I can increase pacemaker monitor timeouts, but I honestly prefer disk to be dropped very quickly if it does not answer. It is natural that drbd goes to a 'Diskless' state if block device does not respond. Before enabling it I'd like to get "Yep, that's correct" here ;) One more thing I care about, is that it would be nice if drbd tries to re-attach disk back after some time (may be several attempts). Although this is probably very rare case (I was really surprised that hardware may do that), it would be very helpful for such brain-dead devices (Yes, I already ordered some new adapters and backplanes, but still want to reuse existing ones for less mission-critical tasks as they was quite expensive). Thank you, Vladislav