2013-06-05 10:14 PM
Assume for a moment, either by severing the interconnect cables or by a mobo/nvram issue that fries c0a and c0b on one head, that `storage failover show` says SFO is dead on your 2-node cluster. (don't say it can't/shouldn't/won't happen. Really. Don't.)
The docs say, for a 2-node, "both nodes are continuously polled to ensure that if takeover occurs, the node that is still up and running has full read-write access to data as well as access to logical interfaces and management functions." But if there's no interconnect, then what?
If you don't have SFO, but do have space, you can vol move over and reboot the damaged head. But without interconnects, when that damaged head goes away and the clustered interfaces stop responding, haven't you just gone split-brain?
For that matter, what DOES an OOQ NAS cluster look like if you go poking around ngsh while it's OOQ?
2013-07-25 10:33 PM
Well, partly solved after talking to Support. Don't keep cluster HA enabled if this happens. Disable HA, set epsilon on the designated survivor, then reboot the damaged head. Epsilon will win the election and life will go on in the reboot, then undo it all after the head comes back.
Would still be good to know the fast way to determine that you're not serving because you went OOQ. I learned this one from painful experience.