I had an incident couple of days ago that I cannot explain and I'm looking for some help. One of my shelves disconnected from the system (power issue). I’ve got a raid 5 that splits on two shelves (3 +2) and the shelves that failed was the one with the 2 disks. Even though the 1 hot spare was immediately started I don’t think the system could manage recover from losing 2 at the same time. When the lost shelve was re powered everything went back to normal. My question is: How the raid managed to continue to work even after losing 2 drives at the same time on raid 5? Many thanks, Eran
Did the host report any issues? Did it lose access to the LUN when you had the power issues, then come back when the power came back?
If so that wouldn't be a big deal, as no writes would be going to the LUN. I think the CLARiiON is smart enough to try to fix issues with unowned LUNs when disks go offline, then come back in a non-degraded state.
EMC support would need spcollects to review and make sure you are correct with which enclosure and raid groups were affected . Maybe one of the drives was already failed over to the Hot Spare
The power issue you saw, did it only affect 1 rail in the cabinet? If so, then the other side would have been fine and this may explain why you still were up. SPCollects and a support engagement could confirm.
Disks has two buses - if one bus on the DAE went down, you would still have the other bus - the LUN/RG would be "degraded" - if you had two Navisphere sessions open, one started on SPA and one on SPB, you could look at each GUI in the hardware section for each disk - one disk might be listed as faulted in one GUI, but not faulted in the other GUI.
driskollt1
131 Posts
575
0
Posted September 3rd, 2010 07:00
Did the host report any issues? Did it lose access to the LUN when you had the power issues, then come back when the power came back?
If so that wouldn't be a big deal, as no writes would be going to the LUN. I think the CLARiiON is smart enough to try to fix issues with unowned LUNs when disks go offline, then come back in a non-degraded state.