The big Hitachi array went down a few years ago due to admin upgrading both paths. First time they allowed no failbackk time. Took out everything. 2nd time they waited a day but didn;t notice that some paths had failures elsewhere and took out several hosts. Another time we lost 3 disks in 6140 because noone noticed and lost a whole parity group. Failover won't help the first or third but would the 2nd.
----- Original Message ----- From: [email protected] <[email protected]> To: [email protected] <[email protected]> Sent: Wed May 18 13:38:22 2011 Subject: Re: [ldoms-discuss] Solaris clusters On 5/18/2011 12:46 PM, Hudes, Dana wrote: > Sorry I was unclear. All LUNs from all arrays (we have 4 kinds of arrays on > the FC SAN no 2 alike) are configured to be visible to the entire LDOM > cluster. I was discussing impact of a failure of some kind which made some > LUNs unreachable (e.g. Failure of the entire array perhaps because the SAN > admins upgraded the array firmware in an improper manner and took out both > paths or perhaps a power transient tripped us to battery and we have too may > KVA and one array gets hit or even better the whole array doesn't go away > only the one parity group loses 2 disks in a 6+1 config and your LUN is now > history). > In such case failover won't help. sure if you want array protection then one need to do SW mirroring between array or you pay big $$$ for enterprise array that claim never went down, due to controller failure or firmware upgrade > > ----- Original Message ----- > From: > [email protected]<[email protected]> > To: [email protected]<[email protected]> > Sent: Wed May 18 12:26:10 2011 > Subject: Re: [ldoms-discuss] Solaris cluster > > in addition in order to move zones between hosts(real or ldom) it is > much better to have shared SAN > > > On 5/18/2011 12:21 PM, Hung-ShengTsao (Lao Tsao) Ph.D. wrote: >> hi >> it seems that you are make life too complicated >> 1)if you want to failover zone, you will need SAN that can be access >> by both zone (with or without ldom) >> 2)if you want failover ldom then you also need SAN that can be access >> by both LDOM >> >> On 5/18/2011 12:01 PM, [email protected] wrote: >>> Is it possible (and a good idea or bad idea) to run Solaris cluster >>> in failover mode for LDOM2 guests and also separate Solaris Cluster >>> of zones in the guests? Auto failover would seem tricky if the LDOM >>> can failover though criteria are tricky. >>> For example if Zone A in a LDOM has storage from SAN array S1 and >>> Zone B in same LDOM has storage from SAN array S2 where there is one >>> Fiber Channel fabric (dual attach hopefully to 2 switches) for both >>> arrays and then zone A loses connection to S1 but B is happily using >>> S2 zone A should failover (or should it?) but not the whole LDOM. >>> >>> Load balancing of zones in LDOMs is desirable. Perhaps do not allow >>> failover of zones? >>> >>> Sent from my BlackBerry® smartphone with Nextel Direct Connect >>> _______________________________________________ >>> ldoms-discuss mailing list >>> [email protected] >>> http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss > _______________________________________________ > ldoms-discuss mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss _______________________________________________ ldoms-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss
