Overprotective, friendly and very very bad

[ozihcs]

.. well pardon the inflammatory subject, but currently I'm a bit annoyed because I've just realized that my entire production farm will be running in slow motion for the entire weekend because I have a faulted disk being replaced by hotspare.

I can understand the rationale behind forcibly disabling the write cache if there are faults, but in this case the protective measures are by themselves much more catastrophic than the drive failure which caused it. And to top it off there doesn't even seem to be any way to reenable the cache and accept the risk of data loss if another 2 drives should fail.


Oh well. I'm sure that it seemed like a very good idea at the time..

 

[comtec17]

Write cache on a Clariion should only be disabled if your SPS battery is low or down and/or you have an SP down. A failed drive should not cause write cache to disable.

 

[onefortheroad]

in addition to the above, the rebuild to the hotspare shouldn't take long so it should be back to normal performance within an hour tops (depending on the rebuild priority set for the raid group with the failed) if your cache is disabled it may be wise to check the SP event logs and see if anything else has happened i.e SP reboot.

 

[ozihcs]

Still waiting for my replacement drive, but I've discovered why the cache was disabled. If the HA Cache Vault option is enabled on the storage array (default), any faulted disk will cause a disable of the write cache. With this option off, the cache will stay enabled. Unfortunately, once the cache has been disabled, clearing the HA Cache Vault tick won't allow the write cache to enable - only a replacement of the failed drive will fix it. Safe I guess, but I may choose to disable it since the lack of a write cache in our case is far worse than the lack of a free hotspare drive.

 

[squiddog]

As ozihcs noted, CLARiiONs require that the vault be present and accounted for.

If you have only 10 drives, well, any failure of the first 9 will kill your write cache.

If you lost a drive on a second DAE, then it wouldn't affect it.

I was reminded of this when I played around with drive rebuilding on my 4500 (I run one at home)