2

My RAID card does not currently have a battery installed and its write mode is set to write-through. Data is rarely written to the RAID - large batches of data come in maybe every other week. Daily use is read-only. The card hosts 4 different RAID volumes, each being around 12 TB large.

I'm trying to assess the risk of my RAID being unrecoverable if the card fails. Have you ever seen a RAID card turn bad and render the RAID unusable when the card was replaced?

Phil
  • 1,013
  • 2
  • 12
  • 16
  • What are you replacing it with? I've never had issues replacing a failed RAID card with a same model replacement, but have never used that brand, or tried replacing a failed RAID card with a different make or model before... – HopelessN00b Jun 29 '12 at 16:29
  • I'd replace it with the same model. Just wondering how likely it would be. – Phil Jun 29 '12 at 16:32
  • This isn't an answerable question. There are hundreds of ways that any computer component can fail, from simply keeling over dead to exploding in a rain of fire and sparks. The only possible answer would be "it depends on how it fails." Just remember that **RAID is not a backup mechanism** and plan accordingly with off-machine backups. – Charles Jun 29 '12 at 16:36
  • @Charles I know there are lots of ways for a RAID to fail, but the question is intended to gauge how much risk is involved. If there is a lot of risk, I can spend more money on mitigating that risk. If there's hardly any risk at all, I can worry about other things. – Phil Jun 29 '12 at 16:38
  • 1
    What if it's not the RAID card that fails? What if the expansion slot it's plugged into has an electrical fault? What if it's actually the cable that's gone bad? You're focusing on a single point of failure and trying to mitigate it. You should be trying to mitigate the *result* of failure, not just the cause. If you're paranoid about losing your RAID array, keep it backed up. – Charles Jun 29 '12 at 16:42
  • Charles is correct. The risk of the RAID card *specifically* corrupting your data when it fails is low. But doesn't rule out rule out other sources of corruption or possibilities of failure. If you care about the data, back it up. RAID is not a backup. – HopelessN00b Jun 29 '12 at 16:51
  • Thanks for the insight. You're correct. For what it's worth, I do keep a backup, but to restore everything after a catastrophic failure would require restoring _and_ reprocessing all the backup data. That would take a lot of time. I'm trying to decide if I need to keep a more current backup of the RAID on a new nearby server (which would be expensive but eliminate the reprocessing) or if the multi-week restore time is acceptable considering the small size of the risk. – Phil Jun 29 '12 at 18:34
  • Sounds like a business decision which I'd pawn off on my boss. "If you want to be able to restore this system in hours, rather than weeks, it will require keeping a current backup on disk, costing $X. If not, we save the money, but we'll be restoring for weeks in the unlikely event of a data failure." CYA, not to mention save yourself the headache. – HopelessN00b Jun 29 '12 at 19:15

1 Answers1

3

With write-through and a modern file system your drive should always be left in a consistant state on power failure as data is written both to the cache and the disk. It is write-back that requires a BBU because data is written to cache and then to disk.

Obviously it can still be nonsensical or not what you had open depending on the functionality of any open apps.

As people have said in the comments RAID is not a backup, drives do fail in groups under the added load of a rebuild, users do delete data by mistake, etc.

JamesRyan
  • 8,166
  • 2
  • 25
  • 36
  • Just to add to that line about nonsensical data, I have actually seen this corrupt the data on a RAID 5 array and destroy years of work because no one bothered to back it up. POS app (which was running from the RAID 5 array) crashed really hard one day and required a hard boot of the server... when the server came back up, the file system was totally hosed. "It's on RAID, so there's no risk of data loss!" Yeah, not so much. Many man-years of work lost to save a few grand on tape backups. – HopelessN00b Jun 29 '12 at 17:47