How many failed drives can take down a RAID5 array?
The answer, of course, is “two or more”. And it’s not nice when it happens.
Two of the drives on a nice shiny FC array failed at approximately the same time (possibly within about two minutes of each other), and both were in the same RAID5 array. Definitely not good. On the other hand I confirmed that PostgreSQL can run off NFS (both server and client on FreeBSD) without problems so far (this is the temporary setup until we get a new array).
Replication is what you want if your data is realtime valuable to you:
http://linuxfinances.info/info/slony.html
It’s a pain to setup the first two times, but then it works like a charm.
Edwin
Comment by Edwin Groothuis — June 13, 2008 @ 9:26 pm
Right. Unfortunately in this case the other component in the replication pair would have been on a VMware virtual machine stored on the same array even if we had it
Luckily the backups were good and recent.
Anyway I’m hoping the build-in replication will be done in 8.4. Last time I looked at third-party replicators they had problems with schema changes.
Comment by ivoras — June 13, 2008 @ 11:06 pm