> I've moved our database from a single 30G drive with reiserfs (3.6.25 on
> kernel 2.4.9) to a 4-drive RAID 5 of 80G disks with reiserfs on that
> (yeilding 167G). The performance is fantastic, but the backend server
> is periodically crashing.
>
> With almost no information whatsoever, anyone care to speculate on if
> this is:
> 1) hardware problem (seems unlikely since the drives are < 1 year old)
> 2) reiserfs or kernel bug
> 3) problem with a) shutting down postgres, b) cp -arv /olddata /newdata,
> c) edit /etc/rc.d, d) start postgres
> 4) something else entirely
We'll really need more information to go on. There are about a hundred
possibilities, so we'll need to know the make/model of board, driver
version, kernel version, all of that.
. I had a similar problem with AMI MegaRAID controllers. I tried three
different models, and they all had the same problem - they would lock up
if the array was left *idle* for too long. It turned out to be a problem
in the Linux drivers. I did a little tweaking and got it to stop crashing
more or less, but the downtime from a single lock-up cost us more than the
boards, so we just bought brand-new Mylex controllers. Later, I noticed
that some changes pretty much like mine had made it into the official
Linux kernel, but it still didn't solve the problem 100%.
One thing that I learned awfully quickly was that it was imperative to
disable Linux's screen-blanking. Many times, the machine would lock up
hard while the screen was blanked, and I'd never get to see any messages.
steve