1

I have a VPS on CentOS 5, running LAMP. Twice now in the last 3 months the server has had to be rebooted due to being out of memory. Downtime ends up being an hour, before someone can look at it. I have munin running to create some graphs, and am trying to figure out where to be looking (which logs) to find out what went wrong. There are some signs that it's busy when the crash happens, but need more/better diagnosis.

See the following munin graphs: https://i.stack.imgur.com/mIrwL.jpg https://i.stack.imgur.com/JAH2A.jpg https://i.stack.imgur.com/xDAEb.jpg https://i.stack.imgur.com/oHkv7.jpg

Firewall and Disk are high just before the gap,and after the restart, munin doesn't autorestart, which is why there's a gap.

Bart De Vos
  • 17,911
  • 6
  • 63
  • 82
joelhaasnoot
  • 15
  • 2

1 Answers1

1

VPSes are pretty notorious for going down as soon as your memory is used up. Once your memory is gone, it tries to hit the disk swap, and as soon as that happens on a VPS you are going to be dead.

I can't tell exactly based on that graph - do you have 256M or 512M RAM on your vps? If you are at 256 it isn't going to take much to overload your machine. Not much better at 512. Due to the bandwidth graph, I would say something is going on to trigger it, such as a backup or nightly cron job.

Two ways you can fix it.

The first is to figure out what is going on to cause the bandwidth usage and then stop that.

The second is to increase your VPS instance to give it some leeway before crashing.

Dave Drager
  • 8,375
  • 29
  • 45