3

I'm experiencing an issue when sometimes the memory gets 100% full, and the swap file also, and the server becomes non-responsive and has to be restarted (causing also problems in database). This is what Cacti shows:

Physical Memory Swap File

The server is running a web-app (database + apache) and during that specific moment didn't experience any ir-regular traffic or usage. This scenario happened twice in the last week. What can cause this? How can I resolve the issue?

Noam
  • 129
  • 1
  • 7
  • I would suggest you have a cron job send you outputs of ps aux, cat /proc/slabinfo etc. every minute or so, either to an email address or to sync-mounted storage. This should make it easy to see what process ran amuck after a crash. If you want to gain time, it might help to increase swap space (you can use a regular file as swap too, just loopback mount it!). – rackandboneman Jan 10 '13 at 10:27
  • 2
    @rackandboneman: Loopback mount swap file? Aieee! Don't do that. Just `dd if=/dev/zero of=/path/to/your/swapfile bs=1024 count=1048576` or so and then `mkswap /path/to/your/yourswapfile`. Then `swapon /path/to/your/swapfile` should work. – Janne Pikkarainen Jan 10 '13 at 10:33
  • If it works, all the better - I did not want to assume it will work on every linux system the TO could have (no version info given), since as per the docs a device file is expected for swapon. – rackandboneman Jan 10 '13 at 10:39
  • If you can log in interactively, you could also run top (press shift-m once it's running to sort processes by memory consumption). When the server stops responding, the top display will freeze, so you can examine it at your leisure. – Ansgar Esztermann Jan 10 '13 at 10:45

1 Answers1

-2

you can clear a cache of linux machine and pu it into a crontab entry to run in regular interval.

ramkumar crk
  • 1