Ticket #72 (new developer task)

Opened 4 years ago

Last modified 2 years ago

Make OOM-killing happen quicker (Reduce swap partition size?)

Reported by: tobixen Owned by: tobixen
Priority: major Milestone:
Component: ServerSetup Keywords:
Cc:

Description

We have just been down for some 60-90 minutes - server being so swamped it took some 30 minutes just to process a simple "sudo /etc/init.d/apache restart" written on the command line. This is not acceptable; when the box is that swamped, the OOM killer should be activated.

Maybe we can reduce the swap partition size? Or maybe there are some parameters under /proc/sys/ that can be tuned?

Change History

comment:1 Changed 4 years ago by anonymous

I'm not sure if the partition size is the reason for problems. We will reach out of memory situations a lot faster if we resize the swap partition. Often probably even when it's not necessary. I would concentrate on limiting apache process memory usage instead. There are some ways in linux to limit memory for some processes, or am I wrong?

comment:2 Changed 4 years ago by tobixen

  • Owner changed from hannu to tobixen

You are right ... ulimits. I will eventually look into it.

comment:3 Changed 2 years ago by fake51

  • follow_up set to none
  • freq_reported set to 1
  • show_on_bw set to 0

Did anything ever get done on this? Cant recall any recent problems with bull but would still be nice to avoid

Note: See TracTickets for help on using tickets.