SUMMARY: What causes /tmp: File system full, swap space limit exceeded

Here are the responses received. Thanks to all of you. I am going to setup a
tmpfs max size using (size=) option in vfstab.

Jay Lessert  response:

Regarding finding root cause, Not much, now.  The evidence is gone.

  The problem is either large processes, or files in /tmp.

  Write a little script that does:

      swap -s (and looks for low "available" number)
      df /tmp/. (and looks for low "avail" number)

  Check every 10 minutes and log to a file (or e-mail you if
  either one gets lower than you want).

Grzegorz Bakalarski reponse:

  WHile I cannot answer your question I'd like to tell you
  that if the problem is that someone overfull /tmp (which is not
  separate but on swap - standard solaris installation) and this
  causes no swap space, you can limit size of  /tmp fs using e.g
  such setting in /etc/vfstab
  swap    -       /tmp   tmpfs   -       yes     size=4096m

  which keeps your /tmp under 4GB limit
  Other solution is not using tmpfs but separate /tmp from swap
  putting it on regular partitition ...


Hargrave, Mark's response

   John,

   If you haven't already, and /tmp is smaller than swap,
   add this to the /etc/vfstab file:

   swap    -       /tmp    tmpfs   -       yes     -

   Thanks,

   Mark


Rich Teer's response:

   One of our 420R was hard rebooted as we got /tmp: File system full, swap
 space limit exceeded messages and lost conectivity.

   System has 2GB memory and 4GB swap and database and apache web server
 are running on the system.

   How would i find out now what was caused the problem.

   System accounting if off, sa/sar files not collected.

   What can i do prevent these?

  By rebooting your machine, you unfortunately destroyed all
  evidense that will help you!  (Needless reboots are the
  M$ solution to problems...)

  A guess would be that some process(es) filled up /tmp,
  eating away your available VM swap space.  If that's the
  case, you can limit the size of a tmpfs file system,
  see the mount_tmpfs man page.


varad.rajan's response:


  First look at the messages file to see what time it rebooted. If it is the
  same time everyday?. Then write a script to clean up the tmp file before
 an
  hour of the reboot.
  Otherwise if it gets to 90% full then cleanup. I hope it helps.

  ..Vard


Andres Rojas's response:

  I've seen sporatical runaway processes that eatup the swap space
 exhausting the /tmp at the same time. This is due to the fact that your
machines
 would be configured with a common physical space for both: swap and /tmp
(see
 "df  -n").

  One chance is to periodically (and very carefully) remove unused temporary
  files from /tmp which will free swap space. Another is to add more swap
  space (see "man swap") that will increase the /tmp total size.


  A good solution we took as a standard is to reserve disk space for swap in
  a given partition and, for /tmp in a dedicated separated partition.


Original Question:
  One of our 420R was hard rebooted as we got /tmp: File system full, swap
  space limit exceeded messages and lost conectivity.

  System has 2GB memory and 4GB swap and database and apache web server are
  running on the system.

  How would i find out now what was caused the problem.

  System accounting if off, sa/sar files not collected.

  What can i do prevent these?

  TIA
  John
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers