Re: [ProgSoc] On rebooting servers

[Paul Mclachlan <subtle@nospam.netaus.net.au>]

At 13:47 12 Dec 1998 +1030, daedalus wrote:

| I realise there are certain circumstances in which reboots become 
| necessary but surely these are abnormal conditions which should be 
| repaired? Memory leaks in applications are cited as a justification for 
| reboots, as are zombie processes. Though I'm not right up on either of 
| these, I'm sure other less drastic measures could be taken to repair a 
| leaky machine than to bounce it.

Memory leaks should, in theory, be cleaned up by the kernel when the
process exits.  If the application must run constantly, maybe scheduling
a restart of that application is warranted (I believe apache uses a
similar approach: a child process is only trusted to handle X jobs
before it is killed and restarted).

Zombie processes often won't be cleaned up by the kernel - they have to
be handled by the calling process.  Situations with lingering zombies
only occur when a programmer errs, so they definately should be fixed.

Other than that, I'd reset unix boxes when the C time() integer clicks
over in 2040 or something. ;0

Now, NT boxes I'd schedule a regular reset for.  Strange things can
happen on those beasts if you leave them running.  Maybe this is where
the practice started?  I know at a company I used to work for, the reset
button was the first (and often only) action taken if a system failed.
But on the other hand, their systems -are- still running, I suppose.

Paul

--
You are subscribed to the progsoc mailing list. To unsubscribe, send a
message containing "unsubscribe" to progsoc-request@nospam.progsoc.uts.edu.au.
If you are having trouble, ask owner-progsoc@nospam.progsoc.uts.edu.au for help.

This list is archived at <http://www.progsoc.uts.edu.au/lists/progsoc/>