[plug] Unstable (not as in distribution) Debian Box

[James Devenish]

In message <1063335672.1065.27.camel at vaio>
on Fri, Sep 12, 2003 at 11:01:12AM +0800, Ryan wrote:
> Often processes go zombie during the morning updatedb stuff after random
> amounts of uptime (normally 1-2 weeks+) and I end up with several
> (sometimes hundreds) of sh and find calls that fork a fair bit then go
> to bed.

Sounds like an I/O or interprocess communication problem?

> My uneducated guess - all the zombie things filled up lots of memory and
> swap filled up and it had troubles dealing with it?

A Unix zombie process has no physical manifestation, so it occupies no
memory at all (other than an entry in the OS' process table or
equivalent). However, perhaps the kernel gets confused nevertheless?

> Why would everything be falling in a heap in the first place though?

It might be best to diagnose this as soon as you realise it's happening.
For instance, what is the parent of all those zombie processes -- the
existence of zombie processes is the parent process(es)' fault (or a
kernel bug). Presumably the parent is cron? What happens if you run
updatedb by hand? Can you periodically save dumps from `ps` and
/proc/meminfo, for example?