re: vserver-stat output corrupted - increasing loadavg values

About this list Date view Thread view Subject view Author view Attachment view

From: Jacques Gelinas (jack_at_solucorp.qc.ca)
Date: Mon Feb 04 2002 - 11:22:53 EST


On Sat, 2 Feb 2002 16:15:52 -0500, Ron Arts wrote
> Hi,
>
> I'm running vserver (very enthusiastic about it BTW).
> I'm experiencing ever increasing loadavg values. It does not
> seem to do any harm, also the machine does not seem to be
> actually slowing down, very responsive, no diskactivity.

Indeed, vtop shows a very high load, yet, shows that both cpus
are pretty much doing nothing.

I have seen this sometime. This is probably a kernel bug. Basicaly, the loadavg
is computed from all the processes READY at a given time. In general, a process
is either Sleeping or Ready to run. For some reason, a process get in a Ready state
yet, is not really ready to run. It is locked somewhere in the kernel, in a driver
for example.

So the loadavg account for this process, but given it is not really ready, it
does not use any CPU at all.

To see this, using the /usr/sbin/vps command

        /usr/sbin/vps ax

You will see 46 lines with the status R instead of S. These are the offending
processes. Now why they are blocked this way, I can't tell. But knowing which
processes are stuck may help.

> vtop output:
>
> 3:54pm up 17:25, 1 user, load average: 48.83, 36.04, 17.78
> 168 processes: 167 sleeping, 1 running, 0 zombie, 0 stopped
> CPU0 states: 0.0% user, 0.0% system, 0.0% nice, 100.0% idle
> CPU1 states: 0.4% user, 6.2% system, 0.0% nice, 92.5% idle
> Mem: 384832K av, 377360K used, 7472K free, 0K shrd, 43440K buff
> Swap: 522072K av, 35492K used, 486580K free 45392K cached
>
> PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
> 16944 root 14 0 1132 1132 836 R 7.1 0.2 0:02 top
> 1 root 8 0 524 524 496 S 0.0 0.1 0:05 init

> Previously all was running well, but it started when I renamed 4 vservers
> (renamed top level directory, and /etv/ververs/ config filenames).
> I did not change anything else inside any vservers.

When you rename a vserver and it is running, you must also rename the
/var/run/vservers/name.ctx file. Unless you do so, the vserver script will
think the vserver is not running.

> After rebooting the machine the vserver-stat output was still corrupted,
> but the loadavg did not increase any more.

Are you using vserver 0.10. Some fixes were made to the vserver-stat utility
(guillaume ?)

---------------------------------------------------------
Jacques Gelinas <jack_at_solucorp.qc.ca>
vserver: run general purpose virtual servers on one box, full speed!
http://www.solucorp.qc.ca/miscprj/s_context.hc


About this list Date view Thread view Subject view Author view Attachment view

This archive was generated by hypermail 2.1.4 : Mon Aug 19 2002 - 12:01:00 EDT