On Thu, 30 Oct 2003, Mathew Thomas wrote:
> I have got squid running on HP ML530 with two Xeon 3.0 GHz, 8GB Mem, and
> 8x36 Gb hard disk. Six disks have been dedicated for Cache and using
> reiserfs filesystem and mounted as /cache1 to /cache6.
Total overkill server... unless ofcourse you plan on running 8 instances
of Squid on this box..
> Yesterday the system crashed. Other than ping response, it was not doing
> anything. I couldn't login into the console. I had to cold boot the
> server in order to bring up the server. I checked the logs, and I can't
> see anything unusual or wrong. Squid is working fine after the cold
> boot.
These kinds of symptoms is almost always a kernel or hardware issue.
I would recommend to always keep the console logged in (in text mode) and
disable the screen saver, or alternatively have a serial console always
connected. Also enable Sys-RQ support. Enabling kernel crash-dump support
may also be helpful (even if maybe not very practical with such large
amounts of memory).
> The server is not under heavy load, ( probably less than one tenth of
> the load we are planning put to the box) . I don't know where to start
> and look at the problem. Please help. I am attaching the output of the
> top command and my squid.conf file.
Neither really matters. You need to gather data on what was going in when
the system hang, not when it operates normally.
Regards
Henrik
Received on Wed Oct 29 2003 - 18:32:59 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:20:45 MST