On Fri, 2008-04-11 at 13:19 -0500, Matthew Rich wrote:
> Hello all,
>
> I have a 2-machine cluster of RHEL 5.1 servers (Dell PE 2850s) that
> serve up a whole bunch of web sites.

<snip>

> 3 times in the last two weeks, the main web server has crashed.

<snip>

> What happens is the "crashed" machine remains pingable but does not
> respond to any other requests, and at the console it only prints memory
> status messages every few seconds but doesn't allow logins (IE it looks
> like all processes are stopped).

<snip>

> Has anybody else seen anything like this? Any pointers on where to begin
> looking?

I would configure kdump on your servers. When a host can be pinged but
you cannot run interactive commands, you can manually crash the box and
analyze the vmcore (or if you have a support contract, Red Hat can help
you analyze it).

http://kbase.redhat.com/faq/FAQ_105_9036.shtm

Many systems can be configured such that if you send them a SysRq-C on
the console, they will crash.

/Brian/

--
    Brian Long                   |     |
                           . | | | . | | | .
                              '     '
                              C I S C O

_______________________________________________
rhelv5-list mailing list
rhelv5-list@(protected)
https://www.redhat.com/mailman/listinfo/rhelv5-list