Re: Sendmail with high load averages.



On Nov 12, 10:01 pm, Bill Cole <b...@xxxxxxxxxxxxx> wrote:
In article <1194909354.433970.174...@xxxxxxxxxxxxxxxxxxxxxxxxxxx>,





dlhirsch <dlhir...@xxxxxxxxx> wrote:
On Nov 12, 4:21 pm, Res <r...@xxxxxxxxxx> wrote:
On Mon, 12 Nov 2007, dlhirsch wrote:

We recently have been having issues with sendmail. For some reason our
RedHat Linux machine has been spiking high load averages. The load
average can get as high as 40-45 but usually around 10-15. Not sure
where to start troubleshooting. The top command does not shed any
light on this issue. I have also looked at vmstat and iostat but
cannot tell what would be causing the spikes. Any suggestions/
questions would be greatly appreciated. Thanks!

This is better of on a redhat list, since they like to
(butcher|hack_the_crap_out_of|customize_beyond_belief) most things that go
into their distro.

Not to mention you've not told us any versions, how many processes etc..

--
Cheers
Res

I could have guessed what information would have been helpful but it
would have been only a guess. Asking for specifics like version
numbers is still kind of vague. The version of sendmail is
"sendmail-8.9.3-20" on RedHat 7.3.

That makes the top guess very simple: you no longer are in control of
this system, and whatever you can see about what is happening on it is
limited to what the functional owner of the system wants to allow you to
see.

A machine running that software has been sitting for many years with
security flaws that make it possible for the entire system to be taken
over remotely. That state is survivable for a short time on pure luck,
but a machine neglected like that for years with any exposure to the net
has almost certainly been found at least once by someone eager and able
to exploit those flaws. If that has happened, then the tools on the
system that you would normally use to analyze a normal performance
problem are very likely to have been replaced with programs of the same
names in the same places which are modified so as to hide the details of
what is really going on. You cannot trust what you are told by the
versions of top, ps, ls, vmstat, iostat, netstat, or anything else on
that host. A highly proficient sysadmin might be able to reclaim control
and salvage such a system from whoever has taken it over, but the fact
that this system is in this neglected state argues that it has not been
touched by even a minimally competent sysadmin in a long time.

Your best chance at this point is not to attempt to fix this system, but
to wipe it and build something fresh.

--
Now where did I hide that website...- Hide quoted text -

- Show quoted text -

To compound the issue we are running Spamassassin and Kaspersky
Anitvirus. I guess the best bet would be to start from scratch on a
different machine since the machine with the issue may be overloaded.
It is a web server, mail server, fax server (Hylafax), proxy server
(Squid), and many others. I wanted to basically get a second opinion
in case there was something simple. It appears that most emails are
from "User unknown" which seems to flood the server. Below is the
output from a script we wrote to count different processes.

Sendmail Server Stats for Tue Nov 13 08:36:18 CST 2007

output from the "w" command

8:36am up 3:50, 2 users, load average: 7.50, 6.45, 5.48
USER TTY FROM LOGIN@ IDLE JCPU PCPU WHAT
root pts/1 130.0.1.67 8:12am 5.00s 0.12s 0.12s -bash
root pts/2 130.0.1.54 8:12am 0.00s 0.20s 0.01s /bin/
sh /usr/bi

***********************************************************
ps -e | grep <process> | wc -l

The number of:
Sendmail processes running is: 745
Procmail processes running is: 166
Kavkeeper processes running is: 146
Spamd proccesses running is: 3
Spamassassin processes running is: 0

***********************************************************
ls -1 <dir> | wc -l

Number of files in the mqueue: 851
Number of Files in the Kav mqueue: 594


Not sure if this helps or not.
Thanks

.



Relevant Pages

  • Re: Monitor system
    ... about processes running on various network servers. ... > central server, and have each client navigate to an aspx page on the ...
    (microsoft.public.dotnet.framework.aspnet)
  • Re: SQL queries getting stuck
    ... Sounds like one or more of your other processes running on the server might ... > The place it gets stuck can be anywhere in the SP. ... >> Profiler, some sort of logging for the procedure. ...
    (microsoft.public.sqlserver.programming)
  • Re: Compromised Windows Server
    ... Internet for machines listening on tcp 139 and 445. ... machine I noticed the following processes running. ... You should be wary of the server being re-compromised. ... compromise the server in the first place. ...
    (Incidents)
  • Re: Zombie Jails - why dont they disappear?
    ... On Thu, 22 Feb 2007, 01:21+0800, LI Xin wrote: ... There are no processes running under any of them, ... server# ps -auxww | grep J ... It was discussed millon times already and there is at least one open ...
    (freebsd-current)
  • Re: Zombie Jails - why dont they disappear?
    ... On Thu, 22 Feb 2007, 01:21+0800, LI Xin wrote: ... There are no processes running under any of them, ... server# ps -auxww | grep J ... It was discussed millon times already and there is at least one open ...
    (freebsd-hackers)