10/14 ok, WHO is fukcing w/ sendmail? load average is 80+ and there are
100's of sendmail processes... same thing happend this morning at
8am EDT. Someone just add some recursive procmail rule or some
shit? This problem has been happening for a couple weeks but just
got alot worse. - rory
\_ also ps, is this normal?
$ mailq | wc -l
26070
\_ no one responded to this... isnt this a ridiculously high
number of messages to have queued on our system? w/ ~100
users and a reasonable load? - rory
\_ Yes, it is. Mail on soda ... is fucked. No one currently
working on soda has the sendmail fu to fix this issue -
apparently it has something to do with forged headers
and virus emails. It is not clear that the hardware plan
will fix the problem. My advice is to use !soda for all
important email from now on.
\_ I... want... an... upgrade... -the soda machine
\_ Faster hardware is not the answer to bad configurations or
poor programming. Unless you work for MS.
\_ Soda's sendmail is completely fux0red. I don't believe anyone
on current politburo knows how to unfux0r it. Sendmail is
a dark art that has not been passed on to the younger
generation, apparently. The new hardware will simply cause
the problem to expand to fill it.
\_ It's more that email in general is fux0red. The sendmail
flurries are because of spam storms. This is made worse
by rampant use of SA on a mainly user-centric box. Offload
this task onto another server more or less dedicated to
mail handling and this problem will be mollified.
\_ I emailed pburo about this a while ago (I am sure the
others have also). The response I got is that
'they are working on something.' I was under the
impression they had a plan to offload spam filtering
to scotch or another machine or something. I don't
think that's happening though. -- ilyas
\_ Yes, the plan is to move mail processing to scotch,
but that will not happen until we get a new scotch.
I am told that last time we tried moving over mail
processing to scotch (a Vectra), the poor machine
fell to its knees in no time. The upgrade plan, as
it stands involves not only getting new hardware for
both soda and scotch, but moving the mail load over.
I know people are going to bitch and moan, but mail
processing needs to get off soda. We're also going
to be adding email antivirus options (I think njh
has been poking around with ClamAV) to deal with the
plethora of attachment-rich virus email. - jvarga
\_ alright well what can we do about it? Dont alot more
people have root than who are in the politburo? Has
politburo come to by synonymous with "soda sys admins"?
Thank you for pointing that out _/
It was more of a question but thanks _/
for the snarky commentary
I know very little about sendmail... I'm guessing maybe
someone w/ access could look in the logs at when these
huge performance spikes are occurring? Is it a regular
time daily/weekly? Does it correspond w/ mail to a
\_ Rumor has it that it has something to do with virii and
"virii" is not a word. The word you were _/
looking for is "viruses"
spoofed Reply-To: headers that aren't being properly
dealt with, but I know nothing for sure.
certain user? Does a procmail problem seem like a
certain user? Does procmail problem seem like a
possible candidate? - rory
\_ We are busy testing NFS. Thanx! -politburo!
\_ NFS for mailspools? yeesh...
\_ On the subject of mail, weren't people supposed to use spamc
for spamassassin? I see a whole lot of "perl spamassassin" jobs.
A conservative antivirus thing would be good. Also, on the scotch
subject, is that "dual Xeon" choice just an imperial fiat? It
doesn't seem to be under any debate (hint: use AMD).
\_ No no, by all means, use Xeons. It will be so fun when Soda
melts a hole in the machine room wall.
\_ Uh, yeah. I own AMD stock and no Intel, but really. Let's
be rational about this. |