Bad start for a day!
Land at office in the morning only to see my X killed as my workstation had run out of swap space. A quick “top -o size” reveals no specific hoggers but spamd and its children were taking a lot of CPU. Hmm, mailq shows 3000+ mails in there.
Killed fetchmail before further damage happens and poked around in the mail logs. Spamd was taking a long time to do the network based tests for each of these mails. The 3000+ mails itself was a result of a script going bad. But the surprising part was the damage that spamd ended up doing to my box.
I am assuming it was spamd which ate up all the RAM and the FreeBSD VM decided to kill X for starters. Since the logs get rotated at 100KB, I had only a few hours worth of logs to rely on. They had nothing new to say about when the problem actually happened.
To prevent such things in the future, I have moved the spamc recipe in .procmailrc way down to the bottom and added addresses from which I receive a huge number of alerts to the whitelist in SpamAssassin user_prefs.
Have to now STFW to see how to increase Sendmail 8.13.x deliveries to use available spamd threads. During the queue bursts, I increased spamd count to 8 but I hardly saw 2 of them being in a run state at any point of time. The other 6 threads were in select state.
Shanker Balan
Shanker Balan is a devops and infrastructure freelancer with over 14 years of industry experience in large scale Internet systems. He is available for both short term and long term projects on contract. Please use the Contact Form for any enquiry.
Follow Me:
