mod_security memory usage

[breun]

We mostly do Plesk 9 with qmail at the moment, so I'm afraid I'm not of much help for that.

[premierhosting]

I was using qmail and not postfix, but drweb was always an offender. Mail system definitely the thing, and Plesk 10 on Centos 5.5 under Xen. Wow. Looks like it was the perfect storm.

[scott]

Thats good information, the other thing that I suspect is the mail-hander system since it is operating in a ram disk. That would be common across both qmail & postfix and as I understand it has changed between 9 and 10.

[premierhosting]

When I brought up a real live server, I first installed it as Plesk 10 for Business, which used postfix as the MTA. I then reimaged with the "regular" Plesk 10 and got back to familiar qmail. The real live server is setup almost identically, without any problems. I thought the ART kernel was the big difference because now I get to use it and before I couldn't, but mail in RAM disk makes a lot of sense. If the Xen isn't handling RAM correctly that could cause issues.

That explanation would explain why my logs showed processes hanging for a long time, which are problems caused either by a crappy kernel, which I tried 3 or 4 different non ART kernel versions from the Centos repos, OR caused by bad ram. Of course on a Cloud environment, how do you test for one stick of bad RAM? N'est pas possible. Best I could tell, this "bad ram" situation caused processes to back up so far that http requests, spamd requests, clamd requests, etc. would just pile up and bring the system down.

RAM disk mail handling + Xen + Centos 5 sounds like a likely culprit.