Wrote a monitoring script which checks the load every 15 sec. It will kill the offending daemon if the load is higher, than 8. Let's see if it helps ...
Currently it looks very fast.
For the whole yesterday I got a "account suspended" message from this address. I thought at least a warning was said about the earlier discussion. But now I guess it wasn't a complaint but something technical really?
Yeah, it was the overload I mentioned in the blog and here. You got the message yesterday because the old IP was cached, the site was up on a new IP a day before.
Anyway, the script works. Still not sure what the actual problem was/is. Probably a hogged resource.
The script works (kills processes) quite effectively. I did not have to reboot the server since yesterday.
I do not understand what periodically causes the enormous load, but it's not the network traffic. Probably the shared resources. Most likely disk. During one of those blackouts when the load jumped to 30 the disk was 100% busy. Blocked "write" mysql queries?
Connection persistence was set to On in php.ini. Let's see if setting it to Off will help.
No, neither of the above helped. But! I hada closer look at httpd.conf and figure out that the settings that the hoster uses by default are more than strange. Changed. Waiting for the results now.
The blog is utterly broken. I do not exclude a DoS attack as a possibility, though.
Reinstalling WordPress ATM.
Hm. Have you looked at mysql Slow Queries log? Also, you can change mysql engine for some tables to MEMORY - it will make them fast, but not persistent across reboots. I think, that it can be done, for example, for forum sessions and shoutbox, or something along that.
And, in my experience, switching from Apache to Nginx with PHP in FastCGI mode can improve speed. But complexity of this setup may vary greatly.
It works now... I don't get it ...
