Long-winded rant post ahead, you have been warned.
So for those hosting with us you would have noticed an increase in the unexpected outages with 2 large ones today. Man what a pain this has become, but I think I about have it licked.
So a few months ago we started having to do 4 hour safe restarts of Apache because of the famous php5ts.dll error. Yes we host on Windows and we are proud of it!
So shortly after that in an effort to correct this problem we changed our Apache install from the http://www.apache.org version to the http://www.apachelounge.com version and we upgraded our PHP from 5.2.x to 5.3.1 which allowed us to use the VC9 Apache build. For a short time there was rejoicing as we had stability, which was quickly ended 4 days later with the return of the errors.
After much reading, tweaking and playing around I finally think I nailed it today after our first crash. Most of the material related to our error came from bad/mis-configured/outdated PHP extensions. In our case we were only running 4ish of them, 2 where MySQL based. Today I pulled the plug on the MySQLi extension and converted all our hosted products back to MySQL. Since then we were doing great until a few moments ago when I got a call we were down again.
After some looking it seems MySQL was hung after several scripts exceeded their 60 second run time, it was also interesting that ClaimAV was also performing an update at the same time the hangs occurred.
Acting quickly to restore service ClaimAV was shutdown (it only filters our mail server) and we defaulted back to AVG.
I am actively monitoring now to see how things go, but fingers crossed it should work.
Related posts:
Tags: Hosting