Apache2 + PHP 5.x + XCache + Load = segfaults? Not any more!
I have been working on client problem that causes their site to segfault every time any serious load hits. We have tried all kinds of fixes for this to no avail. What we have been doing to rectify the problem was to run the logwatcher script from Workhabit (The company formerly known as Firebright).
However, that was a mere bandaid and I could not figure out what was going on. After some testing in our new lab, I was able to replicate the problem which is always a postive first step. I stumbled upon some release notes for XCache that claimed that you don't need to disable auto_globals_jit for XCache any more. When I tried setting auto_globals_jit to off, it completely fixed the problems we were seeing in the lab. Hopefully this fix will be a permenant solution.
It sounds like this fix might work for Drupal with other opcode caches as well. According to the PEAR APC logs, it sounds like this might be fixed in APC, but this PHP directive does not add any benefit to a system like Drupal which uses the globals $_SERVER and $_ENV.
I will report back here if this fix does not continue to work for our client.
ha! very happy to find out
ha! very happy to find out that my 'voodoo' config settings have been shown to work on a repeatable bug.
For future readers, to be clear, I wasn't experiencing segfaults with Drupal at all; they only seemed to occur when the SugarCRM pages were loaded, although we are running a Drupal install on the same box. Of course, if someone is experiencing a similar white-page-of-death with Drupal I'd definitely give it a shot.
Curious to find out how you were able to replicate the problem?
ragaskar
Any further validation of
Any further validation of fixing the problem on segmentation fault with Drupal + eaccelerator by disable auto_globals_jit ?
Post new comment