Ticket #241 (assigned defect)

Opened 1 year ago

Last modified 2 months ago

php 5.2.1 - zend_mm_heap corrupted = huge cpu overload, apache crash

Reported by: naox Assigned to: bart (accepted)
Priority: blocker Milestone:
Component: eAccelerator Version: 0.9.5
Keywords: Cc:

Description

When you use eaccelerator 0.9.5 or eaccelerator-svn301 with a php 5.2.1 working as httpd module it causes huge cpu overload (like 100%*120 = 12000%) over time. For me it happened like every 60 minutes on one server and on another it didnt happened yet (server wasnt on much php usage). Other ppl on internet said that it happens like every 10 hours so it probably depends on usage on php... In httpd log there is:

zend_mm_heap corrupted zend_mm_heap corrupted zend_mm_heap corrupted

This only happens with php 5.2.1. Its not happening with 5.2.0. I'm just reporting that stuff thats all.

Change History

02/28/07 18:13:10 changed by JoHnY

  • priority changed from blocker to major.

I confirm that there are some issues with the latest svn301 version. I experienced the same unusual loads (although I was not aware until I've read this ticket that it was caused by eAccelerator, but then I realised it started exactly when I put the new version up)

Also, when I want to restart apache gracefuly (apachectl graceful), it segfaults and I have to manually kill all httpd processes and then restart apache. When I downgraded to svn282, everything is back to normal.

By the way, the segfault with graceful restart doesn't depend on apache version, I had svn301 on one server with apache 1.3.37 and the other one with 2.0.59 and both segfaulted on graceful restart.

02/28/07 18:15:14 changed by JoHnY

  • priority changed from major to blocker.

08/15/07 23:20:29 changed by bart

Ticket #245 seems to be a duplicate of this one.

08/15/07 23:23:52 changed by bart

  • owner changed from somebody to bart.
  • status changed from new to assigned.

I think it has something to do with using emalloc in a function which is called during initialization. I'll try to make a patch to test if that is causing the problem.

(in reply to: ↑ description ) 05/14/08 23:53:44 changed by terrysduncan

Try the patch I applied to tracker 232. We have no problems after adding this one line fix.