2

We are running APC as an opcode level cache, and have been getting kernel panics (related to shared memory segment). I thought that anything running in the user space (PHP, APC, Apache) could not cause a kernel panic. I have managed to replicate this issue on different hardware and environments. My questions are

  1. Why would APC cause a system-level kernel panic? (If there is shared segment error shouldn't the application just stop running?)
  2. Do you have any advice on how to fix or investigate this issue?

I am running FreeBSD.

TRiG
  • 10,148
  • 7
  • 57
  • 107
Yninja
  • 21
  • 1

2 Answers2

0

This PECL Bug #13474 has a comment from a user who says they resolved a similar problem by recompiling their APC extension with a different file locking type. They suggest "Spin locking".

They link to an issue that seems to relate to this problem on FreeBSD specifically, so perhaps there's something about using APC in that environment that's broken. The link is unfortunately expired.

What versions of PHP and APC are you using? Are you using the latest version? As of this writing (2010-09-14), the latest versions are PHP 5.3.3 and APC 3.1.4.

Bill Karwin
  • 538,548
  • 86
  • 673
  • 828
  • I was using 3.0.19 and have upgraded today to 3.1.4 but hopefully that will solve the problem Thanks a lot Bill! great help – Yninja Sep 15 '10 at 01:16
0

We were having similar panic/segfault issues with APC in our SaaS environment across multiple versions and distros (Debian Etch/Lenny, Ubuntu Lucid), and finally resolved them by switching to XCache. It's been well over a year without trouble.

Jeff Standen
  • 6,670
  • 1
  • 17
  • 18