1

After an upgrade of our Apache server from Debian Wheezy to Jessie the load average has doubled and context switches have tripled. User and system cpu have remained more or less constant.

The server only runs Apache (mpm_prefork). Configuration has been carefully copied across to the new configuration layout.

Significant changes are:

  • Linux 3.2 to 3.16
  • Apache 2.2 to 2.4.10
  • PHP 5.4 to 5.6

I've read the upgrade documentation for all 3 products but nothing seems to explain the increase in load. The number of apache processes hasn't changed. MPM configuration is the same. I'm confident we are still running prefork.

Much of the context switching seems to be coming from the rcd_sched process.

I've spent days trying to find the cause and would appreciate any advice.

Update 7/7/17

Rebooted onto the 3.2 kernel and ran service for 2 days but no significant drop in load average or context switching.

  • Do the load and the context switching drop to normal values if you stop Apache? – Ferenc Wágner Jul 7 '17 at 6:39
  • Yes. After upgrade peak context switching is ~2000/s, if I remove the load it drops to ~250/s and if I stop apache it drops again to ~100/s. Prior to upgrade peek was ~600/s and no load was ~200/s. – Dave Billington Jul 7 '17 at 11:11
0

Upgrading to Debian Stretch returned both the load average and context switching to normal levels.

  • Linux 4.9.30
  • Apache 2.4.25

Unfortunately I can't tell if it was the kernel or apache upgrade that fixed the issues but it's definitely back to normal.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.