OpenVZ Forum


Home » General » Support » iowait/CPU Spikes in Recent Kernels
iowait/CPU Spikes in Recent Kernels [message #51998] Sat, 18 April 2015 03:29
Nick A is currently offline  Nick A
Messages: 1
Registered: April 2015
Junior Member
From: *members.linode.com
Since updating to kernel 2.6.32-042stab104.1, we have been seeing lots of issues on large, fairly busy host nodes with iowait randomly spiking and network connectivity dropping out. kswapd shows at the top of iotop, the load average goes above 900 in some cases, and SSH is lost. When we log in with IPMI, the processes are never the same, and it's never one specific container seeming to cause the problem. After about 3 minutes, the load starts dropping and connectivity returns.

This doesn't happen on all servers, even though some of them have the same number of containers and the same hardware - just ones with the most recent few kernels. We've been using the same hardware/setup (dual E5-2630v2s) for over a year and never had this problem until recently. It seems to get better for a week or so after rebooting, but eventually it starts happening again every 6-12 hours. It is still happening on 2.6.32-042stab106.4. Is anyone else experiencing this problem? It's happening on host nodes in different locations, and no one person has a container on each of them that are having problems. It doesn't seem to be specific to any given host node.
Previous Topic: How to resume migration of big ploop container?
Next Topic: OpenVZ yum mirrors down?
Goto Forum:
  


Current Time: Mon Dec 09 00:28:30 GMT 2019