[one-users] 4.1.80: top pid limit exceeded with more than 20 virtual machines on one node

Michael Kutzner michael.kutzner at virtion.de
Mon Jul 22 00:24:06 PDT 2013


The method of getting the cpu usage of the VMs on the nodes
is problematic with more than 20 VMs on a node as top is called with the list of pids of the 
VMs. Unfortunately, the standard system top (in this case Ubuntu 12.04LTS) is limited to 20
pids - so the call throws an error when more than 20 pids are given.
This leads to an error state of the node.

I solved it by replacing the system top with our own version (setting
MONPIDMAX in top.h of package procpcs - Ubuntu 12.04LTS) - but there must be
a better way (which I did not find this morning due to limited time).

Is this issue already known? I did not find any bug report by now.

Versions: Ubuntu 12.04LTS, OpenNebula 4.1.80

Best, Michael

More information about the Users mailing list