High CPU load on compute node caused by contrail-vrouter-agent process
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Juniper Openstack |
New
|
Undecided
|
Unassigned |
Bug Description
[Description]
CPU load spikes occurred in three episodes on a single day, but have not reoccurred since.
We received a notification from out monitoring system of high CPU load on one of our compute nodes, compute02. On investigation we found that 1 particular process was causing this, namely contrail-
Attached is a screen shot showing the load, the output from top during one of these events, the output from df -h, and the output from a ps ax|grep to check the process causing the problem.
We haven't been able to figure out why this is happening and the only error in the contrail-
Could anyone please assist in figuring out what's going on here?
Seems related to <https:/ /bugs.launchpad .net/juniperope nstack/ +bug/1400633> (contrail- vrouter- agent reporting "No space left on device").
While in the present case the high-load spikes went away, the concern is that a similar occurrence affecting all compute instances simultaneously would bring the environment to its knees, hence a keen desire to get to the bottom of this :-)