[ovs-discuss] Device-Port-Contention issue

[Gerhard Muntingh]

Hi everyone,

We experienced an openvswitch issue on our redhat openstack
infrastructure. Since it's rather exotic, and we haven't been
able to reproduce it, I'm sharing it here.

After some high CPU, openvswitch got into some sort of
device-port-contention looking like this:

  2017-11-01T02:35:03.742Z|1036196|bridge|INFO|bridge br-int: added interface tap6ec8832c-4a on port 64
  2017-11-01T02:35:03.828Z|1036197|bridge|INFO|bridge br-int: added interface tap87e621ea-f4 on port 92
  2017-11-01T02:35:03.853Z|1036198|bridge|INFO|bridge br-int: added interface tap2f90dc94-d3 on port 64
  2017-11-01T02:35:03.913Z|1036199|bridge|INFO|bridge br-int: added interface tap45874cb0-cd on port 92
  2017-11-01T02:35:03.936Z|1036200|bridge|INFO|bridge br-int: added interface tap6ec8832c-4a on port 64
  2017-11-01T02:35:03.939Z|1036201|netdev_linux|WARN|Dropped 271 log messages in last 12 seconds (most recently, 1 seconds ago) due to excessive rate
  2017-11-01T02:35:03.939Z|1036202|netdev_linux|WARN|tap2f90dc94-d3: removing policing failed: No such device
  2017-11-01T02:35:04.040Z|1036203|bridge|INFO|bridge br-int: added interface tap87e621ea-f4 on port 92
  2017-11-01T02:35:04.070Z|1036204|bridge|INFO|bridge br-int: added interface tap2f90dc94-d3 on port 64
  2017-11-01T02:35:04.131Z|1036205|bridge|INFO|bridge br-int: added interface tap45874cb0-cd on port 92

The rate of these messages increased when we drained the
machine. Maybe the main thread got more spare cpu cycles to
use for this after draining.

Restarting openvswitch resolved the issue.

Environment:
 - Redhat 7.4
 - Openstack: newton
 - openvswitch: 2.6.1.10.git20161206.el7fdp (redhat rpm)
 - No DPDK or SRV-IO


Kind regards,

Gerhard Muntingh.