[Openstack-operators] [neutron] openvswitch-agent spins up too many /bin/ovsdb-client processes

Rahul Sharma rahulsharmaait at gmail.com
Mon Mar 14 22:32:28 UTC 2016


Hi All,

We are trying to debug an issue with our production environment. We are
seeing neutron-openvswitch-agent starts failing after some time (1-2 days).
After debugging, we found that there are large number of entries for the
ovsdb-client. On some nodes, it crosses more than 330 processes and then
ovsdb process starts failing.

   1. root     30689     1  0 00:37 ?        00:00:00 /bin/ovsdb-client
   monitor Interface name,ofport --format=json
   2. root     30804     1  0 00:38 ?        00:00:00 /bin/ovsdb-client
   monitor Interface name,ofport --format=json
   3. root     30909     1  0 00:38 ?        00:00:00 /bin/ovsdb-client
   monitor Interface name,ofport --format=json


Pastebin link for the processes: http://pastebin.com/QGQC0Jrt
Pastebin link with openvswitch starting all of them:
http://pastebin.com/repHMkHu

In logs, we start getting errors as:-
Mar 14 05:41:29 node2 ovs-vsctl: ovs|00001|fatal_signal|WARN|terminating
with signal 14 (Alarm clock)
Mar 14 05:41:39 node2 ovs-vsctl: ovs|00001|fatal_signal|WARN|terminating
with signal 14 (Alarm clock)
Mar 14 05:41:49 node2 ovs-vsctl: ovs|00001|fatal_signal|WARN|terminating
with signal 14 (Alarm clock)
Mar 14 05:49:30 node2 ovs-vsctl:
ovs|00001|vsctl|ERR|unix:/var/run/openvswitch/db.sock: database connection
failed (Protocol error)
Mar 14 05:49:32 node2 ovs-vsctl:
ovs|00001|vsctl|ERR|unix:/var/run/openvswitch/db.sock: database connection
failed (Protocol error)
Mar 14 05:49:34 node2 ovs-vsctl:
ovs|00001|vsctl|ERR|unix:/var/run/openvswitch/db.sock: database connection
failed (Protocol error)

Openvswitch version:-
[root at node2 ~(openstack_admin)]# ovs-vsctl --version
ovs-vsctl (Open vSwitch) 2.4.0
Compiled Sep  4 2015 09:49:34
DB Schema 7.12.1

We have to restart openvswitch service everytime and that clears up all the
processes. We are trying to figure out why so many processes are getting
started by neutron-agent? Also, we found that if we restart the host's
networking, one new process for the /bin/ovsdb-client starts. We checked
and found that we don't have any network fluctuations or any nic-flappings.
Are there any pointers where we should be looking into? It occurs on both
controller and compute nodes.

*Rahul Sharma*
*MS in Computer Science, 2016*
College of Computer and Information Science, Northeastern University
Mobile:  801-706-7860
Email: rahulsharmaait at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20160314/c3eea304/attachment.html>


More information about the OpenStack-operators mailing list