openstack
/
tripleo-image-elements
Code Issues Proposed changes

Enable automatic failover of l3 routers 

When l3-agent is not working on a node, neutron server will
automatically move its routers to another l3-agent. Neutron server
detects l3-agent outage after 1-2 minutes, neutron-l3-agent is now
monitored every 30s by pacemaker which will restart the service in case of
a failure (this should cover most of outages). If outage is longer
(e.g. networking issue) and routers are reassigned to another node,
l3-agent will clean up routers from local node when it restarts (or when
it can communicate with neutron server again).

This solution is not perfect and doesn't cover some edge situations
described here: https://bugs.launchpad.net/neutron/+bug/1174591
But I think this still improves current state of l3-agent HA.

This should not be needed once neutron l3 agent HA patch is merged
(I83f2a5d2af42164c42773b385ba7b00872eed54e).

Change-Id: Ifd8d73b3b1a5b13095e87683449aa4cab7579d3e
This commit is contained in:
Jan Provaznik 2014-09-12 09:43:38 +02:00
parent 42ffac22e1
commit dbc1634711
3 changed files with 15 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
elements/neutron-network-node/element-deps
View File

@ -2,4 +2,5 @@ neutron
neutron-dhcp-agent
neutron-openvswitch-agent
os-refresh-config
pacemaker
pkg-map

15
elements/neutron-network-node/os-refresh-config/post-configure.d/80-neutron-networking
View File

@ -1,8 +1,17 @@
#!/bin/bash
set -eux
os-svc-enable -n neutron-metadata-agent
os-svc-enable -n neutron-l3-agent
os-svc-restart -n neutron-metadata-agent
os-svc-restart -n neutron-l3-agent
# this ensures that neutron-l3-agent and neutron-metadata-agent services
# will be monitored by pacemaker on all nodes.
# Because allow_automatic_l3agent_failover is enabled, we want
# to make sure that if l3-agent service fails it's started again by
# pacemaker. If outage is longer and l3 routers are moved to other
# node meantime, l3-agent will clean up routers on local node when
# started again.
# TODO: it would be cleaner to do service restart above by pacemaker
# command, but it's difficult to restart service on single node only
pcmk-service-create -c -n neutron-metadata-agent
pcmk-service-create -c -n neutron-l3-agent

3
elements/neutron/os-apply-config/etc/neutron/neutron.conf
View File

@ -3,7 +3,6 @@
## It has been enabled for passthrough configuration support.
[DEFAULT]
{{#neutron.verbose}}
# Print more verbose output (set logging level to INFO instead
# of default WARNING level). (boolean value)
@ -15,6 +14,8 @@ verbose={{neutron.verbose}}
debug={{neutron.debug}}
{{/neutron.debug}}
allow_automatic_l3agent_failover = True
lock_path = /var/run/neutron/lock
auth_strategy = keystone