Hosting device is never revived from DEAD state if cfg agent is restarted
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
networking-cisco |
Fix Released
|
Undecided
|
Unassigned |
Bug Description
When a device enters a DEAD state it never gets revived if the cfg agent is restarted. The following is observed:
When a device becomes unreachable (via ping) its put into an unknown/
After a period (default 300s) if it remains unreachable its put into a DEAD state
At this point all routers on the device are unscheduled in Neutron.
Subsequent checks then create a list of backlogged devices, which are checked for reachability. Including code to revive dead devices.
AT THIS POINT RESTART THE CFG AGENT
The backlogged devices are determined from the routers JSON blob/dict passed from the driver, but because all the routers have been unscheduled the DEAD device is never again known to the agent and so can never be revived.
summary: |
- Hosting device is never revived from DEAD state + Hosting device is never revived from DEAD state if cfg agent is + restarted |
Reviewed: https:/ /review. openstack. org/511229 /git.openstack. org/cgit/ openstack/ networking- cisco/commit/ ?id=ae72f46ef82 e5329c708a5e9ed eabcd8eec48ba2
Committed: https:/
Submitter: Zuul
Branch: master
commit ae72f46ef82e532 9c708a5e9edeabc d8eec48ba2
Author: Bob Melander <email address hidden>
Date: Fri Sep 15 02:20:14 2017 +0200
Make cfg agent fetch hosting devices to manage during startup
This patch makes changes to the cfg agent so that when it starts it
requests information from the l3 service plugin about the hosting
devices it should manage.
Closes-bug: #1717913
Change-Id: I399c69cc3be9bb 781f15877552cd8 426d1cd4e53