q-agent-cleanup.py script should write consistent logs

[Eugene Nikanorov]

Logs of agent rescheduling script should be much more verbose.

It should report:
1. Found dead agents (agent type, ID, host, last heart beat time)
1.1. DEBUG: resources associated with the dead agent: networks or routers
2. Action to be taken about dead agent: will it be rescheduled and to which node (if it's known)
3. Output rescheduling attempt result for each node to which rescheduling was attempted.
4. Result of rescheduling: <type>-agent ID was moved successfully from node X to node Y
5. Awaiting rescheduled agent to be alive (from Neutron server)

In addition to that the following logs should be improved:
namespaces cleanup, deleted router interfaces (from namespaces)

[Eugene Nikanorov (enikanorov)]

I think this should be high to critical for 6.0

[Bogdan Dobrelya (bogdando)]

In juno, rescheduling was drastically improved, hence, we could get rid of our custom version of this script as well. Setting to medium because of that.

[Bogdan Dobrelya (bogdando)]

ps. rescheduling and cleanup.

[Eugene Nikanorov (enikanorov)]

In fact, there are not much changes to agent rescheduling in Juno, in fact the only patch that addresses some part of the problem is 'removing routers from dead l3 agents', which is not exactly what we want.
SO for now cleanup script makes sense.

[Tomasz 'Zen' Napierala (tzn)]

I need clarification on the status. I don't see any workaround, so why this is in "triaged" status? Should it be just confirmed for now?

[Bogdan Dobrelya (bogdando)]

It was set as Triaged due to the bug description provides all needed info, i.e. which new logs we want.

[Bogdan Dobrelya (bogdando)]

AFAIK, we do not target medium priority bugs for maintanance releases, so, I put the statuses for 5.x.x and 6.0.x as won't fix.
If the status will be high, please update the status to confirmed as well.

[Mike Scherbakov (mihgen)]

Bogdan, master is still open. It means that if you land the fix into master, it's gonna be 6.0, not 6.1.

[Vladimir Kuklin (vkuklin)]

initially this was related to a customer issue with dhcp agent, as we thought that the script was not providing enough logging information. nevertheless, it had nothing to do with the issues that had been discovered, so increasing of verbosity is not so important as script's logging is verbose enough for debugging.