Whenever a VM or project is deleted, we have hooks in designate-sink and keystone to delete any proxies associated with the VM or project.
To do that cleanup, services on cloudcontrol and cloudservices nodes need to be able to talk to the nova-proxy nodes on port 5668. If that port is blocked, the services will hang, timeout, and generally cause ugly races.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | • JHedden | T224981 rabbitmq: connectivity issues between cloudservices1004 and rabbitmq | |||
| Resolved | Krenair | T225168 New OpenStack control-plane nodes can't talk to novaproxy |
Already did the second service node: https://wikitech.wikimedia.org/w/index.php?title=Nova_Resource:Project-proxy/SAL&diff=1828608&oldid=1819110
Mentioned in SAL (#wikimedia-cloud) [2019-06-05T22:50:23Z] <Krenair> Added cloudcontrol1004 IP to match cloudcontrol1003 rule in 'proxy' security group rules for port 5668 T225168
Also did other cleanup of that list visible in https://wikitech.wikimedia.org/wiki/Nova_Resource:Project-proxy/SAL#2019-06-05
List is now:
