Relation removal prevents normal operations
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Glance Charm |
Fix Released
|
High
|
James Page |
Bug Description
After having successfully used ceph-proxy for providing Ceph-backed Glance volumes and Ceph-backed Nova disks I decided to remove the Glance portion so as to use local Glance storage. I discovered that I could not do so. A remnant of the connection to the Ceph cluster appears to have been left behind after the removal of the relation.
The removed relation:
juju remove-relation ceph-proxy:client glance:ceph
The post-removal `juju status` output:
See attachment 'juju-status-
The Glance error:
See attachment 'post-ceph-
I eventually removed the ceph-proxy application but it didn't help. The full `juju status` output in the status attachment was made after ceph-proxy removal.
Changed in charm-glance: | |
milestone: | 20.08 → none |
Changed in charm-glance: | |
milestone: | none → 20.10 |
Changed in charm-glance: | |
status: | Fix Committed → Fix Released |
The issue is that the glance charm does not restart it services when the relation to ceph-* is broken:
@hooks. hook('ceph- relation- broken' ) keyring( service= service) write_all( )
def ceph_broken():
service = service_name()
delete_
CONFIGS.
as a result glance is still trying to use ceph but the keyring and ceph.conf are missing/incomplete on disk.
Adding:
@restart_ on_change( restart_ map())
as a decorator for the function should resolve this.