Some check_http checks missing -S
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
charm-openstack-service-checks |
Incomplete
|
Medium
|
Unassigned |
Bug Description
The charm correctly surmised that 80+ checks of openstack services were https and thus required the -S argument but three services, barbican, designate, and gnocchi are failing because the endpoint is https but the check_http command doesn't have -S.
If I check the endpoint list from keystone, I see all of these services have https schema:
$ openstack endpoint list | grep -e barbican -e gnocchi -e designate
| 3142eda4d15d472
| 3a91c5ac34fb477
| 3c09135964aa48b
| 444cea388ab94ea
| a53cbb734c8042b
| b15be86e0b0c4a6
| c7e1a26378ab4d9
| d3fd5b8a8de4422
| de97d85443214ec
These services were configured with ssl from the beginning but perhaps there is a race condition where a non-https endpoint exists in keystone before it's correctly configured?
I fixed this by going into the debug-hooks environment (it was wanting to run the update-status hook) and clearing the openstack-
root@juju-
root@juju-
lxc
All snaps up to date.
/usr/lib/
warnings.
root@juju-
Is there a better way to regenerate the service checks from the current keystone endpoints?
description: | updated |
Changed in charm-openstack-service-checks: | |
importance: | Undecided → Medium |
Are you using the identity- notifications relation between openstack- service- checks and keystone? In theory, that should clear the openstack- service- checks. endpoints. configured flag upon an endpoint update and rerun configure_ nrpe_endpoints( ) function (pretty much exactly what you did in your manual intervention).