[operator] Automatic updates are broken since 7.15.1

[ibuziuk]

Describe the bug

Reproduced against dogfooding instance on OSD v4 where automatic updates are enabled:

• che-operator version on the cluster - 7.16.2
• che version on the cluster - 7.15.0

Che version

• latest
• nightly
• other: please specify

Steps to reproduce

Expected behavior

both che and operator versions and identical

Runtime

• kubernetes (include output of kubectl version)
• Openshift / OSD v4 (include output of oc version)
• minikube (include output of minikube version and kubectl version)
• minishift (include output of minishift version and oc version)
• docker-desktop + K8S (include output of docker version and kubectl version)
• other: (please specify)

Screenshots

Logs

Not sure if related but there are multiple logs:

time="2020-08-04T09:45:39Z" level=error msg="Failed to detect if self-signed certificate used. Cause: routes.route.openshift.io \"test\" already exists"
time="2020-08-03T20:11:32Z" level=error msg="Failed to create test route 'test': routes.route.openshift.io \"test\" already exists"

might be broken due to test route creation:

	if len(endpointURL) < 1 {
		// Create test route to get certificates chain.
		// Note, it is not possible to use SyncRouteToCluster here as it may cause infinite reconcile loop.
		routeSpec, err := GetSpecRoute(instance, "test", "test", 8080, clusterAPI)
		if err != nil {
			return nil, err
		}

Installation method

• OperatorHub

[tolusha]

That's strange.
I've just checked updating from 7.15.0 to 7.16.2 and it works.
Could you just remove test route?
oc delete route test -n <namespace>

[ibuziuk]

@tolusha well it is definitely an issue with updates on OSD v4 Dogfooding instance - you can check it there. Also, the same issue was reported against OCP v4

[tolusha]

I see. This this the case when a real (not a self-signed) certificate is used.

[tolusha]

Meanwhile pls use the workaround to proceed the installation.

[tolusha]

Fixed by eclipse-che/che-operator@08c51a4

[ibuziuk]

@tolusha will wait for 7.17.0 and verify it against. Is there any impact on CRW 2.2 -> 2.3 update cc: @rhopp ?

[tolusha]

@ibuziuk
that's a good question.
it is better to apply the same fix for 7.16.x branch.

[alexeykazakov]

Meanwhile pls use the workaround to proceed the installation.

What is the workaround?

[tolusha]

The workaround is to delete test route
oc delete route test -n <namespace>

[alexeykazakov]

I've deleted the test route but the che server was not re-deployed. It's still running an old version.

[tolusha]

Scale down/up che-operator deployment.
Eclipse Che will start deploying from the beginning.