HACKING.md

⚠⚠⚠ THIS IS A LIVING DOCUMENT AND LIKELY TO CHANGE QUICKLY ⚠⚠⚠

Hacking on the CEO

Maintenance

Update Alerts

Alerts are stored in jsonnet/custom.libsonnet, you need to update the respective prometheusRules in there and then re-generate the manifest yamls via:

hack/generate.sh

CVO resource removal

Any resource in the manifests/ directory is managed by CVO. Removal of these resources requires code to be added that removes these resources frequently[1]. This must be maintained for 1 release then the code can be removed.

As per [2], a new annotation is added to resources, to be garbage collected by CVO. This approach is being used to remove duplicate etcd-grafana dashboard. See [3],[4]

[1] openshift#429

[2] https://github.com/openshift/enhancements/blob/master/enhancements/update/object-removal-manifest-annotation.md

[3] https://issues.redhat.com/browse/OCPBUGS-8044

[4] openshift#1016

Controller removal

Anytime we remove a controller we should also ensure that we cleanup any stale status conditions. Leaving these statuses to linger can result in degraded controllers that could block upgrade[1],[2].

example:

	staleConditions := staleconditions.NewRemoveStaleConditionsController(
		[]string{
			// IP migrator was removed in 4.5, however older clusters can still have this condition set and it might prevent upgrades
			"EtcdMemberIPMigratorDegraded",
		},
		operatorClient,
		controllerContext.EventRecorder,
	)

[1] https://bugzilla.redhat.com/show_bug.cgi?id=1851351

[2] openshift#517