Its always a good idea to capture a standard OpenShift Must Gather in addition to information below.
An improvement is pending for the OpenShift Must Gather to capture all OLM details: https://github.com/openshift/must-gather/pull/182/files.
The following command results in a very large amount of data and can be unwieldy for troubleshooting:
# oc adm inspect -A olm
If your issue is isolated to installing into a specific namespace try:
# oc adm inspect olm -n <NAMESPACE>
Troubleshooting a Hanging Installation
If an operator isn't installing, the first thing to do generally is look at the status of the various install objects. First look at
InstallPlan, and finally
ClusterServiceVersion. Review the
status section of each resource for errors or warnings.
It's also helpful to look at the logs of the
catalog-operator pod in the
openshift-operator-lifecycle-manager namespace. There should be a message indicating an error or warning that is preventing installation.
There's no general troubleshooting path for all types of failing installations but this article will be updated and link to other knowledge-base entries for specific types of failures. However, most solutions will start with looking at the various OLM resources like above.
Common Mistake with Manual Approval InstallPlans
If you have an
InstallPlan requiring manual approval to upgrade, this will cause all other
InstallPlans in the same Namespace or Project to also require manual approval. This behaviour is expected and exists so that all dependencies for installed or upgrading Operators can be checked properly.
Note that https://issues.redhat.com/browse/RFE-1282 was created to try and reflect this behaviour better in the OpenShift Web Console.