Create an opcap debug mode report for failed or timed out operators

opdev / opcap

Apache License 2.0

3 stars 15 forks source link

Create an opcap debug mode report for failed or timed out operators #331

Closed acmenezes closed 1 year ago

acmenezes commented 1 year ago

While troubleshooting operator installation, generally speaking, engineers will follow some steps like the ones below:

check events
check CSV status
check installPlan status
check operator's pod status
check operator's CR status and conditions
check for messages in OLM package server

The proposal here is to create a report as rich as possible that can be enabled for a short list of operators that are either failing or timing out in order to determine the cause of failure and gather data to approach the partner developer.

acmenezes commented 1 year ago

Include deprecated APIs:

you may find errors on the controller logs.
check for RBAC permissions as well
check OCP kube api version and check release notes
check OCP api endpoints on release notes as well

madorn commented 1 year ago

@acmenezes: @yashoza19 and I were reviewing use of ApiRequestCount API.

It appears too unreliable due to the time it takes to populate with origin user account.

We were thinking we could tie into PR #334 and check installreport for KubeAPIWarningLogger. We did some testing and thsi should not only appear in controller logs when watching a deprecated endpoint, but will also appear on any CRUD function against API (i.e. after CR creation/operand install).