O Must Gather (omg) v2 (WORK IN PROGRESS)
With the previous (v1) implementation of omg, there were some major limitations. A major restructuring is being done to implement some advance features that are necessary, for example:
-
Multi directory support: Often times a single must-gather is not enough to capture all the components that are being investigated hence multiple must-gathers (or combination of must-gathes and oc-adm-inspect) are collected. omg needs to be able to look at them at once (using v1 omg we have to switch to these directories individually, which is very inconvenient).
-
Dynamic CRDs: The CRDs in OpenShift are huge in number and growing. In the current implementation of omg, every CRD needs to be "handled" manually for omg to recoginze it. A resource map to handle each CRD is hard to scale and fails to pick any new/undefined CRDs. In the newer implementation, CRDs will be automatically detected and recognized. This idea was orignally pitched by Robert Bost here.
-
API Groups: Handling API Groups for each CRD is important. For example to differentiate between
oc get config.operator
andoc get config.imageregistry
. -
Components Output: Provide an easy way to see outputs of other components like elasticsearch and ceph when component specific must-gathers are collected (ODF, Logging, CNV, etc.)
-
High Level Summary: A high level summary of a must-gather would be very useful. For exmaple
omg show summary
andomg show issues
. -
Console Logging: A proper console logging library loguru is being used that supports standard/debug output and can separate standard and error outputs.