Cri-o: Add metrics for image pull failures

Created on 28 May 2020  路  4Comments  路  Source: cri-o/cri-o

We should add metrics for image pull failures, so we can add alerts for situations like quay.io going down.

Off the top, we could start with:

  1. A registry is unreachable.
  2. Authentication error.

@mtrmac @vrothberg @saschagrunert Thoughts?

picture mrunalp

All 4 comments

Sounds good to me, we could also count the successful downloads. This way the user doesn鈥檛 have to calculate: pull operations - failures

saschagrunert picture saschagrunert on 28 May 2020
👍1

All proposals sounds like a reasonable addition :+1:

vrothberg picture vrothberg on 29 May 2020
👍1

Proposed a solution in #3809

saschagrunert picture saschagrunert on 29 May 2020
👍1

One more possible thing to consider is mirroring setup for CRI-O through registries.conf. We would want to bubble that up in the metrics so it is clear that mirrors are in use. This is especially important for failure metrics. The first step for that would be to return whether a mirror is in use for a particular image pull, so if there is an error, admin can check the mirror setup.

mrunalp picture mrunalp on 29 May 2020
👍1
Was this page helpful?
0 / 5 - 0 ratings

Related issues

cyphar picture cyphar  路  7Comments
baichinger picture baichinger  路  6Comments
kfox1111 picture kfox1111  路  4Comments
kubealex picture kubealex  路  9Comments
runcom picture runcom  路  3Comments