Operator continuously updates Kubernetes deployments

[avthart]

We are using Kyverno which validates every object on change.

The redis-operator is updating the rfs deployments overy 30 seconds, even when nothing is changed. This creates additional load on Kyverno, especially in environments with a lot of redisfailovers instances.

This can be verified with:

kubectl get deploy -w

Or check the number observedGeneration which is in our case 496690

It would be great if the rfs deployments would be updated only when needed.

Also see https://github.com/spotahome/redis-operator/issues/410#issuecomment-1276374527

[avthart]

Did a diff on two versions:

diff /tmp/1.yaml /tmp/2.yaml 
7c7
<   generation: 496690
---
>   generation: 496693
23c23
<   resourceVersion: "1087217544"
---
>   resourceVersion: "1087219466"
180c180
<   observedGeneration: 496690
---
>   observedGeneration: 496693

[nickvth]

+1

[raghu-nandan-bs]

@samof76 to address this, I can think of couple of improvements.

• using cached informers instead of making direct "get" calls to kubernetes API like here
• use a "hash" of redisfailover object as annotation, and apply the changes whenever this "hash" changes due to CR updates. This way, we dont have to call create / apply call every thirty seconds like here, but only when this resource hash changes. what are your thoughts on this?

[github-actions[bot]]

This issue is stale because it has been open for 45 days with no activity.

[github-actions[bot]]

This issue was closed because it has been inactive for 14 days since being marked as stale.

[avthart]

Please re-open ticket. This is still an issue.

[avthart]

Not stale. Reopen ticket.