@nicolastakashi this PR would introduce a breaking change so might be tricky to get through as is. Out of interest what is your use case for using the VPA for a Deployment over the HPA?

There are currently discussions about introducing focused charts specifically for collection (DaemonSet) and aggregation (StatefulSet/Deployment) with strong SemVer guarantees. The changes you've suggested here are similar to what I was thinking of for the aggregator although I was also going to support using the addon-resizer when in StatefulSet mode in addition to the VPA (not the HPA with the current volume behaviour).

Hi, @stevehipwell when I changed it I thought that could be tricky because of the breaking change, but we can try a different approach to avoid this breaking change.

Regarding the use case for VPA for Deployments in my view, there are two:

1. VPA has a recommendation mode, that only provides recommendations about resource usage, it's nice if you don't want to automatically scale your resources or you just want to give insights to later change it manually.

2. You can combine VPA with HPA using a custom metric or keda, I don't know if fluent-bit is a metric that measures the amount of logs collected, if so the user can horizontally scale the number of replicas by this metric, while using VPA to scale CPU and Memory (I know this could not be useful on the fluent-bit use case)

@nicolastakashi if you're using a deployment (so no persistence) then I'm not convinced about the use cases other than to recommend on the resource settings. The other scenarios are better suited to the HPA especially as restarting the Fluent Bit pod could cause data loss so probably shouldn't be used in a feedback loop situation.

If Fluent Bit is a stateful set with persistence then I can see more value from the VPA as it couldn't use the HPA and would need scaling vertically.

If you really want to land this and you make the changes non-breaking we could add it.

@stevehipwell I'll change the implementation to avoid the breaking change 😄

@stevehipwell can you take a second look pls?

@stevehipwell can you check if I did it properly?

@nicolastakashi it looks like you've lost your code from the commit?

Hey @stevehipwell yeah I made a mistake but I was able to recovery hehe.
Could you take a look again?

@nicolastakashi have you tested the functionality you've added locally?

@stevehipwell yup, just to evidence.

@nicolastakashi assuming the CI passes are you happy for this to be released?

@stevehipwell yeah, let's party 🥳 😄