Feature Request
Describe the problem you need a feature to resolve.
We are currently migrating an internal developer platform from a event driven architecture to a full kubernetes operator architecture.
With the first small operator, we faced the issue that the pod was evicted after 2 to 3 reconciliation runs by filling up the ephemeral storage.
After digging into the issue we found the ansible-operator folder in /tmp. We see, on this small operator, already a consumption of 0.5Mib per artifact. With a sum of 150 entities and a default assigned ephemeral storage of 128Mib we will not survive for long.
We already reduced the maxArtifacts to 1 and decreased the log output, but over time it will fill up heavily. Also we have other operators in the pipeline, which write much more logs in the future.
Additionally we see a lot of other files increasing:
/tmp/k8srcp-.json
/tmp/osrcp-.json
Describe the solution you'd like.
From our pov we see two options:
- Add an option to disable the artifacts completely
- Add an option for how many entities the artifacts should be kept in total, across all custom resources (e.g. the last 10, or even 0)
Feature Request
Describe the problem you need a feature to resolve.
We are currently migrating an internal developer platform from a event driven architecture to a full kubernetes operator architecture.
With the first small operator, we faced the issue that the pod was evicted after 2 to 3 reconciliation runs by filling up the ephemeral storage.
After digging into the issue we found the ansible-operator folder in /tmp. We see, on this small operator, already a consumption of 0.5Mib per artifact. With a sum of 150 entities and a default assigned ephemeral storage of 128Mib we will not survive for long.
We already reduced the maxArtifacts to 1 and decreased the log output, but over time it will fill up heavily. Also we have other operators in the pipeline, which write much more logs in the future.
Additionally we see a lot of other files increasing:
/tmp/k8srcp-.json
/tmp/osrcp-.json
Describe the solution you'd like.
From our pov we see two options: