action #112511
Updated by pdostal about 2 years ago
In Helm test, we are currently installing the charts without TTL, and there could be a situation where the job is cancelled or killed right after the chart deployment and there is no cleanup performed. This can create left-overs in our K8S environments in the Public Cloud providers.
There are 2 solutions
1) Add cleanup of K8S pods in Public Cloud Watcher (might be needed anyways)
2) Investigate how to inject TTL information to the deployment https://github.com/jnummelin/pod-ttl-operator
3) Create own helm chart which will run a container which will exit after defined period of time
Option 1) is preferable, but if we can solve any potential left overs with option 2), then there is no rush in adding this functionality to pcw and we should add that TTL to the helm install command/configuration.