Description
Feature and motivation
There was a discussion around "programmatic solution for killing orphaned sessions after a test suite runs"
The orphaned session could be understood as
the session gets left over by the test suite where driver.quit(?) is not called. as we're testing the implementation, I believe there are instances where too many sessions get stuck in the queue and not quite enough nodes to handle them...
the queue eventually clears but by the end of a 3 hour test (about 600 cases), i sometimes see one or two running for days"
And there was solution with a script and schedule run in a pipeline, something like
we are still kind of benchmarking, in a lab environment i created a batch script that will curl for the status of all nodes then kill any sessions that have been running for too long (15 mins for now).
it runs in the pipeline and uses azure cli to delete the corresponding pod in kubernetes.. since I am using autoscaling a new pod wil automatically take its place
Instead of the pipeline, in Kubernetes has a component CronJob to schedule a task run itself based on cron we set.
Add to chart template for CronJob, which executes a shell script implemented, then everyone can simplify inserting their custom scripts for different tasks there.
Usage example
https://kubernetes.io/docs/concepts/workloads/controllers/cron-jobs/