Description
Since the DevWorkspace operator adds webhooks for pods/exec, we need to stress-test the webhooks server (and to a lesser extent, the controller itself) to get a sense of what effect deploying this to a big cluster would have. This is crucial since every time the webhooks server crashes, all pods/exec requests are blocked until it comes back up.
- What memory/cpu limit is required by the webhooks server to allow it to handle a given amount of
pods/exec requests on a cluster?
- How is throughput affected when the webhooks server is under load? Are we delaying every
pods/exec significantly when we're under load?
Part of the difficulty in testing this is getting a reasonable estimate of how many requests we should actually expect.
Description
Since the DevWorkspace operator adds webhooks for
pods/exec, we need to stress-test the webhooks server (and to a lesser extent, the controller itself) to get a sense of what effect deploying this to a big cluster would have. This is crucial since every time the webhooks server crashes, allpods/execrequests are blocked until it comes back up.pods/execrequests on a cluster?pods/execsignificantly when we're under load?Part of the difficulty in testing this is getting a reasonable estimate of how many requests we should actually expect.