Add template and scaling logic#6414
Add template and scaling logic#6414vofish wants to merge 2 commits intoGoogleCloudPlatform:masterfrom
Conversation
vofish
commented
Jan 30, 2026
vofish
commented
- Add deployment template
- Add basic scale up and down logic
| kubernetes_scale_benchmark.Cleanup(bm_spec) | ||
| container_service.RunRetryableKubectlCommand( | ||
| ['delete', 'deployment', 'app'], | ||
| timeout=kubernetes_scale_benchmark._GetScaleTimeout(), |
There was a problem hiding this comment.
We could consider sharing the command, but we need both the timeout to be different (here you are sharing but that scales based off pod count) & the name ('app' instead of 'scaleup') so I'll leave to you whether you want to refactor into a function taking 2 arguments timeout & name or just duplicate.
| samples = kubernetes_scale_benchmark.ParseStatusChanges( | ||
| 'node', | ||
| start_time, | ||
| resources_to_ignore=set(), |
There was a problem hiding this comment.
Should the initial node(s) be ignored?
| ]) | ||
| kubernetes_commands.WaitForRollout( | ||
| 'deployment/app', | ||
| timeout=kubernetes_scale_benchmark._GetScaleTimeout(), |
There was a problem hiding this comment.
need to swap timeout usage since that scales with kubernetes_scale_benchmark flags. Possibly take a variable in that function like "number of nodes/pods" that defaults to kubernetes_scale_benchmark flags. Also remove the _ since you're making it non public. Otherwise, make a new function in this benchmark.
|
|
||
|
|
||
| def _WaitForScaledNodesDeletion(initial_node_count: int) -> bool: | ||
| timeout = 20 * 60 + kubernetes_scale_benchmark._GetScaleTimeout() |
There was a problem hiding this comment.
This seems to take a while huh? We should quiz Piotr (who's now in the chat) about this. As annoying as it would be, it's possible we're ok with waiting for hour(s).
| ) | ||
| _ScaleDeploymentReplicas(0) | ||
| if _WaitForScaledNodesDeletion(initial_node_count): | ||
| _ScaleDeploymentReplicas(NUM_NODES.value) |
There was a problem hiding this comment.
We need to also collect samples for the second deployment, and for the scale down. Add more ParseStatusChanges calls.
It also looks like that function already takes start_time - great! Expand that function to ignore timestamps older than the start time. Add a unit test to that effect as well.
| # Do one scale up, scale down, then scale up again. | ||
| return [] | ||
| _ScaleDeploymentReplicas(NUM_NODES.value) | ||
| samples = kubernetes_scale_benchmark.ParseStatusChanges( |
There was a problem hiding this comment.
Add some metadata to each set of samples (scale up 1, scale down, scale up 2) indicating which phase it was in. ie for these add metadata like "phase": "scaleup1".
2 participants
