-
Notifications
You must be signed in to change notification settings - Fork 42
Description
The following PR https://github.com/chainguard-dev/mono/pull/28340 disabled autoscaling for our GKE cluster, but that didn't have any effect on the GKE settings. I could still see on dev-eco the node auto-provisioning was enabled.
As a consequence I decided to comment out https://github.com/chainguard-dev/mono/pull/28347 the rest of the node autoprovisiong code but that did NOT have any effect either.
@k4leung4 pointed it out that the plan said there were no changes https://github.com/chainguard-dev/mono/actions/runs/18885466258/job/53899527858?pr=28347. However I could still see these auto provisioning limits when getting the info from GKE gcloud container clusters describe eco --location=us-central1 command:
autoscaling:
autoprovisioningNodePoolDefaults:
diskSizeGb: 200
diskType: pd-ssd
imageType: COS_CONTAINERD
oauthScopes:
- https://www.googleapis.com/auth/userinfo.email
- https://www.googleapis.com/auth/cloud-platform
serviceAccount: [email protected]
upgradeSettings:
strategy: SURGE
autoscalingProfile: OPTIMIZE_UTILIZATION
enableNodeAutoprovisioning: true
resourceLimits:
- maximum: "2000"
resourceType: cpu
- maximum: "8000"
resourceType: memory
There seems to be an issue on the gke terraform module, our changes to disable node auto provisioning were not applied to the GKE cluster (look at our resource limits). To fix that problem we had to manually change the settings on GKE to disable it.