UNCLASSIFIED - NO CUI

Skip to content

CRD Moving

Cody Williams requested to merge reno into main

update "chart" (https://github.com/prometheus-community/helm-charts) from "kube-prometheus-stack-48.2.3" (a9196e6c759914385ecf2695a46952c12000440b) to "kube-prometheus-stack-48.3.1" (1889d55cf74759ca69a061f19af3dc383020d6ea)

Cleaning up leftover CRDs and also upgrading the kube-prometheus-stack to current.

The CRDs during the last Renovate were moved from chart/crds to chart/charts/crd/crd, which transitions from one recommended Helm deployment method (chart/crds as a set folder for CRDs) to another (subchart with CRDs). During the KPT update, both crd forms were saved.

Current working theory is that chart/crds was run first, before subcharts. It began to set up CRDs while helm began loading subcharts. Spoiler: One of the subcharts contained the same exact CRDs. Due to bigbang .Values.monitoring.flux.upgrade.crds (also install.crds) being set to 'CreateReplace', chart/crds created as normal, but chart/charts/crds came in like a wrecking ball mid-deployment, seeing that those CRDs existed. It tried to delete them, but they were not fully deployed yet so the delete failed, causing these errors:

Last Helm logs:
apply CRDs with policy CreateReplace
failed to apply CRD no CustomResourceDefinition with the name "alertmanagerconfigs.monitoring.coreos.com" found

Deleting the old CRD definitions should resolve this issue when deployed with a bigbang chart.

Edited by Cody Williams

Merge request reports