Closed joaocc closed 5 months ago
Helm 3.14.0 comes with Kubernetes 1.29 while Flux controller runtime is on Kubernetes 1.28, we need to migrate all controllers to 1.29 APIs and client-go before we can make a release with this. At best we can try create a RC build for helm-controller after we release Flux v2.2.3 this week https://github.com/fluxcd/flux2/issues/4556
Thanks. Anything that could help in reverting to a stable behaviour in lens helps :)
@joaocc you can use the controller image posted here to switch to Helm 3.14 https://github.com/fluxcd/helm-controller/pull/879
Thanks for the suggestion. We will take a look to the technique, and try to adopt to our deployment strategy (helm + custom bootstrapping from terraform), and point image tag to preview-0d77acf1
.
Thanks for the prompt response.
Hi. The workaround worked very well (sorry for not having provided feedback sooner). Thanks!
Shall we point the image to v0.37.4-helm3140
(as per https://github.com/fluxcd/helm-controller/pull/879) or it's better to wait for next full flux2 release?
Thanks
Hi, After a lot of pain trying to identify why a non-trivial chart with subcharts was resulting in strange values, we found out this issue https://github.com/helm/helm/pull/12480 and https://github.com/helm/helm/issues/12460. It seems that helm 3.13 changed the priority of parent/child values, which was then fixed in helm 3.14.0. We would kindly ask if this could be included in an emergency release of the controller (and then propagated to flux or get some indications how it can be deployed)? We believe this fix is critical, to fix the regression caused by 3.13. Thanks