-
Notifications
You must be signed in to change notification settings - Fork 16.8k
[stable/*] Helm 3 backwards-compatibility for community charts #19008
Comments
The chart maintainers for |
@scottrigby what is the namespace issue? |
@scottrigby RE |
@taylorsilva thanks, removed @ewbankkit Thanks, removed @daixiang0 Thanks for the question. There isn't a specific issue with charts creating namespaces to accomplish application-speciifc goals. I grepped/listed these initially because namespace creation has changed in Helm 3 and I wanted to review charts that created their own. I removed the two above, and checked We really need to look over these individually and more closely after the holidays 😄 |
Removed |
#19212 resolved this issue for |
@scottrigby is there any action for those deprecated charts? Mark as deprecated? For |
@daixiang0 there is not an action for this yet. I have reached out to @krancour about |
@scottrigby I responded to you in slack, but I'm putting my response here for the public record as well:
|
All of the Helm 3 issues with community charts identified so far have been resolved. If more issue categories, or individual chart issues, are identified please ping us on Slack and we'll re-open this. Closing for now. Thanks everyone 💯 🎉 |
Helm 3.0.0 has been released, with a 6 mo bug / 1 yr security deprecation timeline for helm v2.
This repo is intended for charts apiVersion
v1
(notv2
charts), because of the helm/charts deprecation timeilne. This git repo - the source for Helm 2stable
andincubator
repos - has a similar deprecation timeline ashelm
(follow details here: #16720).However, since Helm v3 is the default version of helm now, the stable/incubator
apiVersion
v1
charts should run with Helm 3 as well as Helm 2.CRDs
We'll need to support both
crd-install
helm hook for helm 2, as well ascrds
directory for helm 3 (https://helm.sh/docs/topics/charts/#custom-resource-definitions-crds). A quick grep forcrd-
turned up these charts. Let's look closer and refine this.incubator/cassandra-operator
([incubator/cassandra-operator] Introduce crds directory for compatibi… #19207)incubator/spark-operator
([incubator/sparkoperator] Introduce crds directory for compatibility with Helm v3 #19236)stable/ambassador
([stable/ambassador] helm3 support crds #19225)stable/contour
([stable/contour] manage crds with v3/v2 #19287)stable/etcd-operator
([stable/etcd-operator] Introduce crds directory for compatibility with Helm v3 #19209)stable/jaeger-operator
([stable/jaeger-operator] make compatible with helm3 #19212 [stable/jaeger-operator] fix for helm 3 #19402)stable/prometheus-operator
([stable/prometheus-operator] Introduce crds directory for compatibility with Helm v3 #18721)stable/velero
([stable/velero] Introduce crds directory for compatibility with Helm v3 #19042)Note #18721 uses a neat glob trick I'd like to see the other charts emulate 👍
Namespace creation
Quick grep forredacted several. See comments below:kind: Namespace
. Please comment to help revise (mitigate/add to) this list and I'll keep it updatedstable/magic-namespace
([stable/magic-namespace]: add deprecation notice #19643)Horizontal Pod Autoscaler requires condition
Helm 3's 3-way strategic merge strategy is a huge improvement, and is the correct way to go. However it uncovered that one use case for the incorrect behavior in Helm 2 actually had a nice outcome: HPA-scaled replicas were untouched on
helm upgrade
, whereas Helm 3 now sets the replicas back to the values/template-defined number of replicas.The problem was found by @naseemkullah and discussed with Helm maintainers in helm/helm#7090. The solution is:
Quick grep for
kind: HorizontalPodAutoscaler
. Please comment to help revise (mitigate/add to) this list and I'll keep it updated:incubator/haproxy-ingress
([incubator/haproxy-ingress]fix replica count conflict #19196)stable/ambassador
([stable/ambassador]fix replica count conflict #19204)stable/cloudserver
([stable/cloudserver] fix replica count conflict #19203)stable/contour
([stable/contour] fix replica count conflict #19202)stable/fluentd
([stable/fluentd] fix replica count conflict #19195)stable/neo4j
([stable/neo4j] fix replica count conflict #19200)stable/nginx-ingress
([stable/nginx-ingress] Ignore deployment template's replicas if hpa is enabled #19189)stable/spark
([stable/spark] fix replica count conflict #19199)stable/traefik
([stable/traefik]fix replica count conflict #19198)The text was updated successfully, but these errors were encountered: