I’ve been struggling trying to get the zeebe full helm chart installed onto Azure AKS.
I can leverage the dev profile (https://raw.githubusercontent.com/zeebe-io/zeebe-helm-profiles/master/zeebe-dev-profile.yaml) and it creates a single elasticsearch pod and zeebe pod. However, I noticed the operate pod is erroring trying to perform an index query against elasticsearch using the Http HEAD method. The elasticsearch responds with the fact that only POST is supported. The specific URL it tries to hit is “http://elasticsearch-master:9200/_index_template/operate_template?master_timeout=30s”) This smells of a security issue, but I haven’t found the right setting to tweak.to get this to work. The helm install is OOTB…nothing fancy other than placing it into a specific namespace. The URL issue shows up as a warning in the log…but I can’t seem to bring up the Operate UI so I assume it needs to get passed this issue to come up properly.
It turns out that using “zeebe-full-helm” as the helm chart does not set everything up correctly. When I used “zeebe-full” as the helm chart (ver 0.0.111 gets loaded), then everything got set up and the Operate service started working correctly. I cannot find the repository representing “zeebe-full”, but looking at the difference on what got laid down, it definitely looks like “zeebe-full-helm” was laying down zeebe 1.0.0 and operate 1.0.0. The older “zeebe-full” was laying down 0.24.2 of each image. The elasticsearch image version was consistently 6.8.5 for both charts.
btw, because we had to resort to the older zeebe engine (0.24.x) to get passed the problems running 1.0, the current versions of zbctl and camunda modeler software will not work with the older engine because all the Apis that had “Workflow” in them changed to “Process”. Heads up to anyone dealing with these same issues.
You could upgrade the imageTag in zeebe-dev-profile.yaml from 6.8.5 to 7.10.2, for instance. And then install with:
helm install [NAME] zeebe/zeebe-full-helm -f zeebe-dev-profile.yaml