Janusgraph in Crashloop if helm chart release name is not exactly "weaviate"

[etiennedi]

I've noticed that the pod for janusgraph ends up in a crashloop if the helm chart release name is not exactly "weaviate".

Issue Description

I've tried calling the release "weaviate1", but this breaks the configuration. Investigating the configmap for janusgraph we can find the following hostnames:

• Cassandra: weaviate-cassandra
• Elasticsearch weaviate-elasticsearch-client

➜  weaviate git:(master) ✗ k get cm janusgraph-config -o yaml
apiVersion: v1
data:
  gremlin-server.yaml: |- #removed for better visibility
  janusgraph.properties: |-
    gremlin.graph=org.janusgraph.core.JanusGraphFactory
    gremlin.graph=org.janusgraph.core.JanusGraphFactory
    index.search.backend=elasticsearch
    index.search.directory=/db/searchindex
    index.search.elasticsearch.interface=REST_CLIENT
    storage.backend=cassandra
    storage.directory=/db/cassandra
    storage.hostname=weaviate-cassandra <------------------------
    storage.index.search.client-only=false
    storage.index.search.local-mode=false
    index.search.hostname=weaviate-elasticsearch-client <------------------------
kind: ConfigMap
metadata:
  creationTimestamp: 2019-01-09T12:02:37Z
  name: janusgraph-config
  namespace: billion1
  resourceVersion: "3375"
  selfLink: /api/v1/namespaces/billion1/configmaps/janusgraph-config
  uid: 74b0d432-1406-11e9-b5bf-42010a840fd1

However, if we take a look at the services, the actual hostnames are:

• Cassandra: weaviate1-cassandra
• Elasticsearch: weaviate1-elasticsearch-client

k get svc
NAME                                TYPE           CLUSTER-IP      EXTERNAL-IP     PORT(S)                                        AGE
weaviate                            LoadBalancer   10.39.255.96    104.155.19.75   80:32527/TCP                                   7m
weaviate1-cassandra                 ClusterIP      None            <none>          7000/TCP,7001/TCP,7199/TCP,9042/TCP,9160/TCP   7m
weaviate1-elasticsearch-client      ClusterIP      10.39.247.206   <none>          9200/TCP                                       7m
weaviate1-elasticsearch-discovery   ClusterIP      None            <none>          9300/TCP                                       7m
weaviate1-janusgraph                ClusterIP      10.39.253.116   <none>          8182/TCP                                       7m

Possible Solution

• Use helm templating to build the janusgraph config map, make sure release name is used in hostnames, instead of hard-coded "weaviate".