Failing K8s rabbitmq-peer-discovery-k8s clustering

Question

I'm trying to bring up a RabbitMQ cluster on Kubernetes using Rabbitmq-peer-discovery-k8s plugin and I always have only on pod running and ready but the next one always fails.

I tried multiple changes to my configuration and this is what got at least one pod running

---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: rabbitmq 
  namespace: namespace-dev
---
kind: Role
apiVersion: rbac.authorization.k8s.io/v1beta1
metadata:
  name: endpoint-reader
  namespace: namespace-dev
rules:
- apiGroups: [""]
  resources: ["endpoints"]
  verbs: ["get"]
---
kind: RoleBinding
apiVersion: rbac.authorization.k8s.io/v1beta1
metadata:
  name: endpoint-reader
  namespace: namespace-dev
subjects:
- kind: ServiceAccount
  name: rabbitmq
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: endpoint-reader
---
apiVersion: v1
kind: PersistentVolume
metadata:
  name: "rabbitmq-data"
  labels:
    name: "rabbitmq-data"
    release: "rabbitmq-data"
    namespace: "namespace-dev"
spec:
  capacity:
    storage: 5Gi
  accessModes:
  - "ReadWriteMany"
  nfs:
    path: "/path/to/nfs"
    server: "xx.xx.xx.xx"
  persistentVolumeReclaimPolicy: Retain

---  
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: "rabbitmq-data-claim"
  namespace: "namespace-dev"
spec:
  accessModes:
    - ReadWriteMany
  resources:  
    requests:
      storage: 5Gi
  selector:
    matchLabels:
      release: rabbitmq-data
---
# headless service Used to access pods using hostname
kind: Service
apiVersion: v1
metadata:
  name: rabbitmq-headless
  namespace: namespace-dev
spec:
  clusterIP: None
  # publishNotReadyAddresses, when set to true, indicates that DNS implementations must publish the notReadyAddresses of subsets for the Endpoints associated with the Service.     The default value is false. The primary use case for setting this field is to use a StatefulSet's Headless Service to propagate SRV records for its Pods without respect to     their readiness for purpose of peer discovery. This field will replace the service.alpha.kubernetes.io/tolerate-unready-endpoints when that annotation is deprecated and all clients have been converted to use this field.
  # Since access to the Pod using DNS requires Pod and Headless service to be started before launch, publishNotReadyAddresses is set to true to prevent readinessProbe from finding DNS when the service is not started.
  publishNotReadyAddresses: true 
  ports: 
   - name: amqp
     port: 5672
   - name: http
     port: 15672
  selector:
    app: rabbitmq
---
# Used to expose the dashboard to the external network
kind: Service
apiVersion: v1
metadata:
  namespace: namespace-dev
  name: rabbitmq-service
spec:
  type: NodePort
  ports:
   - name: http
     protocol: TCP
     port: 15672
     targetPort: 15672
     nodePort: 31672
   - name: amqp
     protocol: TCP
     port: 5672
     targetPort: 5672
     nodePort: 30672
  selector:
    app: rabbitmq
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: rabbitmq-config
  namespace: namespace-dev
data:
  enabled_plugins: |
      [rabbitmq_management,rabbitmq_peer_discovery_k8s].
  rabbitmq.conf: |
      cluster_formation.peer_discovery_backend  = rabbit_peer_discovery_k8s
      cluster_formation.k8s.host = kubernetes.default.svc.cluster.local
      cluster_formation.k8s.address_type = hostname
      cluster_formation.node_cleanup.interval = 10
      cluster_formation.node_cleanup.only_log_warning = true
      cluster_partition_handling = autoheal
      queue_master_locator=min-masters
      loopback_users.guest = false

      cluster_formation.randomized_startup_delay_range.min = 0
      cluster_formation.randomized_startup_delay_range.max = 2
      cluster_formation.k8s.service_name = rabbitmq-headless
      cluster_formation.k8s.hostname_suffix = .rabbitmq-headless.namespace-dev.svc.cluster.local
      vm_memory_high_watermark.absolute = 1.6GB
      disk_free_limit.absolute = 2GB

---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: rabbitmq
  namespace: rabbitmq
spec:
  serviceName: rabbitmq-headless   # Must be the same as the name of the headless service, used for hostname propagation access pod
  selector:
    matchLabels:
      app: rabbitmq # In apps/v1, it needs to be the same as .spec.template.metadata.label for hostname propagation access pods, but not in apps/v1beta
  replicas: 3
  template:
    metadata:
      labels:
        app: rabbitmq  # In apps/v1, the same as .spec.selector.matchLabels
      # setting podAntiAffinity
      annotations:
        scheduler.alpha.kubernetes.io/affinity: >
            {
              "podAntiAffinity": {
                "requiredDuringSchedulingIgnoredDuringExecution": [{
                  "labelSelector": {
                    "matchExpressions": [{
                      "key": "app",
                      "operator": "In",
                      "values": ["rabbitmq"]
                    }]
                  },
                  "topologyKey": "kubernetes.io/hostname"
                }]
              }
            }
    spec:
      serviceAccountName: rabbitmq
      terminationGracePeriodSeconds: 10
      containers:        
      - name: rabbitmq
        image: rabbitmq:3.7.10
        resources:
          limits:
            cpu: "0.5"
            memory: 2Gi
          requests:
            cpu: "0.3"
            memory: 2Gi
        volumeMounts:
          - name: config-volume
            mountPath: /etc/rabbitmq
          - name: rabbitmq-data
            mountPath: /var/lib/rabbitmq/mnesia
        ports:
          - name: http
            protocol: TCP
            containerPort: 15672
          - name: amqp
            protocol: TCP
            containerPort: 5672
        livenessProbe:
          exec:
            command: ["rabbitmqctl", "status"]
          initialDelaySeconds: 60
          periodSeconds: 60
          timeoutSeconds: 5
        readinessProbe:
          exec:
            command: ["rabbitmqctl", "status"]
          initialDelaySeconds: 20
          periodSeconds: 60
          timeoutSeconds: 5
        imagePullPolicy: IfNotPresent
        env:
          - name: HOSTNAME
            valueFrom:
              fieldRef:
                fieldPath: metadata.name
          - name: RABBITMQ_USE_LONGNAME
            value: "true"
          - name: RABBITMQ_NODENAME
            value: "rabbit@$(HOSTNAME).rabbitmq-headless.namespace-dev.svc.cluster.local"
          # If service_name is set in ConfigMap, there is no need to set it again here.
          # - name: K8S_SERVICE_NAME
          #   value: "rabbitmq-headless"
          - name: RABBITMQ_ERLANG_COOKIE
            value: "mycookie" 
      volumes:
        - name: config-volume
          configMap:
            name: rabbitmq-config
            items:
            - key: rabbitmq.conf
              path: rabbitmq.conf
            - key: enabled_plugins
              path: enabled_plugins
        - name: rabbitmq-data
          persistentVolumeClaim:
            claimName: rabbitmq-data-claim

I only get one pod running and ready instead of the 3 replicas

[admin@devsvr3 yaml]$ kubectl get pods
NAME                          READY   STATUS    RESTARTS   AGE
rabbitmq-0                    1/1     Running   0          2m2s
rabbitmq-1                    0/1     Running   1          43s

Inspecting the failing pod I got this.

[admin@devsvr3 yaml]$ kubectl logs rabbitmq-1

  ##  ##
  ##  ##      RabbitMQ 3.7.10. Copyright (C) 2007-2018 Pivotal Software, Inc.
  ##########  Licensed under the MPL.  See http://www.rabbitmq.com/
  ######  ##
  ##########  Logs: <stdout>

              Starting broker...
2019-02-06 21:09:03.303 [info] <0.211.0> 
 Starting RabbitMQ 3.7.10 on Erlang 21.2.3
 Copyright (C) 2007-2018 Pivotal Software, Inc.
 Licensed under the MPL.  See http://www.rabbitmq.com/
2019-02-06 21:09:03.315 [info] <0.211.0> 
 node           : rabbit@rabbitmq-1.rabbitmq-headless.namespace-dev.svc.cluster.local
 home dir       : /var/lib/rabbitmq
 config file(s) : /etc/rabbitmq/rabbitmq.conf
 cookie hash    : XhdCf8zpVJeJ0EHyaxszPg==
 log(s)         : <stdout>
 database dir   : /var/lib/rabbitmq/mnesia/rabbit@rabbitmq-1.rabbitmq-headless.namespace-dev.svc.cluster.local
2019-02-06 21:09:10.617 [error] <0.219.0> Unable to parse vm_memory_high_watermark value "1.6GB"
2019-02-06 21:09:10.617 [info] <0.219.0> Memory high watermark set to 103098 MiB (108106919116 bytes) of 257746 MiB (270267297792 bytes) total
2019-02-06 21:09:10.690 [info] <0.221.0> Enabling free disk space monitoring
2019-02-06 21:09:10.690 [info] <0.221.0> Disk free limit set to 2000MB
2019-02-06 21:09:10.698 [info] <0.224.0> Limiting to approx 1048476 file handles (943626 sockets)
2019-02-06 21:09:10.698 [info] <0.225.0> FHC read buffering:  OFF
2019-02-06 21:09:10.699 [info] <0.225.0> FHC write buffering: ON
2019-02-06 21:09:10.702 [info] <0.211.0> Node database directory at /var/lib/rabbitmq/mnesia/rabbit@rabbitmq-1.rabbitmq-headless.namespace-dev.svc.cluster.local is empty. Assuming we need to join an existing cluster or initialise from scratch...
2019-02-06 21:09:10.702 [info] <0.211.0> Configured peer discovery backend: rabbit_peer_discovery_k8s
2019-02-06 21:09:10.702 [info] <0.211.0> Will try to lock with peer discovery backend rabbit_peer_discovery_k8s
2019-02-06 21:09:10.702 [info] <0.211.0> Peer discovery backend does not support locking, falling back to randomized delay
2019-02-06 21:09:10.702 [info] <0.211.0> Peer discovery backend rabbit_peer_discovery_k8s does not support registration, skipping randomized startup delay.
2019-02-06 21:09:10.710 [info] <0.211.0> Failed to get nodes from k8s - {failed_connect,[{to_address,{"kubernetes.default.svc.cluster.local",443}},
                 {inet,[inet],nxdomain}]}
2019-02-06 21:09:10.711 [error] <0.210.0> CRASH REPORT Process <0.210.0> with 0 neighbours exited with reason: no case clause matching {error,"{failed_connect,[{to_address,{\"kubernetes.default.svc.cluster.local\",443}},\n                 {inet,[inet],nxdomain}]}"} in rabbit_mnesia:init_from_config/0 line 164 in application_master:init/4 line 138
2019-02-06 21:09:10.711 [info] <0.43.0> Application rabbit exited with reason: no case clause matching {error,"{failed_connect,[{to_address,{\"kubernetes.default.svc.cluster.local\",443}},\n                 {inet,[inet],nxdomain}]}"} in rabbit_mnesia:init_from_config/0 line 164
{"Kernel pid terminated",application_controller,"{application_start_failure,rabbit,{bad_return,{{rabbit,start,[normal,[]]},{'EXIT',{{case_clause,{error,\"{failed_connect,[{to_address,{\\"kubernetes.default.svc.cluster.local\\",443}},\n                 {inet,[inet],nxdomain}]}\"}},[{rabbit_mnesia,init_from_config,0,[{file,\"src/rabbit_mnesia.erl\"},{line,164}]},{rabbit_mnesia,init_with_lock,3,[{file,\"src/rabbit_mnesia.erl\"},{line,144}]},{rabbit_mnesia,init,0,[{file,\"src/rabbit_mnesia.erl\"},{line,111}]},{rabbit_boot_steps,'-run_step/2-lc$^1/1-1-',1,[{file,\"src/rabbit_boot_steps.erl\"},{line,49}]},{rabbit_boot_steps,run_step,2,[{file,\"src/rabbit_boot_steps.erl\"},{line,49}]},{rabbit_boot_steps,'-run_boot_steps/1-lc$^0/1-0-',1,[{file,\"src/rabbit_boot_steps.erl\"},{line,26}]},{rabbit_boot_steps,run_boot_steps,1,[{file,\"src/rabbit_boot_steps.erl\"},{line,26}]},{rabbit,start,2,[{file,\"src/rabbit.erl\"},{line,815}]}]}}}}}"}
Kernel pid terminated (application_controller) ({application_start_failure,rabbit,{bad_return,{{rabbit,start,[normal,[]]},{'EXIT',{{case_clause,{error,"{failed_connect,[{to_address,{\"kubernetes.defau

Crash dump is being written to: /var/log/rabbitmq/erl_crash.dump...done
[admin@devsvr3 yaml]$ 

What did I do wrong here?

Answer 1

finally i fixed it by adding this in /etc/resolv.conf of my pods:

[my-rabbit-svc].[my-rabbitmq-namespace].svc.[cluster-name]

to add this in my pod i used this setting in my StatefulSet:

dnsConfig:
    searches:
      - [my-rabbit-svc].[my-rabbitmq-namespace].svc.[cluster-name]

full documentation here