loading codes to redis from pod definition file-CodePudding

I have kubernetes job defined as following:

apiVersion: batch/v1
kind: Job
metadata:
  name: redis-master
  namespace: sandbox
  labels:
    app: redis
spec:
  parallelism: 1
  ttlSecondsAfterFinished: 10
  template:
    metadata:
      labels:
        app: redis
    spec:
      containers:
        - name: master
          image: redis
          command: ["sh", "-c"]

          args:
          - redis-server
          - sleep 10
          - redis-cli RPUSH codes $(cat /data/codes)
          env:
            - name: MASTER
              value: "true"
          ports:
            - containerPort: 6379
          volumeMounts:
            - name: codes
              mountPath: /data/codes
              subPath: codes
              readOnly: true
      volumes:
        - name: codes
          configMap:
            name: codes
      restartPolicy: Never

when pod is up and running redis is empty:

/deploy/base$ kubectl exec -it -n box redis-master-mbhrk -- /bin/sh
# redis-cli lrange codes 0 -1
(empty array)
#

this pod should kill him self when the queue codes is empty.

also, i get an error in pod console

can't access tty; job control turned off

how can I load a codes to redis ???

CodePudding user response：

A container only runs one command. You've set that command to /bin/sh -c and passed it a command string to run (redis-server), along with two additional parameters which would be visible inside that command string as $0 and $1.

The first thing you need to do is split the Redis server proper into a separate container, probably managed by a StatefulSet (or perhaps a Deployment if persistence is disabled). You might use something like a prepackaged Helm chart to simplify deploying this. We'll assume this already exists:

apiVersion: v1
kind: Service
metadata:
  name: redis-server
...

Now your Job needs to only run the redis-cli RPUSH command. Because of the way you use a command substitution, you do need a shell here. However, since you're not running the Redis server here, you don't need various bits of machinery around that.

You could probably write the Job as:

apiVersion: batch/v1
kind: Job
metadata:
  name: load-codes
  namespace: sandbox
  # labels:
  #   app: redis
spec:
  template:
    # metadata:
    #   labels:
    #     app: redis
    spec:
      containers:
        - name: loader
          image: redis
          command: 
            - /bin/sh
            - -c
            - redis-cli -h redis-server RPUSH codes $(cat /data/codes)
          volumeMounts:
            - name: codes
              mountPath: /data
              readOnly: true
      volumes:
        - name: codes
          configMap:
            name: codes
      restartPolicy: OnFailure

You'll probably need to set the labels so that they're not picked up by the main Service (since the pod isn't running a Redis server). I've also set the restart policy to OnFailure: if the pod does fail for some reason, maybe because the Redis server isn't available yet, this will automatically restart it, but if it succeeds, the Job will be marked "completed".

CodePudding user response：

The issue was in

args:
  - redis-server
  - sleep 10

Command redis-server is command which brings redis server up. So next command can not be executed because redis-server is still running.