Google Container Engine（GKE-Kubernetes）上的Redis集群

Question

I have set-up redis-cluster on Kubernetes (Google Container Engine). Referenced from : https://github.com/zuxqoj/kubernetes-redis-cluster/blob/master/README-using-statefulset.md

So the current state is that, we have 6 pods under a single node of Kubernetes Cluster, 3 pods as masters and 3 pods as slaves. Have brought all the pods into redis-cluster and things are working fine.

However, when I try to bring down one of the master / slave pod, kubernetes automatically starts a new pod with a new IP. And the new IP is being auto updated in all other pods node configuration, except its own node configuration. Is there something I am missing?

I am using redis 4.0.0

Redis Cluster Yaml: https://drive.google.com/open?id=1oSQzYu-pAJmehaAfU_HdIa3qAJBx1n5C

Answer 1

Try using DNS hostnames instead of IP addresses.

See Helm charts that just do that:

• https://github.com/kubernetes/charts/tree/master/stable/redis
• https://github.com/kubernetes/charts/tree/master/stable/redis-aha

Answer 2

You need to update the IP of the new pod in nodes.conf. You can do it by running this script in the initialization of the pod.

#!/bin/sh

set -e

REDIS_NODES_FILE="/data/nodes.conf"

if [ -f ${REDIS_NODES_FILE} ]; then
  if [ -z "${POD_IP}" ]; then
    echo "Unable to determine Pod IP address!"
    exit 1
  fi
  echo "Updating my IP to ${POD_IP}"
  sed -i.bak -e "/myself/ s/[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}/${POD_IP}/" ${REDIS_NODES_FILE}
fi

Originated from: https://github.com/antirez/redis/issues/5417