SocketIO deployment enters crashloopbackoff when using external redis deployment

[bruxillensis]

Description of the issue

Helm chart socketio deployment crashes when trying to use an existing redis deployment

Context information (for bug reports)

Steps to reproduce the issue

1. Install MariaDB and Redis externally from erpnext
2. Install ERPnext using the helm chart and the following values for Redis:

redisQueueHost: "redis-master.default.svc.cluster.local"
redisCacheHost: "redis-master.default.svc.cluster.local"
redisSocketIOHost: "redis-master.default.svc.cluster.local"

Observed result

SocketIO deployment enters crash loop backoff because it still expects the following host: erpnext-redis-socketio not the one specified in the helm chart. This is also the case if the chart is named anything other than "erpnext".

Expected result

Chart deploys successfully

Stacktrace / full error message if available

listening on *: 9000
node:events:356
      throw er; // Unhandled 'error' event
      ^

Error: getaddrinfo ENOTFOUND erpnext-redis-socketio
    at GetAddrInfoReqWrap.onlookup [as oncomplete] (node:dns:69:26)
Emitted 'error' event on RedisClient instance at:
    at RedisClient.on_error (/home/frappe/frappe-bench/apps/frappe/node_modules/redis/index.js:406:14)
    at Socket.<anonymous> (/home/frappe/frappe-bench/apps/frappe/node_modules/redis/index.js:279:14)
    at Socket.emit (node:events:379:20)
    at emitErrorNT (node:internal/streams/destroy:188:8)
    at emitErrorCloseNT (node:internal/streams/destroy:153:3)
    at processTicksAndRejections (node:internal/process/task_queues:81:21) {
  errno: -3008,
  code: 'ENOTFOUND',
  syscall: 'getaddrinfo',
  hostname: 'erpnext-redis-socketio'
}

[revant]

I am able to use separate Elasticache in one of the clusters.
On another cluster I've setup separate redis server manually and I can access it by private ip.

Check networking of the external helm chart, does it need extra config to make it available for access?

[bruxillensis]

I believe I have solved it. I did not realize that the env variables are saved on the first deployment and not used afterward (documented in the docker images). This would be helpful to document in this repo. I was using an AWS EFS storageclass (to achieve ReadWriteMany on sites_dir) that persisted after a helm uninstall. Originally I installed the helm chart as default (with the redis deployments) and only afterward switched to an external Redis deployment.