OpenCTI fails to start with clean Redis instance

Description

If starting OpenCTI that has already been initialized and the redis instance has been replaced, thus missing stream.opencti, then opencti will fail to start with the following logs:

{"category":"APP","version":"4.4.0","level":"info","message":"[OPENCTI] Servers ready on port 8080","timestamp":"2021-04-16T15:19:58.526Z"}
node:internal/process/promises:245
          triggerUncaughtException(err, true /* fromPromise */);
          ^
ReplyError: ERR no such key
    at ue (/opt/opencti/build/index.js:1:85637)
    at i (/opt/opencti/build/index.js:1:86101)
    at /opt/opencti/build/index.js:1:86401
    at Object.start (/opt/opencti/build/index.js:1:86444)
    at processTicksAndRejections (node:internal/process/task_queues:94:5)
    at async xt (/opt/opencti/build/index.js:1:227086) {
  command: { name: 'XINFO', args: [ 'STREAM', 'stream.opencti' ] }
}

Environment

1. OS (where OpenCTI server runs): Kubernetes 1.18.16-gke.502
2. OpenCTI version: OpenCTI 4.4.0
3. OpenCTI client: frontend
4. Other environment details:

Reproducible Steps

Steps to create the smallest reproducible scenario:

1. Run a clean opencti environment (this works from the docker-compose environment too)
2. Ensure the environment is initialized, has a user, etc
3. Delete redis and its persistent storage
4. Create a new instance of redis
5. Restart opencti

Expected Output

I expected that anything in the current work queue of OpenCTI would be lost, but that it would re-initialize redis with a clean work queue

Actual Output

OpenCTI fails to read information about the stream stream.opencti and crashes.

Additional information

I understand this isn’t a desirable state to be in and it shouldn’t happen, but sometimes Redis crashes and there’s not a smooth process to recover from this.

[richard-julien]

If you had this problem after upgrade you have a problem in your install configuration. Data from redis must not be loss after a container upgrade/restart.

[mwatermolen]

Seems to be affecting docker version 4.50, when I spin up everything in Kubernetes … For visibility https://github.com/OpenCTI-Platform/docker/issues/26