[Bug] 1.14.2: captain-captain crashes on startup ("not 0 instances of captain-certbot"), causing all three system services to restart in a loop

Bug Report: captain-captain crashes on startup with "not 0 instances" error, causing all three services to restart in a loop

Environment

• CapRover version: 1.14.2
• Docker version: 29.6.1
• Docker API version: 1.55 (minimum supported: 1.40)
• OS: Linux (x86_64)

What happened?

After upgrading to CapRover 1.14.2, the captain-captain service crashes immediately on startup with the following error:

Error: There must be only one instance (not 0) of the service running to find node id. captain-certbot
    at /usr/src/app/built/docker/DockerApi.js:108:23
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)

This causes a cascade restart loop across all three CapRover system services:

• captain-captain — crashes with exit code 1
• captain-nginx — force-killed and restarted
• captain-certbot — reports "No such container" and restarts

The cycle repeats approximately every 90 seconds, indefinitely.

Root cause (observed)

During startup, captain-captain queries the node ID of captain-certbot. However, since captain-certbot has not yet fully started (0 running instances at that moment), the check throws an unhandled exception and the process exits with code 1.

This appears to be a race condition introduced in 1.14.2 — the startup sequence does not wait or retry before querying captain-certbot's node ID.

Relevant code path: DockerApi.js:108

Dockerd logs showing the crash loop

level=error msg="fatal task error" error="task: non-zero exit (1)" service.id=<captain-captain-id>
level=info  msg="image pulled" remote="docker.io/caprover/caprover:1.14.2"   ← immediate re-pull
level=error msg="fatal task error" error="task: non-zero exit (1)"           ← crash again

Additional notes

• Also observed a secondary warning in logs: client version 1.38 is too old — this comes from the Docker API version probe in DockerApi.ts which uses v1.38 as a low-version probe. This is a pre-existing cosmetic issue and does not cause the crash, but worth noting alongside this ticket.
• The same setup running CapRover 1.14.1 on Docker 29.3.0 (API 1.54) works without any issues.