Hello,
I’m running a single-node InfluxDB 3 Enterprise deployment under a Home license (non-commercial, self-hosted via Docker) and have hit a deadlock I can’t resolve from the client side.
Setup:
- Single node, node-id: node0
- cluster-id: home-cluster
- Catalog UUID: 974b5bda-b5e7-4633-a004-bd431b75ed8b
- License type: home, 2 licensed cores
- object-store: file (local Docker volume)
What happened:
The node stopped ungracefully (the container was restarting in a crash loop; the exact original trigger is unclear, but it was not a clean shutdown). Since then, the catalog still shows node0 as “running,” so every restart attempt is rejected with:
“Started a node with --node-id node0 and --mode ingest that is still running in another process”
I followed your documented “Recover a crashed node” procedure, but it assumes a second live node can issue stop node/remove node against the crashed one’s registration. In a single-node deployment there is no second node to do this from, so I’m stuck.
As a workaround, I tried starting a second, differently-named node (node0-recovery) against the same data/catalog, intending to use it to mark node0 as stopped. That node does correctly find and load the real catalog (matching UUID above), but it now fails licensing validation instead:
“2 cores licensed, 2 cores in use on other nodes, but 2 were requested. If you would like to start this node with 2 cores, use --num-cores 0 or you can also increase the number of cores for your license by contacting sales@influxdata.com”
Note: --num-cores 0 is rejected outright (“Cannot specify less than 2 for --num-cores”), so the suggested workaround in that message isn’t actually usable, and 2 is always going to equal “already in use” since the stuck node0 is still holding both licensed cores.
What I’m hoping for:
- Guidance on safely clearing node0’s stale “running” registration from the catalog without a second live node to issue the command from, or
- A way to temporarily start a recovery node without hitting the core-count deadlock, so I can clear the original registration myself
My actual data appears fully intact throughout all of this — every attempt has correctly located the same catalog UUID above, nothing has been overwritten, and I have a full backup of the data directory taken before I started troubleshooting. This is purely a registration/licensing deadlock, not a data-loss situation, but it has taken my Grafana dashboards and monitoring offline.
Happy to provide full logs, docker inspect output, or anything else that would help. Thank you for your time.
