v4.26.4 Only 2 dockers hosts shown at the same time

[eilko]

Describe the bug
In v4.25 4 docker hosts were added and they were shown all at the same time. In v4.26.4 there are only shown 2 docker hosts at the top and their information in the block below. The hosts shown are changing every several seconds. All four hosts are shown at a certain point in time, but never all at the same time.

To Reproduce
Steps to reproduce the behavior:

1. Go to 'Docker'
2. See only 2 docker hosts

Expected behavior
See all configured hosts at the same time

Environment:

• Pulse Version: v4.26.4
• Installation Type: ProxmoxVE LXC

Additional context
Also in the settings of the docker hosts only two are shown at the same time. Also here the 4 hosts are switching every x seconds

[rcourtman]

Found a bug that could explain the rotating hosts. The backend was treating the shared API token as the unique identifier for a Docker host. If several agents use the same token, each heartbeat would overwrite the previous host in memory, leaving only one visible per token. That matches what you described whenever multiple hosts share the same token.

The matching code now prefers each agent's own identifier and only falls back to the machine ID plus hostname combination, so sharing a token no longer causes entries to overwrite each other. This fix will be in the next release.

Please try the new build when it's available and let me know whether all four hosts stay visible. If the problem remains, grab the Docker logs around the "Broadcasting state" messages so I can see what the server believes it has in memory.

[eilko]

I think you have found the reason. I have reused the same API token for several of the hosts. I will try the new build when available

[dontpanic42]

I can confirm, having the same issue with docker hosts sharing a common token. It's only showing one host (the last one reporting metrics). The displayed host changes every time a new host reports back new metrics.

[rcourtman]

Commit aa0aa7d already fixes the rotating host records by ensuring reports match on the right identifiers. That removes the immediate overwrite bug people saw.

Commit ee630a2 adds a guardrail so the same API token cannot be used by two Docker agents. Pulse now rejects the extra agent and tells you which host already holds that credential.

Sharing tokens was letting multiple agents rely on one credential. As soon as one of those hosts was removed and its token revoked, the others silently lost access and started failing authentication. Enforcing one token per agent eliminates that silent breakage and keeps the audit data meaningful.

Migration is straightforward. Generate a fresh token for each Docker agent, update the agent config, and restart the container. No need to touch anything else.

These fixes will appear in the next release.