docker agent does not show up after agent reinstall

[njunghausz]

Describe the bug
I performed a new agent installation with generating a new token and clicked docker as well.
The agents didn't show up.

To Reproduce
Steps to reproduce the behavior:
Stopped the agent service.
Removed the old agent from UI.
Generated new token on UI.
Deployed the new agent successfully.

Expected behavior
Both host and docker agent show up.

Environment:

• Pulse Version: 4.34.0
• Installation Type:

curl -fsSL http://192.168.2.96:7655/install.sh | sudo bash -s -- --url http://192.168.2.96:7655 --token xxxxxxxxxxxxxxxxxxxxxxxxx --interval 30s --enable-docker

Additional context

Nov 28 12:28:55 docker1 pulse-agent[243297]: {"level":"info","currentVersion":"v4.34.0","availableVersion":"4.34.0","time":"2025-11-28T12:28:55+01:00","message":"New agent version available, performing self-update"}
Nov 28 12:28:55 docker1 pulse-agent[243297]: {"level":"info","version":"v4.34.0","pulse_url":"http://192.168.2.96:7655","host_agent":true,"docker_agent":true,"auto_update":true,"time":"2025-11-28T12:28:55+01:00","message":"Starting Pulse Unified Agent"}
Nov 28 12:28:55 docker1 pulse-agent[243297]: {"level":"info","time":"2025-11-28T12:28:55+01:00","message":"Host agent module started"}
Nov 28 12:28:55 docker1 pulse-agent[243297]: {"level":"info","component":"pulse-docker-agent","runtime":"docker","daemon_host":"unix:///var/run/docker.sock","version":"29.0.4","time":"2025-11-28T12:28:55+01:00","message":"Connected to container runtime"}
Nov 28 12:28:55 docker1 pulse-agent[243297]: {"level":"info","time":"2025-11-28T12:28:55+01:00","message":"Docker agent module started"}
Nov 28 12:28:56 docker1 pulse-agent[243297]: {"level":"warn","component":"pulse-docker-agent","hostID":"636085b6-e965-409f-b777-3e6d54ef2f7d","pulseURL":"http://192.168.2.96:7655","detail":"docker host "636085b6-e965-409f-b777-3e6d54ef2f7d" was removed at 2025-11-28T11:51:59+01:00 and cannot report again. Use Allow re-enroll in Settings -> Docker -> Removed hosts or rerun the installer with a docker:manage token to clear this block","time":"2025-11-28T12:28:56+01:00","message":"Pulse rejected docker report because this host was previously removed. Allow the host to re-enroll from the Pulse UI or rerun the installer with a docker:manage token."}

There's no such as 'removed hosts' in UI.

On server side I can see the docker1 is registered and reporting but not visible on UI:

root@pulse:~# journalctl -f | grep docker1
Nov 28 12:40:50 pulse pulse[2197]: {"level":"info","component":"pulse","guest":"docker1","cpu":1.38988607707259,"memory":18.276596069335938,"disk":7.665456938997278,"thresholds":{"poweredOffSeverity":"warning","cpu":{"trigger":80,"clear":75},"memory":{"trigger":85,"clear":80},"disk":{"trigger":90,"clear":85}},"time":"2025-11-28T12:40:50+01:00","message":"Checking guest thresholds"}
Nov 28 12:41:01 pulse pulse[2197]: {"level":"info","component":"pulse","guest":"docker1","cpu":2.19613304043368,"memory":18.274497985839844,"disk":7.665503272209991,"thresholds":{"poweredOffSeverity":"warning","cpu":{"trigger":80,"clear":75},"memory":{"trigger":85,"clear":80},"disk":{"trigger":90,"clear":85}},"time":"2025-11-28T12:41:01+01:00","message":"Checking guest thresholds"}
Nov 28 12:41:07 pulse pulse[2197]: {"level":"info","component":"pulse","guest":"docker1","cpu":2.19613304043368,"memory":18.274497985839844,"disk":7.665503272209991,"thresholds":{"poweredOffSeverity":"warning","cpu":{"trigger":80,"clear":75},"memory":{"trigger":85,"clear":80},"disk":{"trigger":90,"clear":85}},"time":"2025-11-28T12:41:07+01:00","message":"Checking guest thresholds"}

[rcourtman]

Fixed in commit f9bd09a. The "Removed Docker Hosts" section now appears in Settings → Agents when there are hosts blocked from re-enrolling.

Workaround until next release: Use the API directly:

curl -X POST http://<pulse-ip>:7655/api/agents/docker/hosts/636085b6-e965-409f-b777-3e6d54ef2f7d/allow-reenroll

Replace the UUID with the host ID from the error message. After running this, restart the agent service and it should connect.

The fix will be in the next release.

[njunghausz]

Fixed in commit f9bd09a. The "Removed Docker Hosts" section now appears in Settings → Agents when there are hosts blocked from re-enrolling.

Workaround until next release: Use the API directly:

curl -X POST http://:7655/api/agents/docker/hosts/636085b6-e965-409f-b777-3e6d54ef2f7d/allow-reenroll

Replace the UUID with the host ID from the error message. After running this, restart the agent service and it should connect.

The fix will be in the next release.

{"error":"Authentication required"} :)

so append -u user:password at the end of the command

[rcourtman]

Thanks for the clarification - the API endpoint requires authentication since it is an admin operation. The curl command with -u user:password will work. The UI fix from commit f9bd09a is already merged and will be in the next release.

[njunghausz]

Might not be related but I see this in all those hosts that I updated to 4.34.0:

Nov 28 22:18:33 seagull pulse-agent[1645838]: {"level":"info","currentVersion":"v4.34.0","availableVersion":"4.34.0","time":"2025-11-28T22:18:33+01:00","message":"New agent version available, performing self-update"} Nov 28 22:18:35 seagull pulse-agent[1645838]: {"level":"info","version":"v4.34.0","pulse_url":"http://192.168.2.96:7655","host_agent":true,"docker_agent":true,"auto_update":true,"time":"2025-11-28T22:18:35+01:00","message":"Starting Pulse Unified Agent"} Nov 28 22:18:35 seagull pulse-agent[1645838]: {"level":"info","time":"2025-11-28T22:18:35+01:00","message":"Host agent module started"} Nov 28 22:18:35 seagull pulse-agent[1645838]: {"level":"info","component":"pulse-docker-agent","runtime":"docker","daemon_host":"unix:///var/run/docker.sock","version":"29.1.0","time":"2025-11-28T22:18:35+01:00","message":"Connected to container runtime"} Nov 28 22:18:35 seagull pulse-agent[1645838]: {"level":"info","time":"2025-11-28T22:18:35+01:00","message":"Docker agent module started"} Nov 28 22:18:43 seagull pulse-agent[1645838]: {"level":"info","component":"pulse-docker-agent","time":"2025-11-28T22:18:43+01:00","message":"Skipping update check - auto-update disabled"}

Auto update is not in sync for both components, despite not being specified during install.

[rcourtman]

The version prefix mismatch causing the update loop has been fixed in c3a5aac.

The standalone docker agent's checkForUpdates() was comparing versions as raw strings ("v4.34.0" == "4.34.0" → false), triggering updates even when versions match. The unified agent got a normalization fix in v4.34.1 (1b86659), but the standalone docker agent wasn't updated.

This fix normalizes both version strings before comparison.

For agents stuck in the loop now: once the Pulse server is updated (or you manually update to the next release), the agent will self-update and stop looping. If you need immediate relief, restart the agent with --auto-update=false.

[njunghausz]

I dunno.. I still have two hosts with the newest agent that keeps disappearing from the UI. In one case docker stays but host is getting in and out, while the other case it's host only agent that keeps disappearing. It's like it's gone for a minute or two and then comes back. While it's gone the token is also gone.
Can I debug this somehow?