Servers and agents
Fix provisioning, health, agent, Docker, and server connectivity problems.
A server is stuck provisioning
provisioning means the agent has never called home. On the machine, run:
systemctl status deplo-agent
journalctl -u deplo-agent -f| What you see | Cause |
|---|---|
| The unit does not exist | The install command never ran, or failed early |
Refusing to run an unverified binary | The download did not match its checksum. Retry; if it persists, something is intercepting it |
| The unit runs but nothing happens | It cannot reach the control plane URL; check outbound access and the URL |
bootstrap token errors | The token was used already, or it expired. Issue a new one from the server's menu |
status=216/GROUP | The unit wants the docker group and Docker is not installed. That is expected on a backups-only host, and its unit is installed without it |
The server is online but every deploy fails
Enrollment uses an outbound connection; deploys also need inbound 9443. See
ports, networks, and files.
Refresh Settings -> Servers to run a new health check. A stored green reading can be stale. The same check applies to the host running Deplo.
From the control-plane machine:
nc -vz 203.0.113.10 9443Open 9443 to the control plane's address and refresh the server.
A server says warning, error, offline, or Unknown
| Value | Means | Where to look |
|---|---|---|
warning | The agent is up and trusted, Docker is not answering | systemctl status docker on the host |
error | Something answered, but its agent is wrong: untrusted certificate, unsupported contract, or an application error | Re-run the install command to re-enroll |
offline | Nothing answered, confirmed by a retry | The machine, the network, or the port |
Unknown | The stored reading is stale | Click Refresh |
Health is a stored reading. A deploy performs a live handshake, so a green card and a failing deploy can still happen together.
Check readiness instead of guessing
Use Check readiness in the server's menu for one live check of the agent, Docker, routing, capacity, build methods, and Deplo configuration.
fail blocks a deploy. warn means it can deploy but is not fully usable.
skip means the check could not be evaluated, not that it passed.
Metrics show dashes
The telemetry stream is not connected. Refresh the server health first. History is held in memory and starts empty after a panel restart.
A deploy says Docker cannot create a network
Deplo uses one Docker network per Environment. If the installer could not restart Docker because containers were running, the configured address pool is not active yet. Restart Docker in a maintenance window; the file is already correct.
A deploy fails with No chain DOCKER-FORWARD
Docker's forwarding chains were removed underneath it. Restart Docker to recreate them.
A removed container still appears
Deplo retries an unconfirmed teardown and raises an alert if it gives up. Removing the server drops its pending teardowns.
The install command returns 503 or cannot resolve a release
The panel cannot reach github.com to resolve or checksum the agent release.
Restore outbound access from the panel's machine, then run the command again.
The install command contains 127.0.0.1
That address is for the machine running the panel. Paste it there only; another server needs the panel's public address.
An agent needs to be reinstalled
Run a fresh install command from the server's Advanced tab. It clears the
old certificates, so an expired or already-used token leaves the agent with
unknown-token. Issue a new token from the server's menu first.
A server is permanently gone
Remove it in Deplo. This revokes trust and does not require the machine to answer. See Remove or uninstall.
Next steps
Did this page help you?