Troubleshooting
Common cdnca problems by failed step: install failures, sign-in, lost passwords, nodes offline, sites not opening, unclear results and whom to contact.
Find the step that failed. Each entry says how to confirm the problem, then what to do; if you cannot confirm it or it persists, contact the right person with the details you noted.
Note these first
- Platform and page, current account, the object (site domain, node IP, order number).
- Time (with time zone) and what you did.
- The exact message and request ID; for install failures, the failed step and error line.
- Never attach passwords, codes, API keys, DNS keys, private keys or the full install command.
Control-plane install failed
The last lines mark the failed step and the cause; follow the matching row in install the control plane: if the install fails. Do not reinstall the OS by default, and do not rerun a fresh install when step 9 (activate) or later failed.
Installer download returns 403
The server's public IP is not registered in the licence centre, or it is not the egress IP. Set the control-plane IP under "My licences" to the server's egress IP and run the same command again.
Preflight "fail" and "warn"
"Fail" items are hard requirements (OS, architecture, memory, ports in use and so on) and must be fixed; "warn" can be confirmed, or skipped with --yes at the end of the command.
The console does not open or you cannot sign in
- The address includes /dashboard/login and, for a port other than 443, the port. The "Control plane URL" printed by the installer omits the port.
- All services are running: systemctl is-active cdnca-control-api cdnca-agent-gateway cdnca-control-nats cdnca-control-web.
- The firewall and security group allow 443 and 80; the domain points at this server.
- You are using the username you set during install (there is no default admin account).
Lost the admin password
- If the control plane has another administrator, ask them to "Reset password" for you under "System → User list".
- If the only platform administrator cannot sign in, follow the control-plane root recovery runbook and run control-api recover-platform-admin as root on the control-plane server. The runbook ships with the package at /data/cdnca-native-root/cdnca/current/docs/platform-admin-recovery.md and can be downloaded from the licence centre at https://auth.cdnca.com/downloads/platform-admin-recovery.md . Check the account's internal details as the runbook describes before running it; recovery revokes all of the account's sessions but keeps MFA and API keys. Do not edit the database by hand.
A node will not install or come online
- The install command failed: read the failed step and error line; the full output is in /var/log/cdnca-install-node-*.log. Fix the cause and run the same command again.
- The node is not under "Pending initialisation": the command came from this control plane, node quota is left, and the node can reach the control plane and gateway ports.
- The node cannot reach the gateway: open TCP 443, 80 and the two gateway ports (enrollment, channel) printed when the control plane was installed. When nodes reach the gateway through a CDN, keep 443 public and set the origin port to the gateway enrollment port.
- It stays pending: click "Initialise" in the node list.
- Its IP is missing from the line group: the node is initialised, has a usable IP and is in the group's region.
See install a node.
A site does not open or shows the wrong content
| What you see | Check first |
|---|---|
| The CNAME column shows "Generation failed" | Click it for the reason, usually missing DNS configuration or CNAME domain; fix it and "Regenerate" (see add a site) |
| The domain does not resolve to the platform | The CNAME record, host, conflicting records and TTL at your DNS provider |
| Errors or origin failures | Origin address and port, origin protocol and Host, and whether the origin allows the node IPs |
| HTTPS certificate errors | The certificate covers all domains, is active, and is applied under the site's "HTTPS settings" (see HTTPS) |
| Changes do not take effect | The publish task result under "System → Background tasks" |
| Visitors see old content | Purge the cache (see purge and prewarm) |
| Legitimate requests are blocked | The matching rule in the block and WAF logs (see false positives) |
Node disk is full
Purge cache from the control plane first, then check the log directories. Nodes need at least 2 GB; add disk if you cache a lot.
No monitoring data
Make sure the node is online and its clock is in sync (large skew is rejected), then check the gateway ports between them. Statistics lag; give a new node a few minutes.
Licensing service unreachable
The control plane renews a lease; within the lease and grace period it keeps creating and publishing, and renews automatically once the service is back. Running node traffic does not depend on the licensing service.
A save or payment result is unclear
- Refresh the details, the order or the background task to see the real state.
- If the page offers "retry with the same request", use it; do not start an identical operation or pay again.
- If it still does not add up, give the order number, time, amount and message to the payee or the platform.
Who to contact
| Problem | Contact |
|---|---|
| Licences, buying licences, top-ups, licence-centre account | Platform support ("Contact" on the website) |
| Sites, plans and permissions on a control plane | That control plane's administrator |
| Subscribed node failures or balances | Your direct supplier, who can involve the platform |
| Plugin features | The plugin provider; your administrator for install permissions |