Skip to main content
Keep a close eye on Telepúrte installation status and controlled nodes. The general summary shows the availability of components and the accumulation of work; node metrics help to disassemble the CPU, memory, disk, processes and services of a particular machine.

How to start a diagnosis

In ZZZZZ, check the park, sources, agents, bastions and active connections. Then proceed to the problem node or task. The green HTTP response of the panel does not confirm that every agent and SSH route is working. For operational automation, the /api/v1/system/metrics, /api/v1/system/metrics/history and /api/v1/system/alerts APIs are available with the relevant rights. Public /healthz and /readyz do not disclose detailed summaries.

How to understand freshness

An online agent or bastion is based on a recent heartbeat, usually within 45 seconds. For an external Axottle source, the data is considered obsolete after three minutes without updating. These states do not prove that a particular SSH input is successful. The operational summary notes a delayed queue when waiting for more than five minutes, an overdue schedule after a minute of delay, a small free space below 512 MiB. These are the thresholds of the installation diagnostics, and not a guarantee of a safe disk reserve for any load. System-wide minute shots are saved up to 24 hours; the history API takes periods of 1, 6 and 24 hours. Don’t confuse this story with graphs of a particular node, which have their own source and lifespan.

Incidents

Repeated minute shots of the same problem should not create a new incident every time; when the cause disappears, the incident closes; reappearance after closure is a new event. Confirmation of the notification means that it is noticed, but does not correct the cause and does not hide the active incident.

What events are notified

Telepúrte retains personalized notifications of failed, expired and uncertain assignments, schedule failures and omissions, configuration divergences and operational incident changes. A routine successful launch does not necessarily create a notification, so missing a message is not a substitute for checking the outcome of a task; the recipients of the task’s events are its author and the current owners within the current policy. The tape is available through /api/v1/notifications, confirmation through its operation ack. HTTPS webhook is provided for the external system; the recipient is configured on the server, and not by an arbitrary address from the browser form.

Connect Webhook

Add to the existing /etc/telepurte/config.json:
Create a separate secret of 32-128 random bytes available to a service user with 0600 mode. Do not put it in the documentation or URL. Check the configuration command telepurte check with the desired --config, then restart the service at the agreed time. The receiver receives the event metadata, checks the signature for the original body and time, and eliminates duplicates using X-Telepurte-Event-ID. The content of the files and terminals is not sent. Delivery can be repeated: the response of 2xx confirms reception, network errors, 408, 429 and 5xx repeat with a delay. After exhausting attempts, the event remains in the queue of failed deliveries. The owner can view /api/v1/notifications/delivery/dead and repeat the selected delivery after the recipient has been corrected. The webhook error does not cancel the original task or internal notification.

Verification

In a test event, check the internal recording, delivery, signature verification, and replay processing. For an incident, make sure that once the cause is corrected, it closes, not just tagged with read.

Rollback

Return the previous webhook configuration and check it before restarting. Deleting the setting suspends external delivery; when you re-enable the remaining queue can continue. Events created before the first onset are not retroactively sent. The commands refer to a standard Linux installation; the actual paths are checked in your configuration.