Diagnose log¶
The diagnose log (/diagnose-log) is a global, filtered event log of Vesana's own components — for troubleshooting right in the UI, instead of digging through container logs over SSH. You only see what your permissions allow (own, tenant-wide, or global scope).
What lands here¶
| Category | Examples |
|---|---|
| Connection | agent/collector auth failures (wrong token, IP), agent/collector offline ↔ online |
| Alert & mail | delivery of a notification — success and failure |
| Check config | check created, changed, deleted |
| Status | a check's status change |
| Check results | every incoming result (see below) |
| Flapping | flapping detected / resolved, including a short history timeline |
| Auth | 2FA lockouts |
| Maintenance | database self-maintenance (retention, compression) |
| Ingest | processing problems for incoming results (see below) |
| Suppressed | messages that did not go out, with the reason (maintenance, host unreachable, dependency, info mode, acknowledgement) |
| AI | problems around the AI features: provider unreachable in chat or analysis, analysis result could not be cached, rejected connection test |
Notifications live here¶
The former dedicated "notification log" page is gone — notifications now show up as the "alert & mail" category in the diagnose log. Every delivery creates an entry: sent successfully or failed, with channel, occasion (alert/recovery/test/SSL certificate), and the affected host/check.
The "all" tab also reads check results¶
The "all" tab shows literally everything: the noteworthy system events and the raw check results — the latter are read directly from the check results at display time, not stored twice (with thousands of checks, a second write would be needlessly expensive). If you want just the notable events without the results flood, filter on a category other than "check results".
Making ingest problems visible¶
If a single result arrives unreadable (e.g. corrupted JSON), only that one result is skipped and noted in the diagnose log — not the whole incoming batch discarded. If an entire batch ends up permanently unprocessable in the dead letters, that shows up here too, instead of silently vanishing.
Filtering¶
Rich filter by severity, category, event type, host, collector, check, status, source IP, actor, time range (relative or calendar), and full-text search. As in the error overview, filter combinations can be saved as your own tab and shared.
Live mode¶
A live button polls for new entries automatically — handy while you're installing a fresh agent or watching a config change take effect.
Detail panel¶
Clicking an entry opens a side panel with full context plus a direct jump to the affected host or check. For a flapping entry, the panel additionally shows a short timeline of how it got there (the last status transitions).
Next¶
- Error overview — current check problems instead of system events
- Notification channels — where notifications go
- Logs — collected application/OS logs from the agent (a different feature from the diagnose log)