For whom: Panel Administrator, SRE. Required rights: Panel Administrator; for check intervals, health policy change; for database management, access to PostgreSQL Panel.
Slow dashboard or panel
The panel actively polls the backend to show the state in real time, which is normal, but under certain conditions it feels like a brake.- Lots of open panel tabs. The dashboard and the health page poll a summary and incidents every 15 to 60 seconds; multiple open tabs mean multiple parallel series of queries. Keep one work tab open.
- A heavy audit log filter. The audit log page filters by substring. On a large table (the audit log does not clean automatically - see below), the broad request is processed slowly.
?q=), use a page-by-page output (limit/offset, maximum 500 entries per page). Journal and audit. - Slow dependency. If a particular screen (Remnawave, CDN) is braked, the cause is often externally dependent, not in the panel: live requests to Remnawave or the provider are waiting for a response. Check connectivity - Connectivity problems.
Slow health checks
Scheduled and manual checks behave differently – this is important for interpretation.- Planned checks are scheduled with default intervals: basic reachability every 60 seconds, service check every 120 seconds, through check circuits every 10 minutes. Intervals are customized in health policy, but the panel does not let them fall below the safe minimum (15 s / 30 s / 60 s) is protection against self-overload. See Health Policy.
- Manual checks (Start check button),
POST .../check) shall be carried out immediately and shall attempt exactly one time without repetition.
- Do not try to speed up the system by lowering the intervals, the minimum will still apply, and the load will increase. Too frequent checks slow down the panel, not speed up detection.
- Long, end-to-end checks almost always rest on the slow network to the node or to the route target, rather than the panel. Runbook health..
Node load
The panel collects system metrics from each node: CPU, load average, memory, swap, disk, network and additional indicators (sockets, conntrack, descriptors, CPU pressure, etc.). Where to watch:- Tab Services - what of the services of the node works and what
absent(undetermined).
Growth of the database
Some of the data Axottle cleans itself, some of it doesn’t consciously, and it’s important to know the difference so that the dashboard doesn’t fill up unnoticed.
Key point: The audit log does not have an automatic persistence — this is done intentionally so as not to lose your history. In an active installation, the table becomes large over time and slows down the filtering by setting. If you need a persistence, this is an operator’s task: plan the periodic archiving/traction of the audit log on the PostgreSQL side as a serviced operation.
Storage recommendations
- Give backup profiles a reasonable
retention_count- old runs are deleted automatically after each new one. See Backup copies of the panel. - Keep backups outside of the panel host (S3-compatible storage) so they don’t take up a local drive.
- Keep an eye on the drive:
axottle doctorwarns of an ending location before it becomes an accident. - If you’re planning an audit log retention, first remove the verified base backup, then truncate, and do it as a scheduled maintenance operation. Service.
Verification
The panel is responsive,
axottle doctor does not show Failed over disk/memory/CPU/base, node metrics are collected without omissions, and the volume of the audit log is controlled by the selected retention.Rollback
Return the previous intervals in the tab Management → Settings → Health policy. Changes in the node performance profile are rolled back by the button Rollback in the tab Performance.Related pages
Metrics and services
CPU, memory, disk, network on the node.
Health policies
Inspection intervals and their minimums.
Journal and audit
Audit log and its credential.
Service
Doctor, backups, regulations.
