Logs

Use Logs to correlate dedicated database behavior with operational events from its provider-backed machines and TigerBeetle runtime.

Availability by plan

Shared databases remain observable through their tenant-scoped Metrics, API results, and database status. Moving to a different URL does not bypass the shared-plan restriction.

Before you begin

You need:

• a signed-in Parix account with access to the organization and database
• a Dedicated or Production database
• a database profile; a new database can have an empty log view until provisioning creates its profile and collectors start
• a time range that overlaps the event you are investigating

Logs is read-only. Viewing, zooming, or following the stream does not queue a workflow and does not restart or otherwise affect the database.

Open the Logs page

1. Select the organization.
2. Open the dedicated database from Dashboard.
3. Select Logs in the database navigation.

The page opens to the last hour. Its header shows:

• Window: the selected preset label
• Fetched: when the current page data was loaded
• Mode: Following latest, Follow paused, or Historical

Choose a time range

The Time Window card provides four presets:

• 15m
• 1h (default)
• 6h
• 24h

For an exact interval:

1. Enter a local date and time in Start.
2. Enter a later local date and time in End.
3. Select Apply Range.

If the range is invalid or Start is after End, Parix falls back to the selected preset instead of querying a reversed interval.

The timeline divides the selected interval into 36 buckets. Hover a bar to see its event count and exact interval. Drag the brush handles over a spike to navigate to that narrower range.

Read and follow the stream

The Log Stream is ordered from oldest at the top to newest at the bottom. Each row contains:

The initial view loads up to 100 rows. Scroll to the top to load older rows in additional chunks of 100 while preserving your scroll position.

Live follow

A window whose end is within two minutes of the present is treated as live. While you remain at the bottom, the page follows new events and reports Following Latest. It checks for new logs every five seconds and can also react to observability updates.

If you scroll away from the bottom, the badge changes to Live Follow Paused so new rows do not move the content under you. Select Jump To Latest to return to the bottom and resume following.

A historical window does not move forward automatically. Change the range when you want a newer interval.

Expected result

When events exist, the timeline shows their distribution and the stream shows canonical event order. No logs in the selected time range and No logs found in this window yet are valid empty states.

Retention and current limitations

• The current production collector keeps log rows for up to seven days and at most 50,000 rows per database profile. High-volume databases can reach the row cap before seven days.
• Presets stop at 24 hours, but the Start and End fields can request another retained interval.
• The current viewer filters by time only. It does not expose a text search, severity filter, component filter, or export control.
• Log availability depends on collector ingestion. The page does not open a shell, SSH session, or direct connection to a provider machine.
• AWS may not be offered for new production databases in the live create catalog even though logging paths support existing or non-production AWS-backed deployments.

Do not use dashboard log retention as a compliance archive. Export or route required records through an approved external system before the collector retention or row limit removes them.

Queued workflows and status behavior

Logs does not queue a control-plane workflow. The page performs an initial read, then incremental reads for older or newer rows. A page-load failure returns a safe message instead of raw backend details; an incremental failure leaves already loaded rows in place and shows an error above the stream.

Troubleshooting

The Logs item is missing

Confirm the database plan. Developer/shared databases intentionally omit Logs. Use Metrics and API error responses for tenant-scoped diagnosis, or use a dedicated-family database when infrastructure logging is required.

The page is empty

1. Select 24h to widen the range.
2. Confirm Start and End overlap the expected event and that your browser timezone is correct.
3. Return to Dashboard and confirm the database is Ready.
4. Generate a known operation, then return to a live range and wait for collector ingestion.
5. Check Metrics for machine state and recent API activity.

An empty result without an error is not a collector failure by itself.

Live follow says paused

Select Jump To Latest. Live follow pauses whenever you scroll away from the bottom, even if the selected interval is still live.

Older rows do not load

If the stream reports Start of selected range reached, widen Start. If the event is older than the current retention window or was displaced by the 50,000-row cap, it is no longer available in this viewer.

The page says “Failed to load logs”

Refresh once, then try a smaller time window. If the error persists while the database is Ready, record the database name, requested interval, and fetch time for support. Do not treat the safe error message as the same state as a successful empty stream.

For the shared checklist, see Observability, backups, and CDC.

Next steps

• Compare the event time with Metrics.
• Use Cluster configurations only after you understand whether the event indicates a capacity or topology problem.
• Review Backup and restore before a recovery operation.