Collector progress alert (SQL0101)
Edit this pageSystem · Related pages
metrics_collector_progressReports that metrics collection has not completed within the allowed age while the system manager remains active. It checks collector completion separately from the collector’s own work, so partial collections cannot keep the alert healthy.
Definition
| Property | Value |
|---|---|
| Code | SQL0101 |
| Severity | warn |
| Owner | System. |
| Evidence | Latest retained sqlstreams.metrics.collector.completed_timestamp measurement and retained system-manager lease intervals. |
| Unhealthy condition | Completion is absent or at least MaximumAge old, with current manager lease coverage. |
Default MaximumAge |
Greater of 2m and three times the collector’s declared PollRate. |
| Default activation | 2m of established unhealthy time with continuous manager lease coverage. |
| Default evaluation schedule | @every 1m. |
| Active reminders | Every 4h by default. |
Example
Using an initialized client, evaluate collector progress:
snapshot, err := client.System().Alerts().MetricCollectorProgress().Snapshot(ctx)if err != nil { return err}fmt.Println(snapshot.State, snapshot.MaximumAge)Snapshot does not write or resolve a report. Use Latest or History on
the same selector to read retained reports. See Read alerts
for their returns, fields, and errors.
Configuration
Set SystemConfig.MetricCollectorProgressAlert
when registering the system. MaximumAge resolves during evaluation from
the registered collector’s poll rate unless explicitly set.
PendingDuration requires a continuous manager lease span. DisablePending
removes the pending delay but still requires current manager coverage.
There is no MaximumGap setting for this alert. Queued checks retain their
policy, while Snapshot reads the current schedule declaration.
Evaluation and resolution
| Evidence | Result |
|---|---|
Valid completion younger than MaximumAge |
Healthy. Resolve an active report. |
| Missing or old completion, with enough continuous manager coverage after it became unhealthy | Active. Write a report or a due reminder. |
| Current coverage exists but the unhealthy span is too short | Pending. Leave the report unchanged. |
| No current manager lease coverage | Insufficient evidence. Leave the report unchanged. |
| Malformed or future completion evidence | Insufficient evidence with EvidenceInvalid set. Leave the report unchanged. |
A valid completion must have the declared gauge name and seconds unit, with a positive integer Unix timestamp no later than database evaluation time.
When a completion exists, unhealthy time starts after its MaximumAge has
elapsed. Time before continuous manager coverage does not count. Touching
or overlapping manager leases preserve coverage across instances.
This alert does not detect a completely stopped installation. Without a manager, its scheduled check cannot run. Use external availability monitoring for that failure.
Reported data
Active Alert reports add:
| Data key | Type | Meaning |
|---|---|---|
completed_at |
Timestamp | Last completed pass, or the zero time when none is retained. |
maximum_age |
Number | Effective allowed age in nanoseconds. |
The report message distinguishes an overdue completion from no retained completion. Resolved reports do not retain the active-report data fields.
CLI
sqlstreams alert latest metrics_collector_progressThis is a system-owned alert, so no stream flag is needed. No retained report
exits with status 1. See alert commands.
Related topics
- System metrics — The collector completion measurement.
- Run the system manager — Worker lifecycle.
- Worker is not running — Inspect worker state and failures.