System logs
Edit this pageSystem · Related pages
sqlstreams explain SQL0065System events report registration changes and background processing across the installation. Use worker logs for individual worker ownership and tick failures.
System manager
level=ERROR msg="system manager stopped -- nothing reconciles the deployment's workers until the loop re-claims the row after its backoff" code=SQL0065 attempt={attempt} delay={delay} error={error}The manager’s current run ended with an error without its context being
cancelled. The outer Run call stays active and retries after backoff.
The event carries attempt, delay, and error. A requested shutdown
does not emit this event.
Registration changes
level=WARN msg="stream config replaced -- the newest declaration wins; if this is unexpected or repeats on every restart, two services declare this stream with different configs and overwrite each other" code=SQL0061 stream={stream} stream_id={stream_id} retention_ttl={retention_ttl} allow_drop_past_committed={allow_drop_past_committed} idempotency_key_ttl={idempotency_key_ttl} empty_compaction_head_ttl={empty_compaction_head_ttl} delivery_log_mode={delivery_log_mode}level=WARN msg="schedule config replaced -- the newest declaration wins; if this is unexpected or repeats on every restart, two services declare this schedule with different configs and overwrite each other" code=SQL0062 schedule={schedule} schedule_id={schedule_id} next_scheduled_at={next_scheduled_at} expression={expression} concurrency={concurrency} timeout={timeout} payload_changed={payload_changed} metadata_changed={metadata_changed}Changed settings are rendered as old → new values.
SQL0061 reports changes to retention_ttl, allow_drop_past_committed,
idempotency_key_ttl, empty_compaction_head_ttl, and delivery_log_mode.
Only changed fields appear.
SQL0062 reports changed expression, concurrency, and timeout.
Payload and metadata changes use payload_changed and metadata_changed
booleans without logging their contents. This event does not enumerate
every mutable schedule field. Read the stored registration for its full state.
Repeated warnings on startup indicate differing declarations. Identical settings do not produce these replacement warnings. See the stream and schedule registration contracts.
Scheduling
level=WARN msg="schedule message was already produced by an earlier ambiguous commit" code=SQL0037 schedule_id={schedule_id} schedule={schedule} scheduled_at={scheduled_at}level=WARN msg="schedule target stream keeps no success rows -- Client.Scheduler(name).Status counts no successes for it; set DeliveryLogMode all on the stream to count them" code=SQL0058 schedule={schedule} stream={stream} delivery_log_mode={delivery_log_mode}| Code | Condition |
|---|---|
SQL0037 |
Scheduled production finds its idempotency key already present and skips another append. |
SQL0058 |
Register declares a schedule on a stream whose DeliveryLogMode is not DeliveryLogModeAll. |
SQL0058 does not reject registration. It qualifies the accuracy of
schedule outcome reporting.
Metrics
level=WARN msg="abandoned-routine events dropped -- abandoned-routine snapshots undercount" code=SQL0052 dropped_count={dropped_count}level=WARN msg="measurements cannot be exported -- the current collection omits the rejected metric family; healthy families still export" code=SQL0104 metric_names={metric_names} detail={detail}| Code | Condition |
|---|---|
SQL0052 |
The in-memory event queue overflows, or a batch of abandoned/cleared events cannot be recorded. |
SQL0104 |
Export validation omits a metric family from the current collection. |
level=WARN msg="abandoned-routine events dropped -- abandoned-routine snapshots undercount" code=SQL0052 dropped_count={dropped_count} error={error}SQL0052 means retained abandoned-routine snapshots can undercount. Events
are not requeued after a failed batch. SQL0104 leaves healthy families
available, as defined by Export metrics.
Alerts
level=WARN msg="alert condition holds -- nothing was published; the scheduled check is what publishes and resolves an alert" code=SQL0063 alert={alert} alert_message={alert_message} detail={detail} hint={hint} owner={owner} severity={severity}level=WARN msg="alert evidence does not meet measurement requirements -- recorded alert unchanged" code=SQL0109 alert={alert} owner={owner} owner_kind={owner_kind} system_id={system_id} stream_id={stream_id} group_id={group_id} detail={detail}| Code | Condition |
|---|---|
SQL0063 |
A producer or consumer registration’s advisory alert evaluation finds an unhealthy condition. |
SQL0109 |
A scheduled check rejects semantically invalid evidence. |
The registration-time check is advisory. Its failure does not fail Register,
and its warning does not write an alert. Missing or stale evidence alone
uses a Debug message rather than SQL0109.
level=DEBUG msg="alert evidence is insufficient -- recorded alert unchanged" alert={alert} owner={owner} owner_kind={owner_kind} system_id={system_id} stream_id={stream_id} group_id={group_id} detail={detail}level=INFO msg="alert active" alert=worker_liveness alert_message={alert_message} hint={hint} owner={owner} severity=infolevel=WARN msg="alert active" alert={alert} alert_message={alert_message} detail={detail} hint={hint} owner={owner} severity={severity}level=INFO msg="alert resolved" alert={alert} alert_message={alert_message} owner={owner}After a committed alert state transition, the check logs alert active at
the alert’s severity: Info for worker liveness and Warn for the other built-ins.
It logs alert resolved at Info. These lines have no separate SQL event code.
Active reminders with unchanged status do not repeat transition logs.
The detail attribute appears only when the alert supplies a detail message.
Suppression
System-manager Warn/Error repeats use one-minute suppression.
Registration and exporter events use their supplied logger. A suppressed
repeat can later surface through suppressed_count.
Related topics
- Run the system manager — Lifecycle and retry behavior.
- Read alerts — Evidence, evaluation, and retained reports.
- Worker is not running — Diagnose missing background work.
- Schema version mismatch — Resolve database/build compatibility failures.