Metric and trace names
This page lists every span and instrument. Both the ActivitySource and the Meter are named Deedbox.
// With OpenTelemetry: subscribe to the "Deedbox" ActivitySource and Meter.// .WithTracing(t => t.AddSource("Deedbox"))// .WithMetrics(m => m.AddMeter("Deedbox"))const string SourceAndMeter = "Deedbox";| Span | Tags |
|---|---|
deedbox.append, deedbox.execute, deedbox.load |
deedbox.stream_type, deedbox.stream_id, deedbox.events |
deedbox.delete_stream, deedbox.erase_stream |
deedbox.stream_id |
deedbox.batch |
deedbox.consumer, deedbox.from_position, deedbox.to_position |
deedbox.handle <consumer> |
deedbox.consumer, deedbox.event_type, deedbox.global_position. Its parent is the append that wrote the event. |
deedbox.job |
deedbox.job.kind, deedbox.job.id |
Instruments
Section titled “Instruments”| Instrument | Unit | Tags |
|---|---|---|
deedbox.append.duration |
ms | deedbox.stream_type |
deedbox.events.appended |
events | deedbox.stream_type |
deedbox.append.conflicts |
conflicts | deedbox.stream_type |
deedbox.execute.retries |
retries | deedbox.stream_type |
deedbox.counter.duration |
ms | none. Counter wait plus hold time. |
deedbox.consumer.lag |
positions | deedbox.consumer, deedbox.schema |
deedbox.consumer.lag.seconds |
s | deedbox.consumer, deedbox.schema |
deedbox.consumer.status |
0 running, 1 rebuilding, 2 stalled. A retired consumer reports 0. | deedbox.consumer, deedbox.schema |
deedbox.consumer.batch.duration |
ms | deedbox.consumer |
deedbox.consumer.failures, deedbox.consumer.stalls |
count | deedbox.consumer |
deedbox.jobs |
jobs | deedbox.job.kind, deedbox.job.status |
deedbox.erasure.streams |
streams | none |
deedbox.archive.events |
events | none |
deedbox.personal_data.decrypts, deedbox.personal_data.redactions |
fields | none |
Each instance reports the consumer gauges for the consumers it runs. An instance updates deedbox.consumer.lag only after it commits a batch, so the gauge does not grow for a stalled consumer. Alert on the health check and on deedbox status instead; see stalled projection.
Log events have stable IDs: 1 to 5 at start-up, 20 to 29 in the runner, and 30 to 34 for jobs. IDs 40 to 46 are for the heartbeat, the cut-over of an inline projection, and consumers left to a newer version. Every start-up error is a DeedboxException with a DBX code.
Events added in 0.5.0:
| ID | Level | Says |
|---|---|---|
| 5 | Warning | A new subscription starts at the first event and handles the events that the store already holds; the event gives the count. |
| 29 | Warning | A consumer met a transient database error. The runner runs the batch again and counts no attempt. |
| 42 | Warning | Deedbox did not count this instance as live and refused an append. The instance joins again. |
| 43 | Error | This instance could not join again; it retries. Appends fail with DBX038 until it does. |
| 44 | Warning | A cut-over removed the heartbeat rows of the instances it names, because their heartbeat was late. |
| 45 | Warning | An inline projection cannot finish its catch-up: forced cut-overs ran out of time. |
| 46 | Information | This instance leaves a consumer to the instances it names: their version handles events that its own version does not. |
A stall record and a failed job hold the exception type and stack frames. The exception’s message is only in these logs.