yuqi1129 commented on code in PR #13831:
URL: https://github.com/apache/gravitino/pull/13831#discussion_r4291683768
##########
docs/metrics.md:
##########
@@ -78,28 +78,37 @@ The `entity-change-log` source exposes each server's
change-log processing state
`/prometheus/metrics`. For example, `entity-change-log.record-lag` in the
metrics registry becomes
`entity_change_log_record_lag` in Prometheus. Gauges read only in-memory
values; the poller samples
the database tail on partial and empty polls, and at most once per
`pollIntervalSecs` during
-a full-batch drain (including failed sample attempts). If only the tail sample
fails, delivery
-continues and the tail value remains at its last successful sample.
+a full incremental-batch drain (including failed sample attempts). A full
replay page also
+schedules the next cycle immediately; when its incremental query is partial or
empty, that
+cycle still samples the tail. Each such replay-only drain cycle therefore
performs an incremental
+query, a tail query, and a replay query. If only the tail sample fails,
delivery continues and
+the tail value remains at its last successful sample.
| Metric suffix | Type and unit
| Meaning
|
| ------------------------------------------------------------ |
------------------------ |
-----------------------------------------------------------------------------------------------------------------------------------------------------------
|
-| `db-tail-id`, `cursor-id` | gauge, change
ID | Latest sampled database ID and last delivered ID on this server.
|
+| `db-tail-id`, `cursor-id` | gauge, change
ID | Latest sampled database ID and last incrementally delivered ID on
this server; replay does not move the cursor.
|
| `record-lag` | gauge,
records | Sampled tail minus cursor, clamped at zero. Interpret only
while tail sampling succeeds.
|
| `seconds-since-last-successful-poll` | gauge,
seconds | Time since a successful database poll, including an empty
result; `-1` before the first poll.
|
| `seconds-since-last-successful-tail-sample` | gauge,
seconds | Time since `db-tail-id` was last refreshed; `-1` before the
first sample. Trust `db-tail-id` and `record-lag` only while this stays near
the poll interval. |
| `poll-failures-total` | counter,
failures | Failed poll cycles.
|
| `tail-sample-failures-total` | counter,
failures | Failed database-tail samples; fetched batches can still be
delivered.
|
-| `listener-failures-total`, `listener-failures.<class>-total` | counter,
failures | Total failures and failures by registered listener class.
|
-| `records-fetched-total`, `records-delivered-total` | counter,
records | Rows fetched and rows delivered successfully to listeners;
one row delivered to two listeners counts twice as delivered.
|
+| `listener-failures-total`, `listener-failures.<class>-total` | counter,
failures | Callback failures and callback failures by registered
listener class.
|
Review Comment:
Fixed in df4b5f107c: reworded to "Total callback failures, and failures by
registered listener class."
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
