[
https://issues.apache.org/jira/browse/FLINK-23486?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17422150#comment-17422150
]
Roman Khachatryan commented on FLINK-23486:
-------------------------------------------
Hi [~complone], I've extracted back-pressure monitoring into FLINK-24402, as
it's implementation is independent from other metrics but OTH it depends on the
back-pressure. I think it can be reported as back-pressured by backend per
second, similar to how "regular" back-pressure is currently reported
([prototype|https://github.com/rkhachatryan/flink/tree/clsb-bp-test]).
Besides of it, I think we also need:
# low-level (uploader) metrics, such as request rate, errors, saturation, and
latency
# and feature-specific metrics, such as for cross-operator grouping, and
retries
# backend-level latency metrics (similar to existing latency tracking)
I've added a PR to add (1) and (2).
(3) is independent as well and seems less critical.
> Add monitoring/metrics for the ChangelogStateBackend
> ----------------------------------------------------
>
> Key: FLINK-23486
> URL: https://issues.apache.org/jira/browse/FLINK-23486
> Project: Flink
> Issue Type: Sub-task
> Components: Benchmarks
> Reporter: Roman Khachatryan
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.15.0
>
>
> E.g. in-flight requests, request size, latency, number of “Logs” per request,
> errors.
>
> With back-pressure (FLINK-23381) it's very important because the task will be
> shown as busy in the UI.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
