Github user HeartSaVioR commented on the issue:
https://github.com/apache/storm/pull/2203
In this PR, I see that metrics V2 drops most of built-in metrics.
This PR maintains metrics V1 for built-in and V2 for user-defined and some
of built-in, but in metrics V1 all of built-in metrics are provided to the
metrics consumers, so users will lose some metrics after replacing metrics
consumer to V2. Moreover I guess we would want to replace the old metrics (in
storm internal) eventually.
For worker there's no more task-level metrics, only (worker,
component)-level is available, and it only tracks emit, ack and fail. All
metrics regarding latency are dropped. Netty metrics are dropped too. Some
built-in metrics might not be able to be injected by users, so also need to
consider about this when dropping metrics. Maybe would want to turn on and off
recording part of metrics as Bobby suggested if they impact performance.
All of disruptor queue metrics are taken, but there are many disruptor
queues in worker but metrics are worker-level, and it's gauge (not aggregation)
so one of queue metrics would always overwrite.
So would like to see some explanations that which metrics are selected, how
they're aggregated in worker, how decision has been made, etc.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
