Github user lulf commented on the issue:
https://github.com/apache/qpid-dispatch/pull/159
Thank you for your comments. My example metric was perhaps too simplicic,
apologies for that. Here are more examples of the metrics I'd like to be able
to provide:
* Counting the number of connections made (tagged with container id) in
total, even if they are not active now
* The number of failed connection attempts, link attaches etc.
* Internal router errors of different types (could be non-amqp specific)
* Latency metrics (time spent in core thread, in I/O thread etc), possibly
as a histogram (not supported in this example API, but part of the prometheus
metric spec: https://prometheus.io/docs/practices/histograms/)
* Ability to report quantiles and histograms
Is this information that I can retrieve in the same way as qdstat or
specify in the management.json? My impression was that qdstat only allowed for
'current state' type of metrics. If we can represent the above using
management, I agree it could be better to use that.
However, there is also the question of the impact of collecting metrics
(tradeoff between quering for the metrics vs. the cost of reporting the
metrcs). Reporting a metric explicitly may scale better, since a counter can be
incremented and queried without locking in the common case. Correct me if I'm
wrong, but my impression was that management operations either has to run in
the core thread or do some locking?
Would an alternative here be to have management/qdstat retrieve the metrics
from the metrics manager? This would resolve the concerns of reporting metrics
twice, and could also potentially enrich the qdstat output with values from
histograms etc.
For reference, here is the document describing the format:
https://prometheus.io/docs/instrumenting/exposition_formats/
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
