Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4586
I can't think of a better solution either.
---
Github user mbode commented on the issue:
https://github.com/apache/flink/pull/4586
I mean in general it is probably not the best thing to just rely on the
port not being available as a consensus algorithm of who should claim which
port. Then again, I could not think of a straightforw
Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4586
So I went back to configuring a single port, starting a flink cluster (JM +
2 TM) and 1 HTTPServer in the IDE.
Starting the HTTPServer in the IDE before the cluster does in fact lead to
an IO
Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4586
And this fails as well:
```
public static void main(String[] args) throws IOException {
PrometheusReporter p1 = new PrometheusReporter();
PrometheusReporter p2 =
Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4586
This is rather odd. This minimal example does fail as expected:
```
public static void main(String[] args) throws IOException {
HTTPServer s1 = null;
Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4586
I've double checked with DEBUG logging enabled, no warning anywhere.
---
Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4586
I configured 3 ports `9001-9003`. Each reporter logged `2017-10-25
13:10:14,848 INFO org.apache.flink.metrics.prometheus.PrometheusReporter
- Started PrometheusReporter HTTP server on port 900
Github user mbode commented on the issue:
https://github.com/apache/flink/pull/4586
Also there should have been the warning "Could not start PrometheusReporter
HTTP server on any configured port. Ports: ...", wasn't this logged?
---
Github user mbode commented on the issue:
https://github.com/apache/flink/pull/4586
Did you configure a port range with sufficient (i.e.) three ports? By
default, it uses only one port. I added a sentence about this to the readme but
maybe we can make this more explicit?
---
Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4586
The port conflict detection doesn't appear to be working as intended. I
started a jobmanager and 2 taskmanagers on the same machine, and each reporter
used the same port. Prometheus could only pick up
Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4586
WIll try it out now and merge it afterwards.
---
Github user mbode commented on the issue:
https://github.com/apache/flink/pull/4586
[Green Travis](https://travis-ci.org/mbode/flink/builds/290468452)
---
Github user mbode commented on the issue:
https://github.com/apache/flink/pull/4586
I implemented your comments and assembled a [small
setup](https://github.com/mbode/flink-prometheus-example) to test the reporter
again.
It currently clones *master* and build the reporter fr
Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4586
Taking another look now.
---
Github user mbode commented on the issue:
https://github.com/apache/flink/pull/4586
@zentol *ping*
---
Github user mbode commented on the issue:
https://github.com/apache/flink/pull/4586
@zentol It would be great if you could have another look on occasion, I
added better handling for metrics that are registered e.g. by different
subtasks.
[green travis](https://travis-ci.org/m
Github user mbode commented on the issue:
https://github.com/apache/flink/pull/4586
[Green Travis](https://travis-ci.org/mbode/flink/builds/268258386)
---
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 n
17 matches
Mail list logo
