You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@mesos.apache.org by Joseph Wu <jo...@mesosphere.io> on 2016/04/08 02:02:24 UTC
Review Request 45910: Move metrics code into the scheduler body.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45910/
-----------------------------------------------------------
Review request for mesos, Anand Mazumdar, Artem Harutyunyan, and Vinod Kone.
Bugs: MESOS-4982
https://issues.apache.org/jira/browse/MESOS-4982
Repository: mesos
Description
-------
Since the HTTP scheduler is a libprocess process, the sub-class
`Metrics` does not need to be a process anymore. This moves the
process-specific code into the appropriate place in the scheduler.
Diffs
-----
src/examples/long_lived_framework.cpp ef498d63bc5f0a8deb46d71edd85a76a1d38fdd0
Diff: https://reviews.apache.org/r/45910/diff/
Testing
-------
make check
Deployed on a test cluster (see the first review in the chain.)
Thanks,
Joseph Wu
Re: Review Request 45910: Move metrics code into the scheduler body.
Posted by Mesos ReviewBot <re...@mesos.apache.org>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45910/#review127722
-----------------------------------------------------------
Bad patch!
Reviews applied: [45910, 45800, 45799, 45440, 45067]
Failed command: ./support/apply-review.sh -n -r 45067
Error:
2016-04-08 02:53:42 URL:https://reviews.apache.org/r/45067/diff/raw/ [13287/13287] -> "45067.patch" [1]
Total errors found: 0
Checking 1 files
Error: No line in the commit message summary may exceed 72 characters.
Full log: https://builds.apache.org/job/mesos-reviewbot/12398/console
- Mesos ReviewBot
On April 8, 2016, 12:02 a.m., Joseph Wu wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45910/
> -----------------------------------------------------------
>
> (Updated April 8, 2016, 12:02 a.m.)
>
>
> Review request for mesos, Anand Mazumdar, Artem Harutyunyan, and Vinod Kone.
>
>
> Bugs: MESOS-4982
> https://issues.apache.org/jira/browse/MESOS-4982
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Since the HTTP scheduler is a libprocess process, the sub-class
> `Metrics` does not need to be a process anymore. This moves the
> process-specific code into the appropriate place in the scheduler.
>
>
> Diffs
> -----
>
> src/examples/long_lived_framework.cpp ef498d63bc5f0a8deb46d71edd85a76a1d38fdd0
>
> Diff: https://reviews.apache.org/r/45910/diff/
>
>
> Testing
> -------
>
> make check
>
> Deployed on a test cluster (see the first review in the chain.)
>
>
> Thanks,
>
> Joseph Wu
>
>
Re: Review Request 45910: Move metrics code into the scheduler body.
Posted by Vinod Kone <vi...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45910/#review127896
-----------------------------------------------------------
Ship it!
Ship It!
- Vinod Kone
On April 8, 2016, 8:29 p.m., Joseph Wu wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45910/
> -----------------------------------------------------------
>
> (Updated April 8, 2016, 8:29 p.m.)
>
>
> Review request for mesos, Anand Mazumdar, Artem Harutyunyan, and Vinod Kone.
>
>
> Bugs: MESOS-4982
> https://issues.apache.org/jira/browse/MESOS-4982
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Since the HTTP scheduler is a libprocess process, the sub-class
> `Metrics` does not need to be a process anymore. This moves the
> process-specific code into the appropriate place in the scheduler.
>
>
> Diffs
> -----
>
> src/examples/long_lived_framework.cpp ef498d63bc5f0a8deb46d71edd85a76a1d38fdd0
>
> Diff: https://reviews.apache.org/r/45910/diff/
>
>
> Testing
> -------
>
> make check
>
> Deployed on a test cluster (see the first review in the chain.)
>
>
> Thanks,
>
> Joseph Wu
>
>
Re: Review Request 45910: Move metrics code into the scheduler body.
Posted by Joseph Wu <jo...@mesosphere.io>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45910/
-----------------------------------------------------------
(Updated April 8, 2016, 1:29 p.m.)
Review request for mesos, Anand Mazumdar, Artem Harutyunyan, and Vinod Kone.
Changes
-------
Remove `/framework/metrics` endpoint.
Bugs: MESOS-4982
https://issues.apache.org/jira/browse/MESOS-4982
Repository: mesos
Description
-------
Since the HTTP scheduler is a libprocess process, the sub-class
`Metrics` does not need to be a process anymore. This moves the
process-specific code into the appropriate place in the scheduler.
Diffs (updated)
-----
src/examples/long_lived_framework.cpp ef498d63bc5f0a8deb46d71edd85a76a1d38fdd0
Diff: https://reviews.apache.org/r/45910/diff/
Testing
-------
make check
Deployed on a test cluster (see the first review in the chain.)
Thanks,
Joseph Wu
Re: Review Request 45910: Move metrics code into the scheduler body.
Posted by Vinod Kone <vi...@gmail.com>.
> On April 8, 2016, 7:11 p.m., Vinod Kone wrote:
> > src/examples/long_lived_framework.cpp, lines 136-149
> > <https://reviews.apache.org/r/45910/diff/1/?file=1331774#file1331774line136>
> >
> > I'm still not convinced that duplicating this information in /metrics/snashot and /counters is a great idea. We haven't done this for other metrics in the code base.
> >
> > Is it a limitation of the monitoring system you are using that disallows it to slurp both counters and gauges from the same endpoint?
>
> Joseph Wu wrote:
> There is no duplication here. (The `/counters` endpoint doesn't give metrics, only the name of metrics.)
>
> As for the monitoring system, there are a couple types of queries which assume a monotonically increasing value (counters). So the data is less useful if we cannot differentiate between counters/gauges. For now, this code isn't super-critical, so I'll remove it here.
>
> Do you want me to remove it from the earlier review too?
just removing here is fine.
- Vinod
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45910/#review127845
-----------------------------------------------------------
On April 8, 2016, 8:29 p.m., Joseph Wu wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45910/
> -----------------------------------------------------------
>
> (Updated April 8, 2016, 8:29 p.m.)
>
>
> Review request for mesos, Anand Mazumdar, Artem Harutyunyan, and Vinod Kone.
>
>
> Bugs: MESOS-4982
> https://issues.apache.org/jira/browse/MESOS-4982
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Since the HTTP scheduler is a libprocess process, the sub-class
> `Metrics` does not need to be a process anymore. This moves the
> process-specific code into the appropriate place in the scheduler.
>
>
> Diffs
> -----
>
> src/examples/long_lived_framework.cpp ef498d63bc5f0a8deb46d71edd85a76a1d38fdd0
>
> Diff: https://reviews.apache.org/r/45910/diff/
>
>
> Testing
> -------
>
> make check
>
> Deployed on a test cluster (see the first review in the chain.)
>
>
> Thanks,
>
> Joseph Wu
>
>
Re: Review Request 45910: Move metrics code into the scheduler body.
Posted by Joseph Wu <jo...@mesosphere.io>.
> On April 8, 2016, 12:11 p.m., Vinod Kone wrote:
> > src/examples/long_lived_framework.cpp, lines 136-149
> > <https://reviews.apache.org/r/45910/diff/1/?file=1331774#file1331774line136>
> >
> > I'm still not convinced that duplicating this information in /metrics/snashot and /counters is a great idea. We haven't done this for other metrics in the code base.
> >
> > Is it a limitation of the monitoring system you are using that disallows it to slurp both counters and gauges from the same endpoint?
There is no duplication here. (The `/counters` endpoint doesn't give metrics, only the name of metrics.)
As for the monitoring system, there are a couple types of queries which assume a monotonically increasing value (counters). So the data is less useful if we cannot differentiate between counters/gauges. For now, this code isn't super-critical, so I'll remove it here.
Do you want me to remove it from the earlier review too?
- Joseph
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45910/#review127845
-----------------------------------------------------------
On April 8, 2016, 1:29 p.m., Joseph Wu wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45910/
> -----------------------------------------------------------
>
> (Updated April 8, 2016, 1:29 p.m.)
>
>
> Review request for mesos, Anand Mazumdar, Artem Harutyunyan, and Vinod Kone.
>
>
> Bugs: MESOS-4982
> https://issues.apache.org/jira/browse/MESOS-4982
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Since the HTTP scheduler is a libprocess process, the sub-class
> `Metrics` does not need to be a process anymore. This moves the
> process-specific code into the appropriate place in the scheduler.
>
>
> Diffs
> -----
>
> src/examples/long_lived_framework.cpp ef498d63bc5f0a8deb46d71edd85a76a1d38fdd0
>
> Diff: https://reviews.apache.org/r/45910/diff/
>
>
> Testing
> -------
>
> make check
>
> Deployed on a test cluster (see the first review in the chain.)
>
>
> Thanks,
>
> Joseph Wu
>
>
Re: Review Request 45910: Move metrics code into the scheduler body.
Posted by Vinod Kone <vi...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45910/#review127845
-----------------------------------------------------------
src/examples/long_lived_framework.cpp (lines 132 - 145)
<https://reviews.apache.org/r/45910/#comment191218>
I'm still not convinced that duplicating this information in /metrics/snashot and /counters is a great idea. We haven't done this for other metrics in the code base.
Is it a limitation of the monitoring system you are using that disallows it to slurp both counters and gauges from the same endpoint?
src/examples/long_lived_framework.cpp (line 446)
<https://reviews.apache.org/r/45910/#comment191217>
seems weird for this to be a gauge when this is actually a counter.
just do, ++metrics.tasks_launched in launch().
- Vinod Kone
On April 8, 2016, 12:02 a.m., Joseph Wu wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45910/
> -----------------------------------------------------------
>
> (Updated April 8, 2016, 12:02 a.m.)
>
>
> Review request for mesos, Anand Mazumdar, Artem Harutyunyan, and Vinod Kone.
>
>
> Bugs: MESOS-4982
> https://issues.apache.org/jira/browse/MESOS-4982
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Since the HTTP scheduler is a libprocess process, the sub-class
> `Metrics` does not need to be a process anymore. This moves the
> process-specific code into the appropriate place in the scheduler.
>
>
> Diffs
> -----
>
> src/examples/long_lived_framework.cpp ef498d63bc5f0a8deb46d71edd85a76a1d38fdd0
>
> Diff: https://reviews.apache.org/r/45910/diff/
>
>
> Testing
> -------
>
> make check
>
> Deployed on a test cluster (see the first review in the chain.)
>
>
> Thanks,
>
> Joseph Wu
>
>