Github user revans2 commented on the issue:
I am really nervous about this patch in general and I am still a -1 on it.
For me to remove my -1 at a minimum all of the unit tests have to pass and
all critical functionality has to be here (the system bolt metrics are being
reported properly). Beyond that I really want to see some benchmarks that are
closer to the real world so I can weigh the cost vs the benefits of this change
in an analytical way.
Right now my gut is telling me that this is too much change all at once for
a performance gain in a single micro benchmark that I believe will have a
minimal impact at best on real topologies, and at a cost that I am not sure I
or my customers are willing to pay.
I would love to be proven wrong and see that everything just gets better
with this patch, and the differences in how topologies behave will be small
enough to help our customers adjust easily. But I am very skeptical.
I plan on spending the next few weeks fixing this patch (at least enough
that I can get system metrics out of it and shuffle does not fail all of the
unit tests), and then try and run what I think is a somewhat representative
sample of topologies on a cluster to see what the numbers actually are. I will
post my results here when I am done.
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 infrastruct...@apache.org or file a JIRA ticket