Github user squito commented on the pull request:
https://github.com/apache/spark/pull/7770#issuecomment-127243422
I didn't realize the point of that other PR was so we could start switching
all task metrics over ... accumulators have their own set of confusing
semantics around retries. Plus You lose all type info on the values, now they
come through as just strings, and also the "key" goes from being a member of
the developer api that you can check against, to being a string constant that
is `private[spark]`.
I don't see how we will migrate without breaking backwards compatibility.
It seems like we'll just be stuck with some metrics in each until we go to 2.0.
At that point we might be better off redesigning a metrics system with the
semantics we actually want.
I can see the appeal of adding some untyped way to get metrics, like MR
counters, which even users could add to, I just don't know if accumulators
actually give it to us. But it sounds like you are telling me this is already
a fait accompli?
---
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]
