Github user phrocker commented on the issue:
https://github.com/apache/accumulo/pull/260
Have you addressed Josh's concern about RPC calls? Further if you profile
the context switching involved with the threads migrating is the net benefit
better than stopping the scan entirely? It stands to reason that a long running
scan will continue to be long running.
Finally, since this is mostly uncooperative yielding, is there any way to
measure the fringe conditions, as per metrics that Josh alluded to? Such as
average time before forced yield, etc? I think if you're baking this into SKVI
that should be an expectation of any initial implementation to address the
concern of provability.
---
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.
---
