Github user hbdeshmukh commented on the issue:
https://github.com/apache/incubator-quickstep/pull/33
That's true. I don't think it is too bug-prone. We just need to just
increment and decrement the count at a couple of places (which I missed,
admittedly).
What TMB gives us is perhaps more accurate snapshot of the queue lengths of
each worker compared to the queue lengths as viewed by Foreman. However the
scheduling decisions themselves are much faster than a typical WorkOrder
execution, so the lag (and corresponding difference in queue lengths) may not
be too big an issue.
On the other hand, I am not sure of the performance implications of
checking the queue lengths through TMB repeatedly, e.g. to find the least
loaded worker before every work order execution.
We may not be able to get rid of WorkerDirectory completely either.
Consider a case when you want to find out the least loaded worker on a specific
NUMA socket. TMB doesn't recognize this case and I don't want to modify TMB
code either. In such a case we need a level of indirection that gives us a list
of workers on that NUMA socket.
---
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.
---
