Github user CodingCat commented on the pull request:
https://github.com/apache/spark/pull/2851#issuecomment-74185015
Hi, @andrewor14, thanks for the review.
To answer your question:
> how the broadcast blocks info is sent to the driver before and after this
PR?
before this PR, the broadcast blocks was not sent to the driver from
executors,
in this PR, I am exactly using the strategy mentione by @shivaram, i.e.
setting tellMaster=true
> Would all the new logic just go away if we simply set tellMaster = true
for all broadcast blocks,
No...the complexity comes from 1) updating the page, and 2) as mentioned by
@rxin , we need to avoid introduce more complexity to BlockManagers (at the
beginning I simply piggyback the broadcast info in the BlockManager heartbeat)
to address 2), I use the listenerBus to notify the StorageStatusListener
and StorageListener that some of the blocks are updated, they need to apply the
operation respectively (StorageStatusListener => update
executorIdToStorageStatus, StorageListener => update internal data structure)
---
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]
