[core] Revert change to use ready to create plasma_object_ids #50467
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: dayshah <[email protected]>
dayshah
commented
Feb 12, 2025
| const auto &obj_id = *iter; | ||
| auto found = memory_store->GetIfExists(obj_id); | ||
| if (found != nullptr && found->IsInPlasmaError()) { | ||
| plasma_object_ids.insert(obj_id); | ||
| ready.erase(iter); | ||
| memory_object_ids.erase(obj_id); |
There was a problem hiding this comment.
we don't use memory object ids after this so no need for this
|
i think a possible fix for this could be to pass two vectors into plasma provider wait one with all the objects in plasma - we'll use this to start doing the pulling one with just the number of objects we need from ready, we'll use this to actually make the WaitRequest |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Why are these changes needed?
Reverting the change to use ready to create plasma_object_ids here. #49218
An example ray data workload that highlights the problem with the issue is here https://gist.github.com/dayshah/1080db0cd3fb561119bca17c85215117. 5 seconds without, 50 seconds with.
The problem is that ready is capped to num_returns, and we do object pulling based on plasma_object_ids which was now being created from ready. Ray data calls
ray.wait(10_refs, num_returns=1). If we create plasma_object_ids with only ready, it'll only contain one object in this situation. If we use memory_store_ids to create plasma_object_ids, we'll have plasma_object_ids with 10 objects and all 10 will start being pulled even if the ray.wait call returns immediately after getting one.Related issue number
Checks
git commit -s) in this PR.scripts/format.shto lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/under thecorresponding
.rstfile.