[object store refactor 1/n] Introduce IAllocator and PlasmaAllocator #17307
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
scv119
added
@author-action-required
scv119
force-pushed
the
test-wheels/introduce-allocator
branch
2 times, most recently
from
8bb8d09
to
ae21707
Compare
scv119
removed
the
@author-action-required
rkooo567
reviewed
Jul 27, 2021
| @@ -64,10 +65,22 @@ ray::ObjectID GetCreateRequestObjectId(const std::vector<uint8_t> &message) { | |||
| return ray::ObjectID::FromBinary(request->object_id()->str()); | |||
| } | |||
|
|
|||
| void ToPlasmaObject(const ObjectTableEntry &entry, PlasmaObject *object, | |||
There was a problem hiding this comment.
Why don't we just return the Plasma object instead of accepting mutable arg?
There was a problem hiding this comment.
There is some subtlety at the caller of ToPlasmaObject. Let's defer that to follow up PRs.
| void *pointer = PlasmaAllocator::Memalign( | ||
| kBlockSize, PlasmaAllocator::GetFootprintLimit() - 256 * sizeof(size_t)); | ||
| RAY_CHECK(pointer != nullptr); | ||
| // TODO(scv119): this leaks details of PlasmaAlloctor, |
There was a problem hiding this comment.
Why don't we just fix it here?
There was a problem hiding this comment.
is fixed in next PR!
fishbone
reviewed
Jul 28, 2021
| @@ -49,4 +41,6 @@ struct MmapRecord { | |||
| /// and size. | |||
| extern std::unordered_map<void *, MmapRecord> mmap_records; | |||
|
|
|||
| /// private function, only used by PlasmaAllocator | |||
| bool GetMallocMapinfo(void *addr, MEMFD_TYPE *fd, int64_t *map_length, ptrdiff_t *offset); | |||
There was a problem hiding this comment.
If it's a get, please mark it as const
There was a problem hiding this comment.
this is unfortunate a C function. I'll make addr const.
|
I feel fine about this one. Only one nit. |
rkooo567
approved these changes
Jul 30, 2021
ericl
added
the
@author-action-required
scv119
force-pushed
the
test-wheels/introduce-allocator
branch
from
54bb716
to
28fe635
Compare
|
Let me know when it is mergeable! |
|
This should be mergable once CI passes! |
scv119
removed
the
@author-action-required
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.
Problem
There are two major issues with plasma store:
Which makes the IO layer unstable & vulnerable to race conditions & slow to develop.
Goal
The goal is to make local object management unit testable, easy to extend, and hide implementation detail from the upper layer.
Phase 1: Unit test all major components:
This is the first PR that refactors object store. Next PR #17313
Specifically in this PR, we creates IAllocator interface thats allocates mmaped memories, and refactor the PlasmaAllocator to implement this interface.
Test Plan
-[x] existing tests
-[ ] on demand stress tests
-[x] more unit test coming in follow up PR #17313
Note: I'm trying out stack PRs in ray-project/test-wheels branch so that we can get signals from CI, so we need merge PRs in sequence.