fix: Throw exception if inadequate GPUs #1508
Open
+1
−1
Conversation
This file contains hidden or 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
|
yinggeh
requested review from
hutm,
biswapanda,
ishandhanani,
julienmancuso,
hhzhang16,
nnshah1 and
mohammedabdulwahhab
as code owners
WalkthroughThe change updates the GPU allocation logic in the Changes
Sequence Diagram(s)sequenceDiagram
participant User
participant ResourceAllocator
User->>ResourceAllocator: assign_gpus(requested_count)
ResourceAllocator->>ResourceAllocator: Check remaining GPUs
alt Requested > Remaining
ResourceAllocator-->>User: Raise ResourceError (insufficient GPUs)
else Requested <= Remaining
ResourceAllocator->>ResourceAllocator: Assign GPUs
ResourceAllocator-->>User: Return assigned GPUs
end
Possibly related issues
Poem
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
Documentation and Community
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 0
🔭 Outside diff range comments (1)
deploy/sdk/src/dynamo/sdk/cli/allocator.py (1)
80-86: 🛠️ Refactor suggestionCorrect move from soft-warning to hard failure, but emit a log entry before raising.
Switching from
logger.warningto an immediateResourceErrorfulfils the PR goal 👍.
However, because the raise short-circuits the method, the message is no longer captured in the logs unless the caller logs uncaught exceptions. A singlelogger.error(...)right before the raise keeps parity with the previous observability.if count > self.remaining_gpus: - raise ResourceError( + logger.error( # visible even when the caller swallows the exception + f"Requested {count} GPUs, but only {self.remaining_gpus} are remaining. " + f"Serving may fail due to inadequate GPUs. Set {DYN_DISABLE_AUTO_GPU_ALLOCATION}=1 " + "to disable automatic allocation and allocate GPUs manually." + ) + raise ResourceError( f"Requested {count} GPUs, but only {self.remaining_gpus} are remaining. " f"Serving may fail due to inadequate GPUs. Set {DYN_DISABLE_AUTO_GPU_ALLOCATION}=1 " "to disable automatic allocation and allocate GPUs manually." )
🧹 Nitpick comments (1)
deploy/sdk/src/dynamo/sdk/cli/allocator.py (1)
86-87: Integer cast drops fractional capacity.
self.remaining_gpus = int(max(0, self.remaining_gpus - count))silently floors fractional
remainders. Two consecutiveassign_gpus(0.5, …)calls on a 1-GPU system would mark the
machine as fully consumed after the first call (1 GPU → 1.5 →int()→ 1).Either keep
remaining_gpusas a float or track whole & fractional capacity separately.-self.remaining_gpus = int(max(0, self.remaining_gpus - count)) +self.remaining_gpus = max(0.0, self.remaining_gpus - count) # keep precision
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
deploy/sdk/src/dynamo/sdk/cli/allocator.py(1 hunks)
⏰ Context from checks skipped due to timeout of 90000ms (2)
- GitHub Check: Build and Test - vllm
- GitHub Check: Build and Test - vllm
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.
Overview:
Throw exception if inadequate GPUs in
dynamo serve.Details:
3 GPUs are needed for disagg EPD serving. 1 gpu for each E P and D. If the number of GPUs available to the user is less than the number of GPUs that an EPD service needs, the service "starts up" normally and hangs when the user tries to make an inference call, without any sign or error that the user doesn't have enough GPUs available to run the service.
Expected Behavior
When serving, an error should be thrown at/around the GPU allocation step.
Where should the reviewer start?
Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)
Summary by CodeRabbit