[fix] Increase mooncake_master client_ttl to mitigate client heartbeat timeout #99
Merged
Conversation
Signed-off-by: 0oshowero0 <o0shower0o@outlook.com>
CLA Signature Pass0oshowero0, thanks for your pull request. All authors of the commits have signed the CLA. 👍 |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR adjusts the MooncakeStore auto-initialization path to reduce unexpected Mooncake client disconnects/segment unregistrations under heavy Python-side load by increasing the mooncake_master client TTL.
Changes:
- Add
-client_ttl=30to themooncake_masterstartup command used whenMooncakeStore.auto_initis enabled.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
146
to
150
| cmd = [ | ||
| "mooncake_master", | ||
| "-client_ttl=30", | ||
| "-default_kv_lease_ttl=999999", | ||
| "-default_kv_soft_pin_ttl=999999", |
0oshowero0
changed the title
mooncake_master client_ttl to prevent client segment unmountmooncake_master client_ttl to relief client segment unmount
0oshowero0
changed the title
mooncake_master client_ttl to relief client segment unmountmooncake_master client_ttl to mitigate client heartbeat timeout
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Background
During the integration of MooncakeStore with verl, we encountered unexpected connection errors and segment unregistrations. After investigation, we identified that the issue is caused by the starvation or stalling of Mooncake's heartbeat mechanism.
Although Mooncake's heartbeat logic is implemented in C++ and theoretically should not be affected by Python-level concurrency models (such as ray, asyncio, or multi-threading/coroutines), our analysis indicates that the blockage likely occurs at the C++/Python interface layer (to be further verified). When the Python side is heavily loaded, it stalls the interface, which inadvertently blocks the underlying C++ heartbeat threads.
Solution
Increase
-client_ttlfrom 10 to 30 during initializingmooncake_masterprocess.