Skip to content

[Bugfix] Report serving after HA rebuild gate opens - #16

Merged
ShuweiShen772 merged 1 commit into
v0.3.12_xllm_patchfrom
bugfix/ha-rebuild-health-state
Nov 14, 2026
Merged

ShuweiShen772 merged 1 commit into
v0.3.12_xllm_patchfrom
bugfix/ha-rebuild-health-state

Conversation

@ShuweiShen772

Copy link
Copy Markdown
Collaborator

Description

When --enable_ha_rebuild_gate=true is enabled on v0.3.12_xllm_patch, the rebuild gate can transition the promoted WrappedMasterService to SERVING while MasterAdminServer remains in leader_warmup.

The /health endpoint then reports service_ready=true but ha_state=leader_warmup. Health checks that require ha_state=serving may consequently treat the active leader as unhealthy and time out or refuse to restart it.

Derive the health snapshot state from the service's actual serving state: while the runtime state is leader_warmup, report serving once the delegated service is serving. The endpoint continues to report leader_warmup while the rebuild gate is closed.

Suggested Merge Priority

  • P2 — Normal: Routine bug fix for HA health reporting.

Module

  • Mooncake Store

Type of Change

  • Bug fix

How Has This Been Tested?

Test commands:

clang-format --dry-run --Werror \
  mooncake-store/src/master_admin_service.cpp \
  mooncake-store/tests/master_admin_server_test.cpp
git diff --check

Signed-off-by: shenshuwei.3 <shenshuwei.3@jd.com>
@ShuweiShen772
ShuweiShen772 merged commit 20039cb into v0.3.12_xllm_patch Nov 14, 2026
1 check failed
@ShuweiShen772
ShuweiShen772 deleted the bugfix/ha-rebuild-health-state branch November 14, 2026 10:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant