wei liu 2232dfc3de

fix: Prevent Close from hanging on etcd reconnection (#45622) ...

issue: #45623
When etcd reconnects, the DataCoord rewatches DataNodes and calls
ChannelManager.Startup again without closing the previous instance. This
causes multiple contexts and goroutines to accumulate, leading to Close
hanging indefinitely waiting for untracked goroutines.

Root cause:
- Etcd reconnection triggers rewatch flow and calls Startup again
- Startup was not idempotent, allowing repeated calls
- Multiple context cancellations and goroutines accumulated
- Close would wait indefinitely for untracked goroutines

Changes:
- Add started field to ChannelManagerImpl
- Refactor Startup to check and handle restart scenario
- Add state check in Close to prevent hanging

---------

Signed-off-by: Wei Liu <wei.liu@zilliz.com>

2025-11-19 12:49:06 +08:00

..

allocator

enhance: [2.5] Use v2 package name for pkg module (#40117)

2025-02-23 00:46:01 +08:00

coordinator/coordclient

enhance: [2.5] Use v2 package name for pkg module (#40117)

2025-02-23 00:46:01 +08:00

core

enhance: optimize term expr performance (#45490)

2025-11-19 11:51:06 +08:00

datacoord

fix: Prevent Close from hanging on etcd reconnection (#45622)

2025-11-19 12:49:06 +08:00

datanode

fix: Fix bulk import with autoid (#44601)

2025-10-09 14:51:58 +08:00

distributed

enhance: [2.5] Add granular flush targets support for FlushAll operation (#44431)

2025-09-28 10:37:06 +08:00

flushcommon

fix: [2.5] flow graph should free function resource after all node close (#42731) (#42775)

2025-06-17 14:40:39 +08:00

http

enhance: [2.5] Make Web UI toggleable via config (#42815)

2025-06-18 12:20:39 +08:00

indexnode

fix: [2.5] Reduce task slot for standalone to 1/4 of normal datanode (#42809)

2025-06-20 16:38:46 +08:00

json

feat: fp32 vector to fp16/bf16 vector conversion for RESTful API (#37556)

2024-11-24 17:46:33 +08:00

kv

enhance: [2.5] Skip remove op if key in save set (#43426)

2025-07-18 17:36:58 +08:00

metastore

fix: [2.5] Use proto.Equal to compare default value of field (#43813) (#43832)

2025-08-12 18:09:44 +08:00

mocks

enhance: [2.5] Add granular flush targets support for FlushAll operation (#44431)

2025-09-28 10:37:06 +08:00

parser/planparserv2

enhance: Use GEOSGeometry directly to skipping the gdal to geos conversion (#44519)

2025-09-24 18:30:06 +08:00

proxy

enhance: [2.5] skip check source id (#45383)

2025-11-07 15:21:42 +08:00

querycoordv2

fix: [2.5] Fix infinite loop in ResourceManager recovery process (#45563)

2025-11-17 15:19:39 +08:00

querynodev2

enhance: [2.5]Make GeometryCache an optional configuration (#45197)

2025-11-03 20:31:34 +08:00

registry

enhance: [2.5] Use v2 package name for pkg module (#40117)

2025-02-23 00:46:01 +08:00

rootcoord

fix: [2.5] Cleanup collection metrics after dropped on rootcoord (#44511) (#44525)

2025-09-24 09:56:06 +08:00

storage

fix: [2.5] Fix import null geometry data (#45163)

2025-10-31 12:34:10 +08:00

streamingcoord

enhance: remove redundant resource key watch operation, just keep consistency of wal (#40503)

2025-03-18 15:46:25 +08:00

streamingnode

fix: [2.5] Fix ants.Pool goroutine leak (#41893)

2025-05-16 19:12:22 +08:00

tso

enhance: [2.5] Use v2 package name for pkg module (#40117)

2025-02-23 00:46:01 +08:00

types

enhance: [2.5] Use v2 package name for pkg module (#40117)

2025-02-23 00:46:01 +08:00

util

enhance: [2.5] skip check source id (#45383)

2025-11-07 15:21:42 +08:00

.mockery.yaml

enhance: broadcast with event-based notification (#39550)

2025-02-07 11:50:50 +08:00