Tanuj Khurana created PHOENIX-7988:
--------------------------------------
Summary: Idle co-active RegionServer stays wedged in
STORE_AND_FORWARD, pinning the group at ACTIVE_NOT_IN_SYNC
Key: PHOENIX-7988
URL: https://issues.apache.org/jira/browse/PHOENIX-7988
Project: Phoenix
Issue Type: Sub-task
Reporter: Tanuj Khurana
Assignee: Tanuj Khurana
An idle RegionServer can stay stuck in STORE_AND_FORWARD mode indefinitely,
wedging the whole replication group at ACTIVE_NOT_IN_SYNC, because its
promotion guard scans a shared directory instead of a per-RS signal. A RS in
STORE_AND_FORWARD promotes back to SYNC_AND_FORWARD via processNoMoreRoundsLeft
only when the caught-up guard passes. That guard scanned the next round's shard
— a shared directory holding
every active RS's live OPENFORWRITE rotation writer. Because that shard is
essentially never empty while any peer is writing, an idle RS which never
processes any file in the OUT directory can remain pinned in STORE_AND_FORWARD,
wedging the group at ACTIVE_NOT_IN_SYNC indefinitely.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
