[
https://issues.apache.org/jira/browse/IGNITE-25241?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-25241:
---------------------------------------
Description: On node stop, TableManager only waits for replicas to stop for
a limited amount of time (currently, it's 30 seconds). If a Raft node cannot be
stopped for some reason, TableManager will give up attempts to stop partitions
and then it will allow the node to proceed with the stop routine. As a result,
Raft log storage factory might be stopped while some Raft node is still alive.
The Raft node will try to access its Raft log, which might crash the JVM as
RocksDB instance related to log storage is already shut down.
> Protect Raft log storage factory with busy lock
> -----------------------------------------------
>
> Key: IGNITE-25241
> URL: https://issues.apache.org/jira/browse/IGNITE-25241
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
>
> On node stop, TableManager only waits for replicas to stop for a limited
> amount of time (currently, it's 30 seconds). If a Raft node cannot be stopped
> for some reason, TableManager will give up attempts to stop partitions and
> then it will allow the node to proceed with the stop routine. As a result,
> Raft log storage factory might be stopped while some Raft node is still
> alive. The Raft node will try to access its Raft log, which might crash the
> JVM as RocksDB instance related to log storage is already shut down.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
