[
https://issues.apache.org/jira/browse/IGNITE-19905?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kirill Gusakov updated IGNITE-19905:
------------------------------------
Description:
*Motivation*
It is possible that handleChangeStableAssignmentEvent() gets an event such that
for the corresponding revision the table is already removed, so an NPE happens
when trying to destroy a partition storage (at the moment it handled by
workaround IGNITE-19906).
*Definition of done*
- The logic about partition raft group and storage cleanup is not event
executed, if the table is not exist already.
*Implementation notes*
- We must check if table still exists at the current revision firstly and only
then execute the cleanup logic
was:It is possible that handleChangeStableAssignmentEvent() gets an event
such that for the corresponding revision the table is already removed, so an
NPE happens when trying to destroy a partition storage.
> Race between stable assignments application and table removal
> -------------------------------------------------------------
>
> Key: IGNITE-19905
> URL: https://issues.apache.org/jira/browse/IGNITE-19905
> Project: Ignite
> Issue Type: Bug
> Reporter: Roman Puchkovskiy
> Assignee: Kirill Gusakov
> Priority: Major
> Labels: ignite-3
> Fix For: 3.0.0-beta2
>
>
> *Motivation*
> It is possible that handleChangeStableAssignmentEvent() gets an event such
> that for the corresponding revision the table is already removed, so an NPE
> happens when trying to destroy a partition storage (at the moment it handled
> by workaround IGNITE-19906).
> *Definition of done*
> - The logic about partition raft group and storage cleanup is not event
> executed, if the table is not exist already.
> *Implementation notes*
> - We must check if table still exists at the current revision firstly and
> only then execute the cleanup logic
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
