[ 
https://issues.apache.org/jira/browse/CAMEL-27284?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Claus Ibsen updated CAMEL-27284:
--------------------------------
    Fix Version/s: 4.24.0

> camel-jpa - JpaMessageIdRepository: a concurrent add of the same message id 
> fails instead of returning false on PostgreSQL and SQL Server
> -----------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: CAMEL-27284
>                 URL: https://issues.apache.org/jira/browse/CAMEL-27284
>             Project: Camel
>          Issue Type: Bug
>          Components: camel-jpa
>            Reporter: shashank
>            Assignee: shashank
>            Priority: Major
>             Fix For: 4.24.0
>
>
> h3. What happens
> CAMEL-24132 (M4) made {{JpaMessageIdRepository.add}} return false when its 
> insert fails on the unique key because another caller (node, thread) inserted 
> the same message id at the same time. It recognises the failure only as 
> {{SQLIntegrityConstraintViolationException}} or {{EntityExistsException}} in 
> the cause chain:
> # The PostgreSQL driver ({{PSQLException}}, SQLState 23505) and the SQL 
> Server driver (SQLState 23000) throw a plain {{SQLException}}, and Hibernate 
> throws its {{ConstraintViolationException}} (a {{PersistenceException}}, not 
> an {{EntityExistsException}}) with it as the cause, so on these databases 
> {{add}} throws {{PersistenceException("Something went wrong ...")}}: the 
> duplicate message fails (error handler, redelivery, dead letter channel) 
> instead of being filtered. H2, Derby, MySQL and Oracle throw the subclass, 
> which is why the module tests pass.
> # The check is inside the transaction: when the EntityManager is the 
> transaction's ({{sharedEntityManager=true}}), the failed flush marks the 
> transaction rollback-only, so the commit throws after the catch returned 
> false, and {{add}} fails on every database.
> {{JpaKeyValueRepository}} (CAMEL-26736) already checks the SQLState class 23 
> and recognises the duplicate outside the transaction.
> h3. How to reproduce
> {{JpaMessageIdRepositoryConcurrentInsertTest}}: the flush throws what 
> Hibernate throws for a PostgreSQL unique violation 
> (ConstraintViolationException -> SQLException state 23505): add throws; same 
> with state 23000 (SQL Server); and with a transaction whose commit fails 
> after the duplicate was recognised.
> h3. Fix
> The SQLState class 23 counts as a constraint violation (as in 
> JpaKeyValueRepository), and a duplicate recognised in the transaction stays a 
> duplicate when the commit then fails.
> _Filed with Claude Code on behalf of allthingssecurity._



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

Reply via email to