shashank created CAMEL-27285:
--------------------------------

             Summary: camel-spring - after a nested PROPAGATION_REQUIRES_NEW 
route of the same policy, the rest of the calling route runs as not transacted
                 Key: CAMEL-27285
                 URL: https://issues.apache.org/jira/browse/CAMEL-27285
             Project: Camel
          Issue Type: Bug
          Components: camel-spring
            Reporter: shashank
            Assignee: shashank


h3. What happens

Route A {{transacted("PROPAGATION_REQUIRES_NEW")}} calls route B 
{{transacted("PROPAGATION_REQUIRES_NEW")}} (the same policy bean) with 
{{to("direct:b")}}. Since CAMEL-8424 the {{TransactionErrorHandler}} of B 
starts a new transaction there. Its {{processInTransaction}} calls 
{{beginTransactedBy(key)}} and {{endTransactedBy(key)}} with the same key as A; 
the unit of work keeps the keys in a Set, so when B ends it removes A's key, 
and {{exchange.isTransacted()}} becomes false for the rest of route A, which 
still runs in A's transaction.

The EIPs that run a step synchronously only for a transacted exchange then hand 
the message to another thread, outside the thread-bound Spring transaction: the 
Delay EIP (asyncDelayed is the default), Throttle, Threads, a delayed 
redelivery (asyncDelayedRedelivery). The steps that run there are not part of 
A's transaction: a database insert is committed on its own and stays when A 
rolls back.

h3. How to reproduce

{{TransactionErrorHandlerNestedRequiresNewTest}} (camel-spring, a small 
{{AbstractPlatformTransactionManager}} that records its transactions and 
supports suspend/resume): A calls B (same REQUIRES_NEW policy), then 
{{delay(1)}} and a step that records the transaction it runs in. 
{{exchange.isTransacted()}} is false after B, and the step after the delay runs 
on a scheduler thread without a transaction, so it is not rolled back with A.

h3. Fix

{{processInTransaction}} marks and unmarks the unit of work only when the key 
was not marked yet.

_Filed with Claude Code on behalf of allthingssecurity._



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

Reply via email to