shashank created CAMEL-26755:
--------------------------------
Summary: camel-support - AbstractCamelClusterService: a view whose
start fails keeps its reference, so it is never started again and never stopped
Key: CAMEL-26755
URL: https://issues.apache.org/jira/browse/CAMEL-26755
Project: Camel
Issue Type: Bug
Components: camel-core
Reporter: shashank
Assignee: shashank
{{AbstractCamelClusterService.getView}} takes a reference on the view of the
namespace and starts the view on the first reference. The reference count is
increased before the view is started, and nothing gives it back when the start
throws. The caller gets the exception and holds no reference, but the count
stays at 1:
* the next {{getView}} only increases the count, so the view is not started: a
{{master:}} route started again (by an operator or the supervising route
controller) is reported started but the node never takes part in the election,
also as the only node and after the cause of the failure is gone; a clustered
route added later on the same namespace is never started either;
* {{releaseView}} can no longer bring the count to 0, so the view is never
stopped;
* after the cluster service is stopped and started again (CamelContext restart,
or the JMX operations), it starts every view with references, including this
one that no route uses, so the node can take the leadership of a namespace it
does not use and the route of the other nodes never runs (what CAMEL-25089 and
CAMEL-26577 made sure does not happen for released views).
A view start can fail when the backend is not reachable yet (for example
{{ConsulClusterView}} creates a session, {{JGroupsRaftClusterView}} connects,
{{InfinispanRemoteClusterView}} writes to the remote cache), or on an invalid
setting ({{FileLockClusterView}} rejects an {{acquireLockInterval}} below 1
millisecond).
Expected: when the view cannot be started, the reference is given back, so the
next {{getView}} starts the view again, and the reference count matches the
users of the view.
Suggested fix: in the view holder, give the reference back (without changing
anything else) when the start throws, and rethrow.
Found with Specula (TLA+ model checked against traces of the real code) and
confirmed with a test on the real classes.
_Filed with Claude Code on behalf of allthingssecurity._
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
