shashank created CAMEL-26573:
--------------------------------
Summary: camel-vertx-websocket - a consumer starting while the
last consumer on its host and port stops keeps being served after its own stop
Key: CAMEL-26573
URL: https://issues.apache.org/jira/browse/CAMEL-26573
Project: Camel
Issue Type: Bug
Components: camel-vertx-websocket
Reporter: shashank
Assignee: shashank
Found with a TLA+ model of the shared host registry of
{{VertxWebsocketComponent}} (consumers on one host and port, each starting and
stopping on its own thread), then reproduced on the real classes with embedded
Vert.x on localhost.
Since CAMEL-24788 the consumers on one host and port share one
{{VertxWebsocketHost}}, which leaves the registry once it serves no consumer.
{{connectConsumer}} gets the host ({{computeIfAbsent}}) and connects to it
afterwards, outside the map; a {{disconnectConsumer}} in between
({{host.disconnect}}, then {{computeIfPresent}} removing a host that serves no
consumer) removes the host the consumer is about to connect to. The consumer
then connects to the removed host and starts its server:
- its own stop finds no host in the registry, so its path stays served after
the stop (messages reach the route of a stopped consumer) and the server keeps
the port until the Vert.x instance is closed (when the CamelContext stops, or
never for a Vert.x instance the application owns);
- the next consumer on that host and port gets a second server on the same port.
Trace: b gets host H (a is still on it); a stops: last route, server closed, H
removed from the registry; b connects to H and starts a server. The disconnect
can also be a consumer failing the {{sslContextParameters}} check of
CAMEL-26109, whose failed start stops it while another consumer has just
created the host.
Route operations of a CamelContext hold its lock, but
{{ThrottlingInflightRoutePolicy}} and {{ThrottlingExceptionRoutePolicy}} stop
and start this consumer (it cannot be suspended) without it, so two routes with
such a policy on one port, or a policy and a route operation, run both at the
same time. A stress loop on main (start one consumer while the last other one
stops) ended with the consumer on a removed host in 295 of 300 runs.
Fix: {{connectConsumer}} and {{disconnectConsumer}} run under a lock per host
and port.
Affected: main, 4.23.x and 4.22.x (the branches with CAMEL-24788); the failed
start path needs CAMEL-26109, which is on main only.
_Filed with Claude Code on behalf of allthingssecurity._
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
