shashank created CAMEL-26302:
--------------------------------
Summary: camel-ehcache - removing the route that consumes from a
cache closes the cache that the other routes still use, so every cache action
fails with "State is UNINITIALIZED"
Key: CAMEL-26302
URL: https://issues.apache.org/jira/browse/CAMEL-26302
Project: Camel
Issue Type: Bug
Components: camel-ehcache
Reporter: shashank
Assignee: shashank
When a route is removed, the core stops and shuts down the endpoint of its
consumer ({{RouteService.doShutdown}}), also when other routes still use the
same endpoint. For an ehcache endpoint this releases caches that other routes
still hold (main c536c99caaaa):
* {{EhcacheEndpoint.doStop}} (line 112) releases the endpoint's reference on
its {{EhcacheManager}}. Without {{cacheManager}}, {{cacheManagerConfiguration}}
or {{configurationUri}} the endpoint has its own cache manager, so this is the
last reference and it closes the {{CacheManager}} with all its caches.
* {{EhcacheEndpoint.doShutdown}} (line 118) shuts the {{EhcacheManager}} down,
which closes every cache it created without a configuration
({{EhcacheManager.doShutdown}}, line 66). With {{cacheManager}},
{{cacheManagerConfiguration}} or {{configurationUri}} the manager is shared by
all the endpoints that use it, so this closes the caches of other endpoints too.
The producers of the other routes keep the {{Cache}} they got when they were
created, so from then on every action (PUT, GET, ...) fails with
{{IllegalStateException: State is UNINITIALIZED}}, also after the removed route
is added again. Nothing else is logged.
Examples:
# {{from("ehcache:orders?configuration=#ordersConfig")}} (a route that reacts
to cache events) and {{to("ehcache:orders?configuration=#ordersConfig")}} in
other routes. As each endpoint uri without a cache manager option has its own
cache manager (CAMEL-26237), a consumer and a producer of the same cache must
use the same uri, so they share the endpoint. Removing the event route breaks
the producer routes.
# {{cacheManager=#myManager}} on all endpoints, with some caches configured in
the manager and others created on demand: removing a route that consumes events
of a configured cache closes the on-demand caches of the other endpoints.
A route is removed by a route reload (dev mode, route watcher), a removed route
template or Kamelet, JMX or the control bus.
*Fix*: the producers and consumers hold a reference on the {{EhcacheManager}}
while they are started (as the endpoints already do), so the cache manager is
closed by the last of them; an endpoint shuts the {{EhcacheManager}} down only
when nothing uses it any more, and the last reference also closes the caches
created without a configuration (an endpoint shut down with a removed route is
not shut down again). {{EhcacheManagerTest}} now counts the producers too (4
references for two endpoints with their producers instead of 2).
*Tests*: new {{EhcacheRemoveRouteTest}}: {{removeEventRouteOfSameEndpoint}},
{{removeEventRouteOfSharedCacheManager}}, {{removeEventRouteThenAddItAgain}}
and {{removeEventRouteThenStopClosesCacheCreatedOnDemand}} fail on main (3
runs, each with {{IllegalStateException: State is UNINITIALIZED}} on the first
action after the removal) and pass with the fix. With the fix camel-ehcache has
77 tests, 0 failures.
*How it was found*: a code sweep for endpoints whose {{doStop}} releases
something that other routes on the same uri still use, after a random
lifecycle-sequence harness found the same kind of defect in camel-disruptor
(CAMEL-26286).
*Affected*: main, camel-4.22.x, camel-4.18.x and camel-4.14.x (same
{{doStop}}/{{doShutdown}}; read, not run).
*Related*: CAMEL-26237 (each endpoint uri without a cache manager option builds
its own cache manager; a consumer on a cache without configuration gets no
events). Sharing the default cache manager per component, as proposed there,
would make the second case above the default one.
_Filed with Claude Code on behalf of allthingssecurity._
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
