[
https://issues.apache.org/jira/browse/KARAF-6074?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16808937#comment-16808937
]
J. Brébec commented on KARAF-6074:
----------------------------------
This issue happen only on the first start, and can happen randomly because it's
a race condition between two threads : the thread activating the osgi framework
(and calling the activators), and the thread doing the provisioning of features.
It happens on the following scenario :
# First start
# The activator of the bundle feature-core is called. It's not blocking : it
spawn a new thread to do the provisioning. the FeaturesService is registered in
osgi here
# Let's assume the thread created in the previous step can't execute (or is
slow), maybe because the hardware have only one core, or whatever reason : the
FeaturesServiceImpl.lastDeploymentStatus is DEPLOYMENT_FINISHED. You can
simulate this step easily with a debugger, with a break-point in
[executor.submit|https://github.com/apache/karaf/blob/karaf-4.2.4/features/core/src/main/java/org/apache/karaf/features/internal/service/FeaturesServiceImpl.java#L994]
# The activator of the bundle "feature-deployment" is called. It can access
the FeaturesService because of step 2.
# It then register a deployment listener : this listener is called back with
the status DEPLOYMENT_FINISHED. Because step 3 is not finished, the root region
doesn't exist and the listener fails with a NPE
I can provide you other details/artefacts, but I don't have access for now to a
running instance.
> Race condition between the FeaturesService and FeatureDeploymentListener
> ------------------------------------------------------------------------
>
> Key: KARAF-6074
> URL: https://issues.apache.org/jira/browse/KARAF-6074
> Project: Karaf
> Issue Type: Bug
> Components: karaf
> Affects Versions: 4.2.2, 4.2.4
> Environment: Karaf 4.2.2 Windows 7 and Equinox
> Reporter: J. Brébec
> Assignee: Grzegorz Grzybek
> Priority: Critical
> Fix For: 4.2.3
>
>
> On the first start of a custom Karaf container (4.2.2), the logs shows a log
> of NPE in FeatureDeploymentListener.
> After some analysis of this Exception, it's look like a race condition
> between the FeaturesService and the FeatureDeploymentListener :
> # The FeaturesService starts and launch a provisioning in another thread
> # a FeatureDeploymentListener is registered, and call bundleChanged for
> every bundle installed (the startup bundles)
> # It calls FeaturesService.state.requirements and get an empty map
> # Updating the root regions in the requirements map fails with an NPE
> # Some times later, the deployment task launched at step 1 is started, and
> FeaturesService.saveState() is called : state.requirements."root region" is
> initialised
> The step 5 is executed in another thread, so it can happens before step 2
> (and it works) or after, causing this NPE
> Moreover, some methods in FeaturesService assume that
> state.requirements."root region" exists. There are probably others npe here.
>
> {code:java}
> 14:37:55.668 ERROR [activator-1-thread-2] Unable to update deployed features
> for bundle: org.eclipse.osgi - 3.12.100.v20180210-1608
> java.lang.NullPointerException: null
> at
> org.apache.karaf.deployer.features.FeatureDeploymentListener.bundleChanged(FeatureDeploymentListener.java:247)
> [25:org.apache.karaf.deployer.features:4.2.2]
> at
> org.apache.karaf.deployer.features.FeatureDeploymentListener.init(FeatureDeploymentListener.java:95)
> [25:org.apache.karaf.deployer.features:4.2.2]
> at
> org.apache.karaf.deployer.features.osgi.Activator.doStart(Activator.java:52)
> [25:org.apache.karaf.deployer.features:4.2.2]
> at org.apache.karaf.util.tracker.BaseActivator.run(BaseActivator.java:292)
> [25:org.apache.karaf.deployer.features:4.2.2]
> at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
> [?:?]
> at java.util.concurrent.FutureTask.run(FutureTask.java:266) [?:?]
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> [?:?]
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> [?:?]{code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
