karaf-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Fabian Lange <fabian.la...@codecentric.de>
Subject Re: Null Pointer on Feature install
Date Tue, 26 Jan 2016 21:57:59 GMT
Hi JB,
it turns out this is most likely the old Issue I reported once. Having now
more insights into what should be done, I propose this solution:
https://github.com/apache/karaf/pull/138

It basically ensures that only one provision is concurrently executing.
What do you think?

Fabian


On Tue, Jan 26, 2016 at 10:19 PM, Fabian Lange <fabian.lange@codecentric.de>
wrote:

> Hi JB,
> not really reproducable. We had this before, and I then introduced a
> locking abstraction around the features manager. This is a programmatic
> installation of features. So I assume that features are no longer installed
> concurrently.
>
> However it might race with karaf/felix bootstrap and feature installation.
> I only synchronized my installation
>
> What I do, is that I add a feature repository, then do a foreach and add
> the requirements, then fire off the install
>
> Fabian
>
>
> On Tue, Jan 26, 2016 at 10:16 PM, Jean-Baptiste Onofré <jb@nanthrax.net>
> wrote:
>
>> It sounds like concurrent update of feature state in the store.
>>
>> Does it mean users install/uninstall multiple features in the same time ?
>>
>> Any use case to reproduce it ?
>>
>> Thanks
>> Regards
>> JB
>>
>>
>> On 01/26/2016 09:18 PM, Fabian Lange wrote:
>>
>>> Hi all,
>>> I am seeing a significant amount of following stacktraces reported by our
>>> users. they seem to go away with restarting, so it looks like a leak or
>>> race condition to me:
>>>
>>> java.lang.NullPointerException
>>>
>>>         at
>>> java.io.FileOutputStream.<init>(FileOutputStream.java:203)[:1.8.0_72]
>>>
>>>         at
>>> java.io.FileOutputStream.<init>(FileOutputStream.java:162)[:1.8.0_72]
>>>
>>>         at
>>> org.apache.karaf.features.internal.osgi.Activator$1.getOutputStream(Activator.java:197)[7:org.apache.karaf.features.core:4.0.4]
>>>
>>>         at
>>> org.apache.karaf.features.internal.service.StateStorage.save(StateStorage.java:56)[7:org.apache.karaf.features.core:4.0.4]
>>>
>>>         at
>>> org.apache.karaf.features.internal.service.FeaturesServiceImpl.saveState(FeaturesServiceImpl.java:297)[7:org.apache.karaf.features.core:4.0.4]
>>>
>>>         at
>>> org.apache.karaf.features.internal.service.FeaturesServiceImpl.saveState(FeaturesServiceImpl.java:1154)[7:org.apache.karaf.features.core:4.0.4]
>>>
>>>         at
>>> org.apache.karaf.features.internal.service.Deployer.deploy(Deployer.java:764)[7:org.apache.karaf.features.core:4.0.4]
>>>
>>>         at
>>> org.apache.karaf.features.internal.service.FeaturesServiceImpl.doProvision(FeaturesServiceImpl.java:1089)[7:org.apache.karaf.features.core:4.0.4]
>>>
>>>         at
>>> org.apache.karaf.features.internal.service.FeaturesServiceImpl$1.call(FeaturesServiceImpl.java:985)[7:org.apache.karaf.features.core:4.0.4]
>>>
>>>         at
>>> java.util.concurrent.FutureTask.run(FutureTask.java:266)[:1.8.0_72]
>>>
>>>         at
>>> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)[:1.8.0_72]
>>>
>>>         at
>>> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)[:1.8.0_72]
>>>
>>>         at java.lang.Thread.run(Thread.java:745)[:1.8.0_72]
>>>
>>>
>>> the NPE is caused by bundleContex.getDataFile(state.json) returning null.
>>> The javadoc says that this could be possible, but because it works most
>>> of
>>> the time I speculate this might be actally a race condition, or maybe
>>> leaked filehandles somewhere.
>>> Anybody got pointers or ideas?
>>> I am willing to make a fix, but I need to understand this more
>>>
>>> Fabian
>>>
>>>
>> --
>> Jean-Baptiste Onofré
>> jbonofre@apache.org
>> http://blog.nanthrax.net
>> Talend - http://www.talend.com
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message