Re: startup failures

[Jean-Baptiste Onofré]

Oh ok.

We improved Karaf feature and bundle start level support in Karaf 2.3.3. 
Can you try with latest Karaf version (2.3.6) ?

Regards
JB

On 09/10/2014 08:18 PM, Mike Wilson wrote:
Thanks for your response, Jean-Baptiste,

I'd just like to clarify that I don't consider these "my" bundles :-) but
rather part of Karaf.
Looking at etc/startup.properties I find:

org/apache/aries/blueprint/org.apache.aries.blueprint.core/1.0.1/org.apache.
aries.blueprint.core-1.0.1.jar=20
...
org/apache/karaf/features/org.apache.karaf.features.core/2.3.0/org.apache.ka
raf.features.core-2.3.0.jar=25

and in other files we have configured our own bundles to start at levels
35-100.

If you look at the event list from my previous mail you can see that the
bundle start order is analogous with the above. But, the asynchronous
Blueprint initialization takes place in the opposite order (which is legal
for asynchronous activities):

2014-09-04 21:44:55,081  BlueprintEvent.CREATING
org.apache.karaf.features.core
2014-09-04 21:44:55,100  BlueprintEvent.GRACE_PERIOD
org.apache.karaf.features.core
...
2014-09-04 21:44:55,123  BlueprintEvent.CREATED
org.apache.aries.blueprint.core

Normally I would expect org.apache.karaf.features.core to "wake up" from its
waiting in GracePeriod once org.apache.aries.blueprint.core has been
initialized but this doesn't seem to happen. So,
- are there any known bugs in this area?
- or, isn't the blueprint-ext namespace actually registered by
   org.apache.aries.blueprint.core?
- or, is something completely different causing this problem?

Thanks :)
Mike

Jean-Baptiste Onofré wrote:
Hi Mike,

it looks like your bundles start before some other.

Do you tune the bundles start level ?

The GracePeriod means that the blueprint bundle is waiting
for services
not yet there.

Regards
JB

On 09/10/2014 06:09 PM, Mike Wilson wrote:
We're having intermittent startup problems on Karaf 2.3.0
and I'd just
want to check with you here how I can investigate this
further. What we
see is that sometimes all bundles won't start correctly. I've been
digging through a number of log files and a typical case seems to be
that central bundles get stuck in GracePeriod and this will in turn
affect other bundles.
Here's an excerpt of one of the logs where you can see that two
karaf.features bundles end up in GracePeriod:
2014-09-04 21:44:54,878 BundleEvent.STARTED
org.apache.aries.proxy.impl
2014-09-04 21:44:54,905 BundleEvent.RESOLVED
org.apache.aries.blueprint.cm
2014-09-04 21:44:54,933 BundleEvent.STARTED
org.apache.aries.blueprint.cm
2014-09-04 21:44:54,937 BlueprintEvent.CREATING
org.apache.aries.blueprint.cm
2014-09-04 21:44:55,081 BlueprintEvent.CREATING
org.apache.karaf.features.core
2014-09-04 21:44:55,100 BlueprintEvent.GRACE_PERIOD
org.apache.karaf.features.core
2014-09-04 21:44:55,122 BlueprintEvent.CREATED
org.apache.aries.blueprint.cm
2014-09-04 21:44:55,123 BlueprintEvent.CREATED
org.apache.aries.blueprint.core
2014-09-04 21:44:55,260 BundleEvent.RESOLVED
org.apache.aries.jmx.api
2014-09-04 21:44:55,260 BundleEvent.RESOLVED
org.apache.aries.jmx.core
2014-09-04 21:44:55,492 BundleEvent.STARTED org.apache.aries.jmx.api
2014-09-04 21:44:55,492 BundleEvent.STARTED
org.apache.aries.jmx.core
2014-09-04 21:44:55,525 BundleEvent.RESOLVED
org.apache.aries.jmx.blueprint.api
2014-09-04 21:44:55,526 BundleEvent.STARTED
org.apache.aries.jmx.blueprint.api
2014-09-04 21:44:55,560 BundleEvent.RESOLVED
org.apache.aries.jmx.blueprint.core
2014-09-04 21:44:55,575 BundleEvent.STARTED
org.apache.aries.jmx.blueprint.core
2014-09-04 21:44:55,873 BlueprintEvent.CREATING
org.apache.karaf.features.command
2014-09-04 21:44:55,939 BlueprintEvent.GRACE_PERIOD
org.apache.karaf.features.command
2014-09-04 21:44:56,878 BlueprintEvent.CREATING
org.apache.karaf.deployer.features
2014-09-04 21:44:56,905 BlueprintEvent.CREATED
org.apache.karaf.deployer.features
The first one is waiting for

namespace=http://aries.apache.org/blueprint/xmlns/blueprint-ex
t/v1.0.0:
2014-09-04 21:44:55,100 | DEBUG | rint Extender: 3 |  |  |
ntainer.BlueprintEventDispatcher  136 | 7 -
org.apache.aries.blueprint.core - 1.0.1 | Sending blueprint
container
event BlueprintEvent[type=GRACE_PERIOD,

dependencies=[(&(objectClass=org.apache.aries.blueprint.Namesp
aceHandler)(osgi.service.blueprint.namespace=http://aries.apac
he.org/blueprint/xmlns/blueprint-ext/v1.0.0))]]
for bundle org.apache.karaf.features.core
which I thought should be successfully registered as
org.apache.aries.blueprint.core had BlueprintEvent.CREATED.
Does this make sense? What else could I look at to
understand the issue?
This is in a production system so I would like to
investigate as much as
possible before taking the step to upgrade versions of
Karaf etc (this
is scheduled for a later project).
Thanks
Mike

--
Jean-Baptiste Onofré
jbono...@apache.org
http://blog.nanthrax.net
Talend - http://www.talend.com



--
Jean-Baptiste Onofré
jbono...@apache.org
http://blog.nanthrax.net
Talend - http://www.talend.com