jackrabbit-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mark Slater (JIRA)" <j...@apache.org>
Subject [jira] Commented: (JCR-318) RepositoryHelper.unregisterRepository() does not allow the repository to be automatically restarted.
Date Thu, 16 Feb 2006 21:15:10 GMT
    [ http://issues.apache.org/jira/browse/JCR-318?page=comments#action_12366685 ] 

Mark Slater commented on JCR-318:
---------------------------------

And that brings us back to the statement in the docs that says:

> Limitations: the application should not be redeployed during the same JVM process to
avoid
> creating duplicate Jackrabbit instances with the same configuration. In case you want
to redeploy
> your application be sure to shutdown the repository when your application is undeployed.
It can
> be done by calling RepositoryImpl.shutdown() (e.g. in the destroy() method of a servlet).

--- from http://incubator.apache.org/jackrabbit/doc/deploy/howto-model1.html

Lets say there's a web app that configured and accessed its private repository with JNDI using
Tomcat's Context approach (i.e. the Reference is only in the web app's Context). A new version
of the .war is dropped in and Tomcat redeploys the web app. If the repository was not shut
down prior to the redeployment, is it possible for the repository to become corrupted, assuming
the new web app also uses JNDI to access the repository?

If the repository can't become corrupted in that situation, I think the docs need to be changed
to avoid confusion... everything else on that page talks about a JNDI referenced repository.

If the repository can become corrupted in that situation, then the docs are correct and RegistryHelper.
unregisterRepository() should be called. Which leads to the problems (NamingException and
duplicate configuration in JNDI and source code) I reported earlier when the web app tries
to use RegistryHelper.registerRepository().


> RepositoryHelper.unregisterRepository() does not allow the repository to be automatically
restarted.
> ----------------------------------------------------------------------------------------------------
>
>          Key: JCR-318
>          URL: http://issues.apache.org/jira/browse/JCR-318
>      Project: Jackrabbit
>         Type: Bug
>   Components: core
>     Versions: 0.9
>  Environment: Tomcat 5.5.x, JNDI
>     Reporter: Mark Slater
>     Assignee: Stefan Guggisberg
>     Priority: Minor
>      Fix For: 1.0

>
> Using JNDI to create a model 1 (embedded) repository as suggested in the documentation
(http://incubator.apache.org/jackrabbit/doc/deploy/howto-model1.html) returns a BindableRepository
instance to the web app. The same documentation page strongly recommends the webapp calls
RepositoryImpl.shutdown() in the servlet's destroy() method. However, since BindableRepository
is restricted to its package, there is no way for the webapp to access the RepositoryImpl.
The RepositoryHelper class provides a unregisterRepository() method which can perform the
shutdown. Unfortunately, the method does not allow the repository to be restarted later on,
forcing a full restart of Tomcat. The exception thrown when trying to access a repository
shut down in this manner in a re-deployed webapp is:
> java.lang.IllegalStateException: repository instance has been shut down
> 	org.apache.jackrabbit.core.RepositoryImpl.sanityCheck(RepositoryImpl.java:481)
> 	org.apache.jackrabbit.core.RepositoryImpl.login(RepositoryImpl.java:1005)
> 	org.apache.jackrabbit.core.RepositoryImpl.login(RepositoryImpl.java:1080)
> 	org.apache.jackrabbit.core.jndi.BindableRepository.login(BindableRepository.java:174)
> 	org.apache.jackrabbit.deployment_test.JackrabbitTest.testRepository(JackrabbitTest.java:212)
> 	org.apache.jackrabbit.deployment_test.JackrabbitTest.doGet(JackrabbitTest.java:80)
> 	javax.servlet.http.HttpServlet.service(HttpServlet.java:689)
> 	javax.servlet.http.HttpServlet.service(HttpServlet.java:802)
> It may be useful to merge the transient capabilities of TransientRepository into BindableRepository,
perhaps making it a factory configuration option. I imagine there's a performance hit to opening
a repository, in a shared deployment environment, where the lifecycle is controlled by the
application server, it would probably be best to not have the repository be transient.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
   http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see:
   http://www.atlassian.com/software/jira


Mime
View raw message