Re: Concurrent writing to a JCR repository

hi behrang,

On 10/1/06, Behrang Saeedzadeh <behrangsa <at> gmail.com> wrote:
> Hi Again,
>
> I changed a code a bit and surprisingly I'm getting a new exception:
>
> javax.jcr.InvalidItemStateException:
> 6aa4218b-93a6-4a34-b078-42e843baa5b2: the item cannot be saved because
> it has been modified externally.

this exception indicates that you have 2 or more separate sessions
that try to modify
the *same* item (in your case adding nodes to the /positions node).

you can avoid such situations if you lock the node before
you start modifiying it.

cheers
stefan

>         at org.apache.jackrabbit.core.ItemImpl.getTransientStates(ItemImpl.java:387)
>         at org.apache.jackrabbit.core.ItemImpl.save(ItemImpl.java:1054)
>         at org.apache.jackrabbit.core.SessionImpl.save(SessionImpl.java:805)
>         at org.behrang.practice.jsr170.UserThread.run(Main.java:66)
>         at java.lang.Thread.run(Thread.java:595)
>
> Here's the changed code:
>
> --- --- --- ---

Re: Concurrent writing to a JCR repository

hi behrang,

On 9/30/06, Behrang Saeedzadeh <behrangsa <at> gmail.com> wrote:
> Hi,
>
> I have written a very simple performance test for Jackrabbit that
> creates 10 threads each saving 1000 simple one-property nodes to the
> repository.
>
> When I run the test, an exception is thrown:
>
> javax.jcr.ItemNotFoundException:

strange, you should be getting an InvalidItemStateException instead
(see my other
reply in this thread).

please open a jira  bug if you're able to reproduce this issue.

> 0cada011-dc27-4531-bfcc-e8dde67532d9/{http://www.jcp.org/jcr/1.0}primaryType
>         at org.apache.jackrabbit.core.ItemManager.createItemInstance(ItemManager.java:464)
>         at org.apache.jackrabbit.core.ItemManager.getItem(ItemManager.java:320)
>         at org.apache.jackrabbit.core.ItemImpl.restoreTransientItems(ItemImpl.java:708)
>         at org.apache.jackrabbit.core.ItemImpl.save(ItemImpl.java:1208)
>         at org.apache.jackrabbit.core.SessionImpl.save(SessionImpl.java:805)
>
>
>
> When I reduce the thread count to 1, the program completes with no
> problem so the problem should be that I am using the JCR API

how does authorization works in jackrabbit?

I would like to learn more how JackRabbit handles authorization to the repository content. Especially
when it comes to searching.
I've seen that there is a plugable AccessManager but I'm not sure how this is used to filter the search result.
Can somebody point me to some documentation or where in the code to look?

Re: how does authorization works in jackrabbit?

Hi Magnus.
Firstly, I think by authorisation in this context you mean access control?
(authorisation is via JAAS I believe).

I worked out a little by plugging in a dummy AccessManager and looked at the
callbacks. I did not try it with searching though, but I gather it will use
it to honour the access rules with returning results - thus your acess list
lookups better be really fast and in memory !

Anyone else with real experience, please do share !

On 10/1/06, Magnus Grimsell <magnus.grimsell <at> idainfront.se> wrote:
>
> I would like to learn more how JackRabbit handles authorization to the
> repository content. Especially when it comes to searching.
> I've seen that there is a plugable AccessManager but I'm not sure how this
> is used to filter the search result.
> Can somebody point me to some documentation or where in the code to look?
>

Subject: Re: Isn't Jackrabbit released

Yeah Jukka said he was going to update it !

He will, he will... patience !

But it is released ;)

On 9/29/06, Behrang Saeedzadeh <behrangsa <at> gmail.com> wrote:
>
> This is an excerpt from the getting started page:
>
> "Since Jackrabbit is not yet released, in order to get your hands on a
> running Jackrabbit instance you have to download and build it
> yourself. Directions for doing so can be found in the Building
> Jackrabbit section. Once Jackrabbit is up and running, you can take
> your first hops."
>
> I guess it needs to be updated 
>
> Regards,
> --
> "We can only see a short distance ahead,
> but we can see plenty there
> that needs to be done." - Alan Turing
>
> "Science is a differential equation. Religion
> is a boundary condition" - Alan Turing
>
> Behrang Saeedzadeh
> http://www.jroller.com/page/behrangsa
> http://my.opera.com/behrangsa

SV: how does authorization works in jackrabbit?

Hi Michael,

Yes, I do mean access control.

After looking through the code it seems like the AccessManager will get invoked for every hit in the result list.
I would like to implement a simple access control scheme based on node type. From a performance point of view 
it would be good to avoid a check for every hit. On the other hand its simple to implement a custom
QueryManager that
narrows the search to only include the node types that the subject has access rights to, right? 
Is there a suitable place to do the access control check that does not get called when performing searches?

/Magnus

-----Ursprungligt meddelande-----
Från: Michael Neale [mailto:michael.neale <at> gmail.com]
Skickat: den 1 oktober 2006 19:00
Till: users <at> jackrabbit.apache.org
Ämne: Re: how does authorization works in jackrabbit?

Hi Magnus.
Firstly, I think by authorisation in this context you mean access control?
(authorisation is via JAAS I believe).

I worked out a little by plugging in a dummy AccessManager and looked at the
callbacks. I did not try it with searching though, but I gather it will use
it to honour the access rules with returning results - thus your acess list
lookups better be really fast and in memory !

Anyone else with real experience, please do share !

Comparing an RDBMS to JCR

Hi,

The concurrent update capability of JCR is a little bit confusing to me.

In an RDBMS, when I execute the following code sinppet simultaneously by a
few threads, no problem rises:

  tx.being();
  executeUpdate("insert into positions(id, name, parentId) values (?, ?,
1)");
  tx.commit();

In JCR, the equivalent to this seems to be:

  tx.begin();
  Node positions = session.getRootNode().getNode("positions");
  Node pos = positions.addNode("position");
  pos.setProperty("name", "some name");
  tx.commit();

But apparantly this throws an InvalidItemStateException (am I right?) when
multiple sessions concurrently execute this snippet of code. If this is
true, then how can one handle the aforementioned SQL operation in JCR?

Regards,
Behi (via Nabble 
--

-- 
View this message in context: http://www.nabble.com/Comparing-an-RDBMS-to-JCR-tf2366850.html#a6593713
Sent from the Jackrabbit - Users mailing list archive at Nabble.com.

Re: Comparing an RDBMS to JCR

Behi,

I think the only thing you're missing is locking. Many (most) RDBMS have
some level of built-in/implicit locking of the table or, better yet the row,
for write operations. This is not true of JCR,  you must explicitly
lock/unlock your nodes to prevent concurrent update attempts.

HTH,
Doug

On 10/1/06, behrangsa <behrangsa <at> gmail.com> wrote:
>
>
> Hi,
>
> The concurrent update capability of JCR is a little bit confusing to me.
>
> In an RDBMS, when I execute the following code sinppet simultaneously by a
> few threads, no problem rises:
>
>   tx.being();
>   executeUpdate("insert into positions(id, name, parentId) values (?, ?,
> 1)");
>   tx.commit();
>
> In JCR, the equivalent to this seems to be:
>
>   tx.begin();
>   Node positions = session.getRootNode().getNode("positions");
>   Node pos = positions.addNode("position");

Re: Comparing an RDBMS to JCR

Doug,

Thanks for the reply. Does it imply that using JCR has a __serious__
performance penalty compared to direct access to the database via JDBC, and
even using ORM tools such as EJB3?

BTW - There's a good comparison of JCR to File Systems and DBs available at
http://www.day.com/site/en/index/products/content-centric_infrastructure/content_repository/crx_faq.html

But I am still confused as to when one should consider using JCR* as a
fundamental component of an enterprise project? For what kind of projects
does it make sense to use JCR? Once we consider using JCR, should we
store/retrieve everything via JCR? Or only a portion of the project should
be so?

*JCR or any other CMS (?) API

Regards,
Behi

Doug Douglass-2 wrote:
> 
> Behi,
> 
> I think the only thing you're missing is locking. Many (most) RDBMS have
> some level of built-in/implicit locking of the table or, better yet the
> row,
> for write operations. This is not true of JCR,  you must explicitly
> lock/unlock your nodes to prevent concurrent update attempts.

jta transaction without using jca

I would like to configure jackrabbit with db based file system and persistence manager that participates
in the ongoing jta transaction.
I tried out the JNDIDatabasePersistenceManager but since it does explicit commits i ran into trouble.

Caused by: java.sql.SQLException: The method 'commit' cant be called when a global transaction is active.
	at oracle.oc4j.sql.spi.ConnectionHandle.oc4j_throwSQLExceptionIfGlobalTxActive(ConnectionHandle.java:407)
	at oracle.oc4j.sql.spi.ConnectionHandle.commit(ConnectionHandle.java:122)
	at org.apache.jackrabbit.core.state.db.DatabasePersistenceManager.store(DatabasePersistenceManager.java:339)

I've read that you are supposed to use JCA to enable jackrabbit to participate in JTA transactions.
However, since our product must
work on a bunch of different application servers I'm a little hesitant to introduce this new requirement.

Has anybody used JTA and jackrabbit without using JCA?

Magnus Grimsell
magnus.grimsell <at> idainfront.se