Mailing-List: contact users-help@jackrabbit.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@jackrabbit.apache.org
Received-SPF: pass (athena.apache.org: domain of mittspamkonto@gmail.com
 designates 72.14.204.228 as permitted sender)
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition;
        b=jiRXgWIqwRjEL4gaKNtQ5vVVrn/J7b6GNZj+z+ZLP5J7l2/xDPP6xJk8d2FEMnrKWA5O9VAkeTPVm0QfGqjwBsmg2whIb+KLKYff9L/5n7ZID1YIkKk8yRz1JHPqq23sz+frRhnHcFIjgo01XvRcjeY8dPsNj2nah0fQNmJ/5Uk=
Message-ID: <2c7d5a310802260322x3f70f9cn10db6c043f1cb5c6@mail.gmail.com>
Date: Tue, 26 Feb 2008 12:22:49 +0100
From: "Foo Bar" <mittspamkonto@gmail.com>
To: users@jackrabbit.apache.org
Subject: Concurrent access problems
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

Hi.

I have just made my first attempt to use Jackrabbit in a bigger
project, and I'm having problems with concurrent access.

I'm using jackrabbit 1.4 with the example repository.xml in JBoss 4.2.2.GA.

The first test I made is simple. I tried to change an implementation
that currently uses a filesystem to store files to use jackrabbit.
Users upload files, these get unique id:s and I save them in a flat
directory:

root
|
files
|-1
|-2
|-3
etc..

Each file-node is created according to the recipe at
http://wiki.apache.org/jackrabbit/ExamplesPage (jcr:content subnode
etc). The save is as simple as possible; I get the session, create the
file subnode with binary content and logout.
The id:s are fetched from a db (auto-increment), so they are unique
across transactions.

The saving works fine, until I start hitting the page with a few
(tenths of) simultaneous uploads. I then get validation exceptions,
telling me about conflicting changes:

 javax.jcr.InvalidItemStateException:
6570ddc8-29a7-4335-a9ca-76dc98687d22: the item cannot be saved because
it has been modified externally.
        at org.apache.jackrabbit.core.ItemImpl.getTransientStates(ItemImpl.java:378)
        at org.apache.jackrabbit.core.ItemImpl.save(ItemImpl.java:1083)
        at org.apache.jackrabbit.core.SessionImpl.save(SessionImpl.java:896)
        at org.apache.jackrabbit.jca.JCASessionHandle.save(JCASessionHandle.java:178)

I've read the discussion here:
http://www.nabble.com/Problem-in-Multithreaded-Environment-td14779040.html
(nabble is down, google cache here:
http://66.102.9.104/search?q=cache:jqbwF-fsAjoJ:www.nabble.com/Problem-in-Multithreaded-Environment-td14779040.html)

This discussion, together with the issue at
http://issues.apache.org/jira/browse/JCR-584 basically says that it
should be ok to have several sessions concurrently adding subnodes to
a node, without any problems. The JIRA issue only explicitly says that
two sessions concurrently can add subnodes, and I assumed this to be
valid for any number of sessions. Is this a correct assumption?

Is there anything I should keep in mind when trying to make this work?
Any properties who need to be set on nodes or in workspace? Any save
order that might cause problems?

I'm having trouble reproducing the problem in unit tests, so any hints
as to why this is happening would be greatly appreciated.

I don't really see serializing access as an option, since the plan is
to have many concurrent saves. I therefore ruled out locks. Is it true
that locks always imply serial access?


Regards

Alexander