Status of concurrent versioning (Was: Re: concurrent writes (JCR-314))

[Jukka Zitting <ju...@gmail.com>]

Hi,

[Taking the thread to dev@]

On 6/25/07, Marcel Reutegger <ma...@gmx.net> wrote:
> Pablo Rios wrote:
> > Regarding the steps below required to support concurrent writes, what
> > considerations should be taken into account in connection with JCR-18 issue ?
>
> IMO JCR-18 should be closed as fixed and new issues should be created as new
> deadlocks are observed. e.g. JCR-962.
>
> as jackrabbit evolved, new synchronization and locking was added to the core. I
> think their usage should be reviewed and changed accordingly. most important,
> the sequence when locks are acquired.

What's the general feeling about concurrent versioning? Do we have
light at the end of the tunnel? I guess reviewing whether JCR-18 is
still relevant would make sense

At the moment my standard recommendation for versioning operations in
concurrent environments is to use guard code like this (AFAIK
concurrent normal workspace access doesn't interfere with versioning):

    synchronized (repository) {
        // versioning operations
    }

It would be very nice if we could drop those synchronization block.

BR,

Jukka Zitting

Re: Status of concurrent versioning (Was: Re: concurrent writes (JCR-314))

[Marcel Reutegger <ma...@gmx.net>]

Jukka Zitting wrote:
> Excellent! It would be nice to have JCR-18 and related issues as fixed
> in Jackrabbit 1.4. The beer is on me if you make that happen. :-)

there's only one thing I can say:

http://www.streettech.com/storypics/simpsonbeer.wav

regards
  marcel

Re: Status of concurrent versioning (Was: Re: concurrent writes (JCR-314))

[Jukka Zitting <ju...@gmail.com>]

Hi,

On 6/25/07, Marcel Reutegger <ma...@gmx.net> wrote:
> Jukka Zitting wrote:
> > What's the general feeling about concurrent versioning? Do we have
> > light at the end of the tunnel? I guess reviewing whether JCR-18 is
> > still relevant would make sense
>
> I'm currently working on issue JCR-962 and I think I can see the light ;) though
> I'm still stumbling over loose rocks now and then...

Excellent! It would be nice to have JCR-18 and related issues as fixed
in Jackrabbit 1.4. The beer is on me if you make that happen. :-)

BR,

Jukka Zitting

Re: Status of concurrent versioning (Was: Re: concurrent writes (JCR-314))

[Marcel Reutegger <ma...@gmx.net>]

Jukka Zitting wrote:
> What's the general feeling about concurrent versioning? Do we have
> light at the end of the tunnel? I guess reviewing whether JCR-18 is
> still relevant would make sense

I'm currently working on issue JCR-962 and I think I can see the light ;) though 
I'm still stumbling over loose rocks now and then...

I'll post a patch tomorrow for review. the general idea is to write lock the 
version manager early in the transaction processing to avoid deadlocks and to 
maintain the same sequence of locking within XA transactions as well as regular 
versioning operations.

I will also create some test cases that cover JCR-18 so we can close the issue 
if the tests pass.

regards
  marcel