[infinispan-dev] Command cancellation

Dan Berindei dan.berindei at gmail.com
Mon Sep 24 07:38:39 EDT 2012


I'm not sure using Thread.interrupt() will work that well for cancelling
arbitrary distributed tasks without the Cache methods also throwing
InterruptedException. If we catch InterruptedExceptions and wrap them in
CacheExceptions, the distributed task won't know that it has been cancelled.

It may retry whatever cache operation it was doing (e.g. if using
optimistic txs), and it will succeed, because the interrupted flag on the
thread has been cleared when the InterruptedException was thrown. It would
be possible to extract the cause of caught CacheExceptions and check the
inner exception type, but that would be easy to overlook.

Cheers
Dan


On Mon, Sep 24, 2012 at 1:46 PM, Manik Surtani <manik at jboss.org> wrote:

>
> On 24 Sep 2012, at 11:39, Vladimir Blagojevic <vblagoje at redhat.com> wrote:
>
>  On 12-09-21 2:34 PM, Manik Surtani wrote:
>
> Looks good, except that the pseudocode for dealing with a
> CancellableCommand (on the recipient node) should look like:
>
>  * Receive command
> * if CancellableCommand, register with CancellationService
> * Perform command
> * If CancellableCommand, un-register from CancellationService
>
>  That last step was missing from your detail below.  I presume that would
> require a CancellationService#unregisterThread(UUID u) ?
>
>  I was looking at the most appropriate place for this logic and I think
> InboundInvocationHandlerImpl#handleInternal method and its
> try/catch/finally clause fits the bill, would you agree?
>
>
> I believe so, yes.
>
> --
> Manik Surtani
> manik at jboss.org
> twitter.com/maniksurtani
>
> Platform Architect, JBoss Data Grid
> http://red.ht/data-grid
>
>
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/infinispan-dev/attachments/20120924/97891457/attachment.html 


More information about the infinispan-dev mailing list