[infinispan-dev] IRC meeting log

Pedro Ruivo pruivo at gsd.inesc-id.pt
Wed Jul 4 14:14:00 EDT 2012


Hi Mircea, Dan

I've completed the rebase to Dan's branch and I've made the pull request 
too (see [1]).

The branch has already the total order code for Distributed mode and you 
can start look at it. However, if you don't wanna make it public 
already, you can disable in Configuration.

I've run the total order tests cases and, occasionally, some tests are 
failing randomly. I'm trying to reproduce the failures to get more 
insights on this, but when I run them standalone, the errors do not show 
up anymore... so I'm currently still trying to verify what goes on 
exactly, but this is not very easy. However, my theory is that, with 
respect to the existing 2PC-based protocol, when we use TO, one of the 
preconditions assumed by the test case is no longer valid, hence I 
suspect that the errors that we are getting are actually false positives.

The point is that with TO, unlike 2PC, when a transaction commits at the 
originator, this does not guarantee that all the nodes have already 
applied the update. With 2PC, instead, the commit command returns only 
after the commit messages have been acknowledged by all nodes. It is my 
understanding that some of these tests rely on this assumption. My 
proposal is therefore to update the totalorder-related tests to change 
the immediate consistency check with an eventually. I have already done 
this in the code for which I have issued a pull request.

Cheers,
Pedro Ruivo

[1] https://github.com/danberindei/infinispan/pull/2


On 6/12/12 10:00 AM, Pedro Ruivo wrote:
> Hi Mircea,
>
> Sorry for the delay in replying to emails but, to be honest, I am too
> busy right now because we have the project review next Tuesday :(
>
> If you don't mind, I will do it in the last week of June.
>
> Cheers,
> Pedro
>
> On 6/12/12 9:07 AM, Mircea Markus wrote:
>> Hi Pedro,
>>
>> Any luck with the rebase?
>>
>> Cheers,
>> Mircea
>>
>> ----- Original Message -----
>>> From: "Mircea Markus"<mircea.markus at jboss.com>
>>> To: "infinispan -Dev List"<infinispan-dev at lists.jboss.org>
>>> Sent: Wednesday, June 6, 2012 2:26:31 PM
>>> Subject: Re: [infinispan-dev] IRC meeting log
>>>
>>>
>>> Hi Pedro,
>>>
>>>
>>> Would it be possible for you add the logic on top of our TOM
>>> branch[1] and issue a pull reques? We won't integrate TOA in a first
>>> stage.
>>>
>>>
>>> [1]
>>> https://github.com/danberindei/infinispan/commits/mmarkus_t_to_dan_final
>>>
>>>
>>>
>>> On 4 Jun 2012, at 16:55, Pedro Ruivo wrote:
>>>
>>>
>>>
>>> Hi Dan, all
>>>
>>> I've finished the implementation State Transfer for the Total Order
>>> [Broadcast/Anycast] based protocol. You can find the code here [1].
>>>
>>> Note: It seems to work but I didn't test it extensively
>>> Note2: This code has TOA too (Total Order based for Distributed mode)
>>>
>>> Feedback is welcome :)
>>>
>>> Cheers
>>> Pedro Ruivo
>>>
>>> [1] https://github.com/pruivo/infinispan/tree/to_state_transfer
>>>
>>> On 6/4/12 4:04 PM, Dan Berindei wrote:
>>>
>>>
>>> Galder, Re:TOB, I believe Pedro has started implementing state
>>>
>>>
>>> transfer according to their design at
>>>
>>>
>>> https://docs.google.com/document/d/1QVALG4xnu2kxXFrj6M8RLXJG3OcKwH9k-678mdbT6I0/edit
>>>
>>>
>>>
>>>
>>>
>>> I have made some comments on the design, but I'm not sure where he is
>>>
>>>
>>> with the implementation.
>>>
>>>
>>>
>>>
>>>
>>> Cheers
>>>
>>>
>>> Dan
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> On Mon, Jun 4, 2012 at 5:43 PM, Galder Zamarreño<   galder at redhat.com>
>>> wrote:
>>>
>>>
>>>
>>>
>>> Meeting bot was not working (??), so find attached the log.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Cheers,
>>>
>>>
>>>
>>>
>>> --
>>>
>>>
>>>
>>>
>>> Galder Zamarreño
>>>
>>>
>>>
>>>
>>> Sr. Software Engineer
>>>
>>>
>>>
>>>
>>> Infinispan, JBoss Cache
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>>
>>>
>>>
>>>
>>> infinispan-dev mailing list
>>>
>>>
>>>
>>>
>>> infinispan-dev at lists.jboss.org
>>>
>>>
>>>
>>>
>>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>>>
>>>
>>> _______________________________________________
>>>
>>>
>>> infinispan-dev mailing list
>>>
>>>
>>> infinispan-dev at lists.jboss.org
>>>
>>>
>>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>>> _______________________________________________
>>> infinispan-dev mailing list
>>> infinispan-dev at lists.jboss.org
>>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>>>
>>>
>>> _______________________________________________
>>> infinispan-dev mailing list
>>> infinispan-dev at lists.jboss.org
>>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev


More information about the infinispan-dev mailing list