[Urgent] BUG in Take operation

[Charith Wickramarachchi]

Folks,

I have found a major bug in the take operation where when we are doing replication some contents does not get removed.This needs to be attend ASAP.

I looked in to the issue and the bug was caused by two reasons

1. Wildcard content removal does not happen properly
2. After replicating the tuple removal does not happen properly 
   

i fixed the 1st issue so that bug is no longer visible to the user. (System is behaving correctly in the user point of view). But still tuple removal is not consistent.Which means even though a tuple may not visible to users it may exist invisibly in the system.

For me its looks like due to the bug in Freepastry gethandles() function since for small netwroks it will not return expected number of handles. This was documented in the code by jeff in 2.1 code.

But i m not 100% convinced that reson behind this is that PAST issue. So im glad if some one can have a look and confirm???


 

 


thanks ,
/Charith

--
Charith Dhanushka Wickramarachchi
http://charithwiki.blogspot.com/

[Pradeep Fernando]

On Thu, Jun 10, 2010 at 5:26 PM, Charith Wickramarachchi
<charith....@gmail.com> wrote:
> Folks,
>
> I have found a major bug in the take operation where when we are doing
> replication some contents does not get removed.This needs to be attend ASAP.
>

so, you mean we were doing 0 replication all this time. didnt look in
to your initialization code.

> I looked in to the issue and the bug was caused by two reasons
>
> Wildcard content removal does not happen properly
> After replicating the tuple removal does not happen properly
>
> i fixed the 1st issue so that bug is no longer visible to the user. (System
> is behaving correctly in the user point of view). But still tuple removal is
> not consistent.Which means even though a tuple may not visible to users it
> may exist invisibly in the system.
>
> For me its looks like due to the bug in Freepastry gethandles() function
> since for small netwroks it will not return expected number of handles. This
> was documented in the code by jeff in 2.1 code.
>
> But i m not 100% convinced that reson behind this is that PAST issue. So im
> glad if some one can have a look and confirm???

i'll look in to this.can you give me a data sequence so that i can
debug & get up to the speed.

>
>
>
>
>
>
>
> thanks ,
> /Charith
>
> --
> Charith Dhanushka Wickramarachchi
> http://charithwiki.blogspot.com/
>
>

--
Pradeep Fernando.
http://pradeepfernando.blogspot.com/

[Charith Wickramarachchi]

On Thu, Jun 10, 2010 at 5:43 PM, Pradeep Fernando <prad...@gmail.com> wrote:

On Thu, Jun 10, 2010 at 5:26 PM, Charith Wickramarachchi
<charith....@gmail.com> wrote:
> Folks,
>
> I have found a major bug in the take operation where when we are doing
> replication some contents does not get removed.This needs to be attend ASAP.
>

so, you mean we were doing 0 replication all this time. didnt look in
to your initialization code.

   It was 0 replication for the default case. but now i set it to 1  while testing with some other apps. So now default is 1

look at BissaFactory -> createEnvironment method which use only ip port info to create environment.

> I looked in to the issue and the bug was caused by two reasons
>
> Wildcard content removal does not happen properly
> After replicating the tuple removal does not happen properly
>
> i fixed the 1st issue so that bug is no longer visible to the user. (System
> is behaving correctly in the user point of view). But still tuple removal is
> not consistent.Which means even though a tuple may not visible to users it
> may exist invisibly in the system.
>
> For me its looks like due to the bug in Freepastry gethandles() function
> since for small netwroks it will not return expected number of handles. This
> was documented in the code by jeff in 2.1 code.
>
> But i m not 100% convinced that reson behind this is that PAST issue. So im
> glad if some one can have a look and confirm???

i'll look in to this.can you give me a data sequence so that i can
debug & get up to the speed.

    Since i fixed the 1st issue that issue will be no longer there.
    for testing purposes what u can do is comment  the code that remove wildcards .(In BissaImpl class take operation)
   
     and start up two nodes.

    then insert 2-4 tuples that matches a same template

    ex : hello,world   hello,world1 hello,world11 hello,world111 hello,world111

    and take them using the template hello,??? from the same node
    and try read it from the same node and the other node using hello,???

    even though u have comment the wild card removal part this  system must not read tuples after the take since original ones should be  gone.

have a try and send the updates to the list.






 

>
>
>
>
>
>
>
> thanks ,
> /Charith
>
> --
> Charith Dhanushka Wickramarachchi
> http://charithwiki.blogspot.com/
>
>

--
Pradeep Fernando.
http://pradeepfernando.blogspot.com/