Re: [javagroups-users] Merge problem

[Victor <dre...@ma...>]

Bela,

sorry, I have one more question:
c) after being shunned, the channel is automatically closed - is this 
"be design"?
In such situation AUTO_RECONNECT option does not work (can not reconnect 
after being closed), so I need to write my own reconnection logic - 
creating a new JChannel instance and calling connect().

Victor N


Victor wrote:
> Bela,
> 
> things seems to be much better than I thought :)
> 
> let me continue:
> 
> 5) I waited until the view on node A updated (node A has a greater FD 
> timeout).
> Now node A was shunned.
> Strange thing here - I did not see "view changed" event on node A at 
> this moment; it was just shunned and I saw the following:
> Feb 26, 2009 7:30:37 PM org.jgroups.protocols.pbcast.GMS 
> castViewChangeWithDest
> WARNING: 192.168.1.1:30001 failed to collect all ACKs (1) for view 
> [192.168.1.1:30001|2] [192.168.1.2:30001] after 2000ms, missing ACKs 
> from [192.168.1.2:30001] (received=[192.168.1.1:30001]), 
> local_addr=192.168.1.1:30001
> 
> 6) after a delay a new Channel was connected on node A (my code works 
> this way),
> first, a singleton cluster was created on node A (node A saw itself 
> only; node B saw itself only)
> 
> 7) after about 1-1.5 minutes, finally, view was merged - nodes A and B 
> see each other now and send messages to each other!
> 
> So seems everything works, but some questions still interest me:
> a) why node A did not receive "view changed" event on the step 5?
> b) why node A did not see node B instantly on step 6?
> 
> Victor N
> 
> 
> Victor wrote:
>> Hello Bela,
>>
>> I did a test to check how the system reacts on network disconnection.
>> I have a typical TCP config with 2 nodes:
>>
>> 1) I started nodes A (192.168.1.1) and B (192.168.1.2) on 2 computers; 
>> but node A has a greater FD timeout than B
>>
>> 2) I switched off my network cable, waited a little, then switched the 
>> network on again
>>
>> 3) view on node B was updated (A is removed from it), view on node A 
>> is not updated - because FD timeout is greater, it's ok
>>
>> 4) after that the nodes can not merge:
>> node A says:
>> Feb 26, 2009 6:01:17 PM 
>> org.jgroups.protocols.pbcast.CoordGmsImpl$MergeTask run
>> WARNING: Merge aborted. Merge leader did not get MergeData from all 
>> subgroup coordinators [192.168.1.2:30001, 192.168.1.1:30001]
>> Feb 26, 2009 6:01:17 PM org.jgroups.protocols.pbcast.CoordGmsImpl 
>> handleMergeCancelled
>> WARNING: merge was supposed to be cancelled at merge participant 
>> 192.168.1.1:30001 (merge_id=[192.168.1.1:30001|1235667667861]), but it 
>> is not since merge ids do not match
>> received msg from 192.168.1.1:30001: hello world
>>
>> and node B says:
>> 26/02/2009 20:01:29| WARN  | 
>> org.jgroups.protocols.pbcast.NAKACK.handleMessage(): 
>> 192.168.1.2:30001] discarded message from non-member 
>> 192.168.1.1:30001, my view is [192.168.1.2:30001|2] [192.168.1.2:30001]
>>
>> What can I do here to help my system to merge?
>> Why don't the nodes merge or shun? I have shun=true in FD and GMS, 
>> auto-reconnection of Channel is disabled.
>>
>> Thanks,
>> Victor N
>>
>> ------------------------------------------------------------------------------ 
>>
>> Open Source Business Conference (OSBC), March 24-25, 2009, San 
>> Francisco, CA
>> -OSBC tackles the biggest issue in open source: Open Sourcing the 
>> Enterprise
>> -Strategies to boost innovation and cut costs with open source 
>> participation
>> -Receive a $600 discount off the registration fee with the source 
>> code: SFAD
>> http://p.sf.net/sfu/XcvMzF8H
>> _______________________________________________
>> javagroups-users mailing list
>> jav...@li...
>> https://lists.sourceforge.net/lists/listinfo/javagroups-users
>>
>>
>>
>>
>