[AG-TECH] Bridge Registry Problem?

[Michael Braitmaier]

In fact I had problems with our AG3.1 bridge server being able to get 
its RegistryClient connected to the peers repository. However this 
seemed to be a temporary probelm has gone away now, which makes it even 
harder to track down. I just realised that at the same time while I had 
the problem, after doing a bridge cache update, around 80% of the 
bridges were unreachable, which seemed for me not to be a coincidence.

Cheers,
Michael



Adam Carter schrieb:
> Hi Chris (cc list),
>
> Thanks for your reply. I can see now the difference between the web 
> service that returns the bridge list and the list of bridges itself, 
> so the fact that the list only has one entry no longer worries me.
>
> However, is it possible that bridges are trying to add their entry to 
> the Argonne server, failing for some reason (possibly due to the way 
> that the server at Argonne has been configured?), and then not 
> launching properly themselves because of this? This would explain the 
> behaviour that I've had reported to me that bridges will not go back 
> up when they've gone down... If others who are using a bridge have not 
> noticed the problem, then I guess it's some other problem with the 
> bridge, but I thought I'd check first. I see that Michael M has 
> restarted his bridge, apparently with no problems, so I assume that 
> it's something at the other end...
>
> Cheers,
>
> Adam
>
>
>
> Christoph Willing wrote:
>>
>> On 16/06/2009, at 11:07 PM, Adam Carter wrote:
>>
>>>
>>> Hi,
>>>
>>> Recently I've noticed that a couple of unicast bridges I've tried to 
>>> contact have been unavailable. When I spoke to Michael Braitmaier in 
>>> Stuttgart he said that his bridge server was not restarting because 
>>> it couldn't access the peers.txt list on the Argonne Server. 
>>> Checking in the Venue Client in my list of bridge servers, nearly 
>>> all of them show up as unreachable. I can still connect to those 
>>> which are still up and running at their original addresses, but I 
>>> suspect this is also related to the peers file. When I put the peers 
>>> file url into a browser it shows only one entry, namely that of 
>>> Argonne. Is this deliberate, or an oversight?
>>
>>
>> Adam,
>>
>> The peers.txt will normally contain just a single entry consisting of 
>> the name of a machine and the port at which a particular web service 
>> is running. Bridges register with that web service when they start up 
>> and, when queried the right way, that web service returns the list of 
>> bridges that have registered.
>>
>> In short, the minimal content of peers.txt is not the reason why some 
>> bridges end up being unreachable.
>>
>>
>> chris
>>
>>
>> Christoph Willing                       +61 7 3365 8316
>> QCIF Access Grid Manager
>> University of Queensland
>>
>>
>
>