Arek Kasprzyk wrote:
> 
> On 1 Sep 2006, at 10:27, David Withers wrote:
> 
>> Hi Arek,
>>
>> I just want to check a couple of things as I investigate this.
>>
>> 1. Are you seeing the above sequence of requests happening 100,000 times?
> 
> no, not the sequence of requests x 100,000, all of them ~100,000.
> given that the sequece consists of 15 requests (most of which are
> configuration
> request, the sequence would happen ~6,000x15

The multiple 'datasets' and 'configuration' requests you are seeing was
due to cached results of those requests being lost when objects were
serialized - this problem has been fixed in the current code.

>> 2. Does each request happen twice or is the logger reporting it twice?
>>
> 
> the request happens twice as far as I can make it.

I can't see anything in the client causing each query to be made twice.
Maybe the workflow does this explicitly.

> I am only trying to signal the problem on the list. Perhaps it would be
> best if I got
> you in touch with the guys who are working on these workflows (offline)
> and they will give
> you more details themselves so you can verify  that such problem really
> exists and if so
> diagnose it further with them and come up with the plan for improvements
> if necessary
> 
> what do you think?
> a.

I don't think we could prevent people from creating queries that have,
for example, an ensembl_transcript_id filter with one id - it's a
perfectly valid query. We could probably restrict the number of requests
that can be made to a biomart service within a set timeframe. However,
it would be better to restrict access on the server side - if the server
starts sending errors back then Taverna, or another client using the
martservice interface, can retry later, using appropriate backoff and delay.

David.
-- 
David Withers
School of Computer Science, University of Manchester,
Oxford Road, Manchester, M13 9PL, UK.
Tel: +44(0)161 275 0145

Reply via email to