Hello Alessandro,

A first, very useful step to diagnose your mpiblast freeze would be to 
run mpiblast with the --debug command-line option.  That will cause a 
verbose debugging log to be printed.  If the debug output is smaller 
than a few KB, please send it to the list.  Otherwise, please send it 
directly to me...

Thanks,
-Aaron

alessandro albiero wrote:
> Dear Developers,
>
> We work for the university of Padua and we are interested in your 
> programm.
> We are developping a web-services platform for the gene annotation and 
> we would like to use mpiblast for making our system faster. In fact we 
> have to use blast but this limit the platform in term of time.
> We hope to resolve this problem using a parallelized blast like mpiblast.
>
> I try to test mpiblast but I have some problems and I ask help to you.
>
> 1) I try to use the mpiblast version 1.4.0 with ncbi tools of oct 
> 2004. In my pc (with debian etch) I can't compile the ncbi tools. I 
> check in the web and I have seen that this problem is know.... so I 
> try to compile ncbi tools and mpiblast in a debian Sarge. The 
> compiling problem are resolved but mpiblast don't work correctly.
> MpiFormatBD seems to work correctly and it produces a fragmented 
> formatted database but mpiblast doesn't produce output.
> I try to run the programm with "mpirun" in a cluster of 8 machines 
> (all with etch...). Mpiblast is correctly launched in every  machine 
> but after 30 minutes I have to kill the processes because they seem 
> freezed. I don't receive any error messages. Using a normal blastn 
> with the same query and database, it completes the operation in 4 
> seconds.
> The configuration file .ncbirc is correct and it includes shared 
> directories all accessible.
> I try to launch mpiblast with STRACE for checking where it freezes: in 
> every case I tried, it freezes with this output
> "write(-1, "cmd=client_bnr_fence_in gid=0 gr"..., 48) = -1 EBADF (Bad 
> file descriptor)"
>
> 2) I try to use the rpm version. I convert all with alien but also in 
> this case the programm doesn't work correctly. When I try to run it, I 
> receive always the same error :
> "Sorry, mpiBLAST must be run on 3 or more nodes
> aborting job:
> application called MPI_Abort(MPI_COMM_WORLD, 0) - process 0"
> even if I try to run it on 8 nodes!!!
>
> We try to use different version of ncbi tools (more recent, if the 
> patch was available...) and different version of mpiblast but we 
> receive always the same messages from the system.
> May this be a problem of debian version ? ( we compile on sarge but 
> the we test the programm on etch.... but we can't downgrade all the 
> machine....).
>
> I hope to receive help.
>
> Best Regards
>
> Alessandro Albiero
> c/o Bioinformatic Lab. of CRIBI, Prof. G. Valle,
> University of Padova,
> V.le G. Colombo 3, 35121 Padova (Pd) Italy,
> tel +39 0498276165,
>
> -- 
> Alessandro
> ------------------------------------------------------------------------
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> ------------------------------------------------------------------------
>
> _______________________________________________
> Mpiblast-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/mpiblast-users
>   


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Mpiblast-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mpiblast-users

Reply via email to