--- Aaron Darling <darling at cs.wisc.edu> wrote: >If large exact matches aren't problematic for standard NCBI blast onyour machine then they shouldn't be troublesome for mpiblast either. One thing you could try is disabling the distributed bioseq lookup. Itmay be that one of the database sequences is particularly large and thewriter crashes when trying to receive a very large message using MPI. Use the command line switch --disable-mpi-db to shut off distributedbioseq lookups. Beware that --disable-mpi-db should only be used with blastn searches... Fortunately, I'm only doing blastn searches. Even more fortunately, --disable-mpi-db seems to have solved the problem. A full run takes about 2 days, and I've only just been able to take a stab at it. But its currently cruising through sequences that used to make it crash. A great big thanks to Alan Kilian and Chris Dwan for offering more insight to the problem. > For what it's worth (maybe not much), this query didn't > ... Its worth a heck of a lot. Certainly a few more drinks, if you can wait about 6 months. And just for the archives, some of you would have seen my earlier messages about a message 'Error allocating X bytes for Dynamic Programming'. This issue was only happening on machines using the 2.6 kernel. I hope to continue to play with configurations and isoate that problem a bit better. But if anyone else sees something comparable, you may want to roll back to 2.4. Mike "I've been chasing this bug off-an-on for 4 months now. In 60 days we have to deliver to the customer." Cariaso ===== Mike Cariaso