For most of our searches SequenceServer (1.0.14) works great but there are occasional queries we run which take a very long time and eventually the browser crashes.
The database is only 1.1g and query is about 500nt, but when I watch the network tab of devtools over 1.1gb of data is returned from the server before my browser halts. On the server site I can watch top and the sequence server uses up to 7GB of RAM while it runs this query before finally returning results. It seems like an enormous amount of hits are being returned but when I do the same query on the command line with blastn it only 55 results over default cutoffs are returned.
Any clue what could be happening here?
Thank you for highlighting this and providing the details. Is that 55 hits or 55 hit sequences each with potentially hundreds of HSPs (local pairwise alignments)?
FWIW, the way we parse and return results is completely rewritten in version 2, in part because of performance/scaling issues.
Our version 2 “prereleases” are actually very stable (we use them for research, but have conservatively stopped short of calling them completely stable because we’re still ironing out a few details).
It would be super helpful if you could try running your analysis in v2 and let us know if you get the same pattern.
If I may to add Yannick’s message:
To get a sense of equivalent results on the command line, you would have to run blastn with -task blastn option. I suspect that would return a lot more than 55 hits.
I am indeed curious to know if version 2.0 of sequenceserver handles your situation more gracefully and if things are a bit better with Firefox than with Chrome (or vice versa).
To get seqserv 2.0:
gem install --pre sequenceserver
You can rollback if you don’t like it:
gem uninstall sequenceserver