[maker-devel] Database disk image is malformed error

Tim Fallon tfallon at mit.edu
Fri Jun 16 09:07:14 MDT 2017


Hi there,

I’ve been running MAKER in a 2 stage way using MPI, to annotate a de novo insect genome.  By two stage, I mean for stage 1 I have a lot of independent folders / maker runs (e.g. individuals reference insect proteomes passed as FASTA with protein2genome=1), and then for stage 2 in a separate folder I am concatenating all that evidence from Stage 1 (using gff3_merge -o) and passing it as GFF parameters.

Stage 2 has been crashing.  It takes a very long time to setup the SQLite DB from the (~24 hours, with 39 MPI CPUs), and then once it is all loaded it works for a couple seconds then crashes with things like this:

"DBD::SQLite::db selectcol_arrayref failed: database disk image is malformed at /lab/solexa_weng/testtube/maker_3.00_beta/bin/../lib/GFFDB.pm line 525.”

I am passing a lot of evidence to Stage 2, probably more than people typically pass (the GFFs together are 44GB, whereas the resulting *.db file is 95G).

Have you seen this error before?  I’m thinking it could be a couple possibilities:
1) Running up against SQLite size / concurrency constraints where the .db ends up being malformed due to MPI / passing too much evidence. Solution -> Load GFFs without MPI, or load less evidence.
2) GFFs are malformed (they pass validation with GT). Solution -> Remove the malformed GFF evidence, although I haven’t been able to track any malformed GFFs down.
3) Identifiers in the GFF that are unique when in a single file, become non-unique. Solution -> Manually rename IDs in passed GFF files to be unique.

Thoughts?

All the best,
-Tim

Timothy R. Fallon
PhD candidate
Laboratory of Jing-Ke Weng
Department of Biology
MIT

tfallon at mit.edu

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://yandell-lab.org/pipermail/maker-devel_yandell-lab.org/attachments/20170616/a6bc9f53/attachment-0002.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1849 bytes
Desc: not available
URL: <http://yandell-lab.org/pipermail/maker-devel_yandell-lab.org/attachments/20170616/a6bc9f53/attachment-0002.p7s>


More information about the maker-devel mailing list