Excepted size prefilter and --disk-space-limit

[michoug]

Hi,
I'm trying to run an annotation with a metagenome filtered with Eukrep of ~1115 MBp with your database based on Uclust.

easy-predict --threads 35 --slice-search -e 100 --metaeuk-eval 0.0001 --min-ungapped-score 35 --min-exon-aa 20 --metaeuk-tcov 0.6 --min-length 40 --disk-space-limit 200G --compressed 1 ASSEMBLY_euk_sim.fasta /mnt/databases/MMETSP/MMETSP_uniclust50_MERC ASSEMBLY_euk_metaeuk temp

I have around 550 Gb free and your software tells me that it may need 30048T

Hard disk might not have enough free space (533G left). The prefilter result might need up to 30048T.

It seems a bit much, knowing that changing --disk-space-limit from 200Gb to 100Gb decrease the size by 2 ;) Should I stop the run now, Also, I was wondering in how many slices will the run be divided into as it may give me a better idea of the expected size of the run?

Another question that I had is the usage of the option --disk-space-limit, does increasing it, speed up the process/decrease the memory or other?

[elileka]

The number is indeed a bit ridiculous. However, it should not affect the run - it is just some printout that is not very useful in some cases (it is the product of the number of queries and --max-seqs). Generally, the average user shouldn't change --disk-space-limit because MetaEuk/MMseqs2 adjusts it on its own.

[ys117vt]

Hi,
by using database as .../MMETSP/MMETSP_uniclust50_MERC, did you have any issue with the swapresults? I got an error message as below:
/work/cascades/.../Cu_Rig/metaeuk/ASSEMBLY/temp_MERC/3071197285104908048/tmp_predict/14365638628511940689/tmp_search/16001091333167059095/searchslicedtargetprofile.sh: line 170: 17396 Killed "$MMSEQS" swapresults "${TARGET}" "${INPUT}" "${TMP_PATH}/aln" "${TMP_PATH}/aln_swap" ${SWAP_PAR}
Error: swapresults died
Error: search step died
Error: predictexons step died

[milot-mirdita]

That error message usually means that the linux out-of-memory killer process killed MetaEuk. What hardware are you using for this run? The MMETSP_uniclust50_MERC database is quite large and needs quite a bit of resources. See #24