traviswheeler
commented
3 years ago Looking back through the history of this issue (and the related notes near the end of #168), I believe the error reported by @RivasLab took one of these forms, depending on input: buildAndWriteFMIndex: Error writing T in FM index. buildAndWriteFMIndex: Error writing BWT in FM index. buildAndWriteFMIndex: Error writing SA in FM index.
These are all the sort of error you'd get if you've run out of disk on the temporary directory. Near the end of the conversation in #168, we made a guess that the problem was related to the location of the temporary directory: if you're on a cluster with massive network drive but small local disk, and /tmp is mounted on the local disk, you'd likely get an out-of-disk error for such a large sequence input. @npcarter suggested setting $TMPDIR to point to a large filesystems (/n/eddyfs03 or /n/holylfs03). Did this overcome the problem?
npcarter
this is a follow up on issue https://github.com/EddyRivasLab/hmmer/issues/168
We are using hmmer 3.3.2, which is a version after the previous makehmmerdb issue was closed.
We are trying to use makehmmerdb for a large dataset of vertebrate genomes (~600GB).
This is the error we get after ~24h running /var/slurmd/spool/slurmd/job22844537/slurm_script: line 14: /n/holylfs03/LABS/eddy_lab/Users/maoaoyue/hmmer-3.3.2/src/makehmmerdb: Input/output error
This is the script we used in the Odyssey cluster
!/bin/bash
SBATCH -n 1 # Number of cores requested
SBATCH -N 1 # Ensure that all cores are on one machine
SBATCH -t 3-00:00 # Runtime in minutes
SBATCH -p eddy # Partition to submit to
SBATCH --mem=16000 # Memory per cpu in MB (see also --mem-per-cpu)
SBATCH --open-mode=append
SBATCH -o %j.out # Standard out goes to this file
SBATCH -e %j.err # Standard err goes to this filehostname
DB=/n/holylfs03/LABS/eddy_lab/home/wgao/vertebrates/vertebrate_genomes.fasta BF=/n/holylfs03/LABS/eddy_lab/Users/maoaoyue/vg.hmmerdb /n/holylfs03/LABS/eddy_lab/Users/maoaoyue/hmmer-3.3.2/src/makehmmerdb $DB $BF