eledef failed. Exit code 256 - with a combo of RepeatModeler version: 2.0.8 RECON version: 1.08

[Stephen-Richards-scripts]

**I had a crash with the error eledef failed. Exit code 256 **

A concise description of the bug, including any error messages.
eledef failed. Exit code 256

Reproduction steps
Can also do this:
eledef seqnames msps.out single
which gives:
Can not open the fragment list file, exiting.

However if I run it the old way:
mkdir -p images summary
imagespread seqnames msps.out
sort -T . -k 3,3 -k 4n,4n -k 5nr,5nr images/spread1 > images/images_sorted
mkdir -p ele_def_res
eledef seqnames msps.out single

After looking at the code it seems there is a runRECON and a runRECONOLD, but the last is not called.
I think it just needs to check the Recon version so if someone has the old one it works OK. RECON seems to be in the process of being updated and I don't think the new version has hit bioconda yet

I changed the line: (in the RepeatModeler perl script)
# Run recon
unless ( -f "$roundTmpDir/summary/eles" ) {
runRECON( workDir => $roundTmpDir,
reconPath => $RECON_DIR );
}

to :
# Run recon
unless ( -f "$roundTmpDir/summary/eles" ) {
runRECONOLD( workDir => $roundTmpDir,
reconPath => $RECON_DIR );
}
For me - so I am fine with the hack. -

Log output
sorry - I forgot and deleted it :(
Here is a snippet I kept -- it is at the end of round 1:
99% completed, 00:0:00 (hh:mm:ss) est. time remaining.
99% completed, 00:0:00 (hh:mm:ss) est. time remaining.
100% completed, 00:0:00 (hh:mm:ss) est. time remaining.
Comparison Time: 01:06:56 (hh:mm:ss) Elapsed Time, 3597 HSPs Collected

• RECON: Running initial definition of elements ( eledef )..
  eledef failed. Exit code 256

Environment (please include as much of the following information as you can find out):
Conda env on hpc at UC Davis

• How did you install RepeatModeler? e.g. manual installation from repeatmasker.org, bioconda, the Dfam TE Tools container, or as part of another bioinformatics tool?

• I installed using the Bioconda earlGrey env

• Which version of RepeatModeler do you have? The output of RepeatModeler without any options will be a help page with the version of the program displayed at the top.

• RepeatModeler version: 2.0.8 RECON version: 1.08

• Which version of RepeatMasker is this RepeatModeler installation using? Have you installed RepBase RepeatMasker Edition for RepeatMasker, or the full Dfam database?

• Operating system and version. The output of uname -a and lsb_release -a can be used to find this.

Additional context
I never submitted a big report before - so please tell me if this is useful, or if you need more info , or if this is just annoying and I shouldn't.

[TobyBaril]

Hi @Stephen-Richards-scripts - I've been working on a fix for this and think I have wrangled the recipe for RECON 1.10. If it passes testing, it should be available in Bioconda soon. In the meantime, I would suggest activating your earl grey environment, and then installing RepeatModeler 2.0.7 which will work with the current version of RECON. I have updated the latest Earl Grey recipe to restrict installation to 2.0.7 until the new RECON is available.

[Stephen-Richards-scripts]

Will do,
Thanks for getting to this so soon!
Stephen

[Stephen-Richards-scripts]

Hi Tobias,

Thank you again for the quick response on the recent RepeatModeler/RECON issue.
I wanted to check something else with you before assuming it was a bug.
While investigating Earl Grey performance on a large spider genome, I noticed that Earl Grey appears to divide the requested thread count by 4 before passing it to TEstrainer:
strainthreads=$(expr $ProcNum / 4)

Then, inside TEstrainer_for_earlGrey.sh, MAFFT parallelism is calculated again as:
MAFFT_THREADS=$(($THREADS / 4))

and each MAFFT process is run with:
mafft --thread 4

So, for my Earl Grey run I requested with 30 threads ends up running the primary MAFFT stage with parallel --jobs 1, i.e. one 4-thread MAFFT job at a time. After changing Earl Grey to pass the full $ProcNum into TEstrainer, the same stage appears able to run with --jobs 7, which seems much closer to using the requested allocation.

Was the initial division by 4 in Earl Grey intentional, perhaps as a memory-safety precaution, or might this be an accidental double division now that TEstrainer also handles MAFFT thread partitioning internally?

Best,
Stephen

[helenabea]

Hi, just wanted to add that I had the same problem and redownloaded repeat modeler with v2.0.7, and now it's running normally. I did tried reinstalling RECON, repeat masker and repeat modeler multiple times, until I ran the RECON Demo and realized it should be something with the source code o v2.0.8

[rmhubley]

I don't believe there is a current problem configuring RepeatModeler 2.0.8 with either RECON 1.0.8 or RECON 1.10 correct? It seems like the last few comments pertain to either Bioconda or Earl Grey. Please let me know if there are any current issues related to installating/using RECON with RepeatModeler as I am about to push a new release and I would like to address any ongoing concerns.

[helenabea]

Hi Robert, I was not using Bioconda or EarlGrey. Installing RepeatModeler 2.0.7 worked with RECON 1.0.8. I did not try RepeatModeler 2.0.8 with RECON 1.10 though.

[rmhubley]

Thanks to everyone for your persistence with this issue. @Stephen-Richards-scripts - your response was spot on and fixes the issue for both versions of RECON. The issue stems from the new RECON APIs and where compatibility was worked in. The code in RepeatModeler was set to work with only the new version of RECON (1.10) and never reverted back to the older RECON API (1.08 and earlier). RECON 1.10 is able to be called using either API so the right thing to do was to revert the RepeatModeler change. In any case, I have fixed this in the release I am preparing (RepeatModeler 2.0.9) which should come out this week.