Eric C. Anderson
- Fresh install of mambaforge
- Get Snakemake
- Get mega-non-model-wgs snakeflow and test it in an interactive session
- Start copying data over
- copy the old resources directory that has the indexed genome
- Make an alpine slurm profile
- Running
- Restarting
- Reboot! The sample IDs were not the same between runs
This documents Eric modifying the workflow to run on alpine, and also getting things setup to run all the SWTH data.
To run this workflow on Alpine, I wanted to start from a completely clean slate. So I started by completely re-installing mambaforge. Since my tmux is in mambaforge, I had to first make sure all my tmux sessions were killed.
(base) [login10: ~]--% which tmux
/projects/eriq@colostate.edu/mambaforge/bin/tmux
(base) [login10: ~]--% tmux kill-server
error connecting to /tmp/tmux-2002325/default (No such file or directory)
# It looks like nothing was actually running
# Wait! That is because I typically login to login11...
# as per my .ssh/config:
Host summit
HostName login11.rc.colorado.edu
RemoteForward 52611 localhost:52698
# So, I got onto that instead.
ssh eriq@colostate.edu@summit
# but I had closed all my tmux windows already. Good
(base) [login11: ~]--% tmux kill-server
no server running on /tmp/tmux-2002325/defaultNow, it is time to reinstall mambaforge.
First, I removed the conda block from my .bashrc, logged out and logged back in again.
Then I removed my mambaforge directory:
[login11: projects]--% pwd
/home/eriq@colostate.edu/projects
[login11: projects]--% ls
afblue Cons-Gen-Comp-2022 DaddySalmon empty_dir java-programs mambaforge README.mdwn rmote-server
[login11: projects]--% rm -rf mambaforgeThat takes a little bit of time. While it was running, I was reading through the Alpine docs.
Once it was done, I went ahead and did a fresh install of mambaforge as detailed here: https://mamba.readthedocs.io/en/latest/installation.html
[login11: projects]--% acompile
acompile: submitting job... salloc --nodes=1 --partition=acompile --ntasks=1 --time=01:00:00 --qos=compile --job-name=acompile --bell --oversubscribe srun --pty /bin/bash
salloc: Granted job allocation 2291926
salloc: Nodes c3cpu-c11-u21-2 are ready for job
wget "https://github.com/conda-forge/miniforge/releases/latest/download/Mambaforge-$(uname)-$(uname -m).sh"
bash Mambaforge-$(uname)-$(uname -m).sh
# after agreeing to the licensing, I had to tell it to put the thing in:
/projects/eriq@colostate.edu/mambaforge
# while it was installing, I noticed that it is using python 3.10
Extracting python-3.10.12-hd12c33a_0_cpython.condaThat was painless and incredibly fast. After it was done, I sourced my .bashrc to initialize it:
[c3cpu-c11-u21-2: projects]--% source ~/.bashrc
(base) [c3cpu-c11-u21-2: projects]--%Once in my base environment, I installed tmux there. I simply installed the latest tmux on conda: 3.5
(base) [c3cpu-c11-u21-2: projects]--% mamba install -c conda-forge tmuxOnce that was done. I logged out and logged back in via my iTerm tmux.
I have recently installed and tested snakemake 7.30.1 on SEDNA and it is working great. That is what is still currently up on Anaconda so I will use it here, as well, for consistency. Following the instructions at: https://snakemake.readthedocs.io/en/stable/getting_started/installation.html I did:
(base) [login11: ~]--% acompile
acompile: submitting job... salloc --nodes=1 --partition=acompile --ntasks=1 --time=01:00:00 --qos=compile --job-name=acompile --bell --oversubscribe srun --pty /bin/bash
salloc: Granted job allocation 2291932
salloc: Nodes c3cpu-c11-u21-2 are ready for job
(base) [c3cpu-c11-u21-2: ~]--% mamba create -c conda-forge -c bioconda -n snakemake-7.30.1 snakemakeThat was painless.
I symlinked scratch in my home directory to alpine’s scratch and am
now there.
(base) [login11: ~]--% cd scratch/
(base) [login11: scratch]--% ls
README.mdwn
(base) [login11: scratch]--% gitup
Agent pid 31431
Enter passphrase for /home/eriq@colostate.edu/.ssh/id_ed25519:
Identity added: /home/eriq@colostate.edu/.ssh/id_ed25519 (SUMMIT)
(base) [login11: scratch]--%
(base) [login11: scratch]--% git clone git@github.com:eriqande/mega-non-model-wgs-snakeflow.gitThe gitup there is an alias I have to refresh my SSH credentials,
which seems necessary occasionally on Alpine. The alias looks like this
in my .bashrc:
# here is an alias for the command you need to do to
# make SUMMIT have access to your SSH keys for GitHub.
# It seems like I have to do this pretty much every new
# session, and sometime after a certain amount of time
# in the same session, so I just made it an alias. Also,
# it requires my SSH passphrase, so I have to do it
# interactively.
alias gitup='eval "$(ssh-agent -s)"; ssh-add ~/.ssh/id_ed25519'At any rate, now I will first get an interactive session with a single core to first install all of the conda environments for the workflow:
(base) [login11: scratch]--% cd mega-non-model-wgs-snakeflow/
(base) [login11: mega-non-model-wgs-snakeflow]--% module load slurm/alpine
(base) [login11: mega-non-model-wgs-snakeflow]--% sinteractive -t 00:20:00 -c 1
/usr/local/bin/sinteractive: waiting for job with ID 2292145 to start.I note that using sinteractive gets screen involved, and I am not a
big fan of that when I am already using tmux. So, let’s see if srun will
work for us:
(base) [login11: mega-non-model-wgs-snakeflow]--% srun -c 1 -t 00:20:00 --pty /bin/bashHere on the fourth of july, 2023 in the early morning, it did not take long to get those resources!
(base) [c3cpu-c9-u1-1: mega-non-model-wgs-snakeflow]--% conda activate snakemake-7.30.1
(snakemake-7.30.1) [c3cpu-c9-u1-1: mega-non-model-wgs-snakeflow]--% snakemake --cores 1 --use-conda --conda-create-envs-only --configfile .test/config/config.yamlWe will see if all of those can get installed in less than 20 minutes before my allocation dies!
That seemed to work. So, now I will get 4 cores on atesting and see how it works:
srun --partition=atesting -c 4 -t 00:30:00 --pty /bin/bash
# dry-run
snakemake --cores 4 -np --use-conda --configfile .test/config/config.yaml
# that looked fine, so I did a full funThat ran without any problems on the little test case. 322 steps done with no problems at all.
So, what we have left to do now is try it using the slurm profile on a real data set, like the SWTHs.
My token needed refreshing. I ended up just making a new rclone remote for my onedrive and then put a link to the RueggLab_BGPshare into my onedrive root directory.
I am going to copy using two different commands in different shells. First is:
(base) [login11: mega-non-model-wgs-snakeflow]--% fp .
/scratch/alpine/eriq@colostate.edu/mega-non-model-wgs-snakeflow/.
(base) [login11: mega-non-model-wgs-snakeflow]--% rclone copy onedrive2:BGP_Share/Genetic_and_Environmental_Data/Novoseq_fastq_2_download/SWTH data/SWTH
# Holy freakazoid that is fast! About 400 Mb/secThe second copy job is:
rclone copy onedrive2:BGP_Share/Genetic_and_Environmental_Data/Novoseq_fastq_2_download/SWTH_Novoseq_Plate2 data/SWTH_Novoseq_Plate2This is up on the sharepoint:
(base) [login11: mega-non-model-wgs-snakeflow]--% pwd
/home/eriq@colostate.edu/scratch/mega-non-model-wgs-snakeflow
(base) [login11: mega-non-model-wgs-snakeflow]--% rm -r resources
(base) [login11: mega-non-model-wgs-snakeflow]--% rclone copy onedrive2:BGP_Share/Genetic_and_Environmental_Data/Species_genetic_data/SWTH/resources resourcesWe have to change the partitions, etc. But more importantly we will need to modify the memory requirements and the tmpdir in resources and maybe the threads.
We are going to do that in the profile.
We do some exploring first. Here are all the places where the resources get modified in the various rules:
awk 'BEGIN {OFS="\t"} /^rule/ {rule=$2; next} /resources:/ {go=1; print FILENAME, rule, $0; next} /[a-z]+:/ {go=0} /=/ && go==1 {print FILENAME, rule, $0}' ../Snakefile *.smk
calling.smk make_gvcf_sections: resources:
calling.smk make_gvcf_sections: time="1-00:00:00",
calling.smk make_gvcf_sections: mem_mb = 4600,
calling.smk make_gvcf_sections: cpus = 1
calling.smk genomics_db_import_chromosomes: resources:
calling.smk genomics_db_import_chromosomes: mem_mb = 9400,
calling.smk genomics_db_import_chromosomes: cpus = 2,
calling.smk genomics_db_import_chromosomes: time = "36:00:00"
calling.smk genomics_db_import_chromosomes: # job. resources: cpus = 2, does not work. Same for above. I set reader
calling.smk genomics_db_import_scaffold_groups: resources:
calling.smk genomics_db_import_scaffold_groups: mem_mb = 9400,
calling.smk genomics_db_import_scaffold_groups: cpus = 2,
calling.smk genomics_db_import_scaffold_groups: time = "36:00:00"
calling.smk genomics_db2vcf_scattered: resources:
calling.smk genomics_db2vcf_scattered: mem_mb = 11750,
calling.smk genomics_db2vcf_scattered: cpus = 2,
calling.smk genomics_db2vcf_scattered: time = "1-00:00:00"
mapping.smk mark_duplicates: resources:
mapping.smk mark_duplicates: cpus = 1
qc.smk multiqc_dir: resources:
qc.smk multiqc_dir: mem_mb = 36800
ref.smk bwa_index: resources:
ref.smk bwa_index: mem_mb=36900,And here are the only places where we are using more than 1 thread:
awk 'BEGIN {OFS="\t"} /^rule/ {rule=$2; next} /threads:/ {go=1; print FILENAME, rule, $0; next}' ../Snakefile *.smk
angsd-ready-bams.smk realigner_target_creator: threads: 4
annotation.smk annotate_variants: threads: 4
calling.smk make_gvcf_sections: threads: 1
calling.smk genomics_db_import_chromosomes: threads: 2
calling.smk genomics_db_import_scaffold_groups: threads: 2
calling.smk genomics_db2vcf_scattered: threads: 2
mapping.smk map_reads: threads: 4All the nodes we would send jobs to on Alpine have 3.74 Gb of RAM per core. So we merely need to make sure that the threads and the memory are in agreement and then off we go…
I modified the profile accordingly.
I ran this first on SWTH from a login node with:
snakemake --profile hpcc-profiles/slurm/alpine --configfile config/config.yamlIt worked better than expected. But:
- I had allowed too many jobs/cores and it submitted more than was allowed. So some jobs were killed right after submission.
- I also got a lot of OOM error on the map_reads rule. The problem there is that the resources don’t seem to be getting set properly by the –profile values.
I will be able to explore that some more. I will try putting them into the –workflow-profile
To kill it gracefully, from another terminal I grepped snakemake out of
ps -ax and found the job id number and then just used kill to
gracefully shut it down. It continued to run the jobs that were running
(and that were queued…but I scancelled those) until those things
finished. All in all, a very nice experience.
There is not good documentation on the max number of jobs, but I will scale it down to 600 in the profile.
Here are some of the samples that failed on map_reads:
(base) [login11: map_reads]--% grep oom * | awk -F"," '{print $2}' | head -n 20
sample=00N0540
sample=00N2156
sample=00N2157
sample=00N2158
sample=00N2159
sample=00N2160
sample=00N2161
sample=00N5917
sample=00N5929
sample=07N31126
sample=07N31128
sample=07N51683
sample=07N51685
sample=07N51690
sample=07N51755
sample=07N51761
sample=07N51764
sample=07N51768
sample=07N51770
sample=07N51771So, I can test a few of these when I crank the memory back up.
It might be that it needs fewer cores but more memory The mem might scale with the number of threads…
I figured out what was wrong with the profile and it is now working much better.
Also, here is how to compute the total number of CPU hours (SU’s) used from a certain date:
sacct -S2023-01-01 -u eriq@colostate.edu -ojobid,start,end,alloccpu,cputime | column -t | awk '!/[a-z.]/ {n=split($5,a,/:/); h=a[1]; m=a[2]; s=a[3]; fh = h + m/60 + s/3600; cumul += $4 * fh; print $0, fh} END {print "TotalCoreHours:", cumul}'Right now I am here:
2318029 2023-07-06T15:08:31 2023-07-06T15:13:14 4 00:18:52 0.314444
2318033 2023-07-06T15:09:46 2023-07-06T15:15:25 4 00:22:36 0.376667
2318042 2023-07-06T15:09:46 2023-07-06T15:14:49 4 00:20:12 0.336667
2318059 2023-07-06T15:09:46 2023-07-06T15:13:35 4 00:15:16 0.254444
2318069 2023-07-06T15:10:30 2023-07-06T15:14:32 4 00:16:08 0.268889
2318095 2023-07-06T15:10:30 2023-07-06T15:12:42 4 00:08:48 0.146667
2318126 2023-07-06T15:11:47 2023-07-06T15:15:57 4 00:16:40 0.277778
2318133 2023-07-06T15:12:40 2023-07-06T15:15:26 4 00:11:04 0.184444
TotalCoreHours: 312.279OK!
That thing ran for about 12 hours and got quite far, but then ended. There had been some errors.
The dry-run says there are 5 mark_duplicates and 7 clip_overlaps to be done, and then all the other stuff downstream of that.
By grepping out of the log file that snakemake automatically saves we can get the slurm job numbers that failed:
(base) [login11: mega-non-model-wgs-snakeflow]--% grep Error .snakemake/log/2023-07-06T145637.260179.snakemake.log | awk '/executing/ {err[$4] = sprintf("%s %s", err[$4], $10)} END {for(i in err) print i, err[i]}'
make_gvcf_sections 2325409 2325410 2325411 2325412 2325413 2325414 2325415 2325416 2325417 2325419 2325420 2325421 2325422 2325423 2325424 2325425 2325426 2325427 2325428 2325429 2325430
clip_overlaps 2325408 2325418
mark_duplicates 2321682 2322604 2330469 2337276 2337861And now we can investigate those failures at our leisure.
Here is another way to print those that makes it easier to use them:
(base) [login11: mega-non-model-wgs-snakeflow]--% grep Error .snakemake/log/2023-07-06T145637.260179.snakemake.log | awk '/executing/ {err[$4] = sprintf("%s *%s*", err[$4], $10)} END {for(i in err) print i, err[i]}' | sed 's/,//g;'
make_gvcf_sections *2325409* *2325410* *2325411* *2325412* *2325413* *2325414* *2325415* *2325416* *2325417* *2325419* *2325420* *2325421* *2325422* *2325423* *2325424* *2325425* *2325426* *2325427* *2325428* *2325429* *2325430*
clip_overlaps *2325408* *2325418*
mark_duplicates *2321682* *2322604* *2330469* *2337276* *2337861*Note that some of the make_gvcf_sections failed.
# explore these:
# The mark_duplicates fails have slurm logs:
(base) [login11: mark_duplicates]--% ls -l *2321682* *2322604* *2330469* *2337276* *2337861*
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 4021 Jul 6 17:27 mark_duplicates-bqsr_round=0,sample=00N5507-2322604.err
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 0 Jul 6 17:26 mark_duplicates-bqsr_round=0,sample=00N5507-2322604.out
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 3986 Jul 6 21:08 mark_duplicates-bqsr_round=0,sample=08N0723-2337861.err
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 0 Jul 6 21:02 mark_duplicates-bqsr_round=0,sample=08N0723-2337861.out
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 4045 Jul 6 17:05 mark_duplicates-bqsr_round=0,sample=13N01565-2321682.err
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 0 Jul 6 17:03 mark_duplicates-bqsr_round=0,sample=13N01565-2321682.out
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 3986 Jul 6 20:28 mark_duplicates-bqsr_round=0,sample=99N0667-2337276.err
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 0 Jul 6 20:24 mark_duplicates-bqsr_round=0,sample=99N0667-2337276.out
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 3986 Jul 6 18:55 mark_duplicates-bqsr_round=0,sample=99N0674-2330469.err
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 0 Jul 6 18:51 mark_duplicates-bqsr_round=0,sample=99N0674-2330469.out
# and, in every case they were out-of-memory kills:
(base) [login11: mark_duplicates]--% grep kill *2321682* *2322604* *2330469* *2337276* *2337861*
mark_duplicates-bqsr_round=0,sample=13N01565-2321682.err:slurmstepd: error: Detected 1 oom-kill event(s) in StepId=2321682.batch. Some of your processes may have been killed by the cgroup out-of-memory handler.
mark_duplicates-bqsr_round=0,sample=00N5507-2322604.err:slurmstepd: error: Detected 1 oom-kill event(s) in StepId=2322604.batch. Some of your processes may have been killed by the cgroup out-of-memory handler.
mark_duplicates-bqsr_round=0,sample=99N0674-2330469.err:slurmstepd: error: Detected 1 oom_kill event in StepId=2330469.batch. Some of the step tasks have been OOM Killed.
mark_duplicates-bqsr_round=0,sample=99N0667-2337276.err:slurmstepd: error: Detected 1 oom_kill event in StepId=2337276.batch. Some of the step tasks have been OOM Killed.
mark_duplicates-bqsr_round=0,sample=08N0723-2337861.err:slurmstepd: error: Detected 1 oom_kill event in StepId=2337861.batch. Some of the step tasks have been OOM Killed.But I don’t have slurm logs for the other failures. So they must have failed before they got launched by SLURM. Weird…
Let’s see if we can find the logs for them:
# here are the logs for them:
(base) [login11: mega-non-model-wgs-snakeflow]--% grep -A 10 'Error in rule make_gvcf_sections' .snakemake/log/2023-07-06T145637.260179.snakemake.log | grep log:
log: results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046221.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046221.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046224.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046224.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/07N51764/NC_046223.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/07N51764/NC_046223.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/00N5500/NC_046258.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/00N5500/NC_046258.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/97N6564/NC_046221.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/97N6564/NC_046221.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046225.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046225.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/07N52116/NC_046233.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/07N52116/NC_046233.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/07N51764/NC_046226.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/07N51764/NC_046226.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046233.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046233.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/97N6564/NC_046224.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/97N6564/NC_046224.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/97N6564/NC_046233.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/97N6564/NC_046233.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/05N5310/NC_046252.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/05N5310/NC_046252.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046235.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/AB16098/NC_046235.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/14N02035/NC_046230.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/14N02035/NC_046230.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/97N6564/NC_046230.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/97N6564/NC_046230.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/96N0608/NC_046247.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/96N0608/NC_046247.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/98N1131/NC_046262.2.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/98N1131/NC_046262.2.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/99N0668/NC_046251.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/99N0668/NC_046251.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/00N5502/NC_046225.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/00N5502/NC_046225.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/07N53531/NC_046230.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/07N53531/NC_046230.1.stdout (check log file(s) for error details)
log: results/bqsr-round-0/logs/gatk/haplotypecaller/21N01277/NC_046246.1.stderr, results/bqsr-round-0/logs/gatk/haplotypecaller/21N01277/NC_046246.1.stdout (check log file(s) for error details)And those are all empty.
So, now let us check sacct for those job IDs.
(base) [login11: mega-non-model-wgs-snakeflow]--% sacct -j 2325409,2325410,2325411
JobID JobName Partition Account AllocCPUS State ExitCode
------------ ---------- ---------- ---------- ---------- ---------- --------
2325409 smk-make_+ amilan csu-gener+ 1 FAILED 0:53
2325409.bat+ batch csu-gener+ 1 CANCELLED 0:53
2325409.ext+ extern csu-gener+ 1 COMPLETED 0:0
2325410 smk-make_+ amilan csu-gener+ 1 FAILED 0:53
2325410.bat+ batch csu-gener+ 1 CANCELLED 0:53
2325410.ext+ extern csu-gener+ 1 COMPLETED 0:0
2325411 smk-make_+ amilan csu-gener+ 1 FAILED 0:53
2325411.bat+ batch csu-gener+ 1 CANCELLED 0:53
2325411.ext+ extern csu-gener+ 1 COMPLETED 0:0Hey! Those jobs just got cancelled before running. Probably no big deal.
And the clip-overlaps:
(base) [login11: mega-non-model-wgs-snakeflow]--% sacct -j 2325408,2325418
JobID JobName Partition Account AllocCPUS State ExitCode
------------ ---------- ---------- ---------- ---------- ---------- --------
2325408 smk-clip_+ amilan csu-gener+ 1 FAILED 0:53
2325408.bat+ batch csu-gener+ 1 CANCELLED 0:53
2325408.ext+ extern csu-gener+ 1 COMPLETED 0:0
2325418 smk-clip_+ amilan csu-gener+ 1 FAILED 0:53
2325418.bat+ batch csu-gener+ 1 CANCELLED 0:53
2325418.ext+ extern csu-gener+ 1 COMPLETED 0:0Same thing—just got cancelled.
And finally, let’s check the other mark_duplicate runs this way. Here are the numbers in an easier to use format:
grep Error .snakemake/log/2023-07-06T145637.260179.snakemake.log | awk '/executing/ {err[$4] = sprintf("%s%s", err[$4], $10)} END {for(i in err) print i, err[i]}'
make_gvcf_sections 2325409,2325410,2325411,2325412,2325413,2325414,2325415,2325416,2325417,2325419,2325420,2325421,2325422,2325423,2325424,2325425,2325426,2325427,2325428,2325429,2325430,
clip_overlaps 2325408,2325418,
mark_duplicates 2321682,2322604,2330469,2337276,2337861(base) [login11: mega-non-model-wgs-snakeflow]--% sacct -j 2321682,2322604,2330469,2337276,2337861
JobID JobName Partition Account AllocCPUS State ExitCode
------------ ---------- ---------- ---------- ---------- ---------- --------
2321682 smk-mark_+ amilan csu-gener+ 1 OUT_OF_ME+ 0:125
2321682.bat+ batch csu-gener+ 1 OUT_OF_ME+ 0:125
2321682.ext+ extern csu-gener+ 1 OUT_OF_ME+ 0:125
2322604 smk-mark_+ amilan csu-gener+ 1 OUT_OF_ME+ 0:125
2322604.bat+ batch csu-gener+ 1 OUT_OF_ME+ 0:125
2322604.ext+ extern csu-gener+ 1 OUT_OF_ME+ 0:125
2330469 smk-mark_+ amilan csu-gener+ 1 OUT_OF_ME+ 0:125
2330469.bat+ batch csu-gener+ 1 OUT_OF_ME+ 0:125
2330469.ext+ extern csu-gener+ 1 COMPLETED 0:0
2337276 smk-mark_+ amilan csu-gener+ 1 OUT_OF_ME+ 0:125
2337276.bat+ batch csu-gener+ 1 OUT_OF_ME+ 0:125
2337276.ext+ extern csu-gener+ 1 COMPLETED 0:0
2337861 smk-mark_+ amilan csu-gener+ 1 OUT_OF_ME+ 0:125
2337861.bat+ batch csu-gener+ 1 OUT_OF_ME+ 0:125
2337861.ext+ extern csu-gener+ 1 COMPLETED 0:0Yep! All out of memories. So we just give those more memory when starting it all up again. It does not look like it ran out of Java memory—the SLURM controller killed it. So we will just restart it with three times as much memory.
It only has the default 3700 Mb so, let’s give it 11220 Mb. We can do that on the command line with –set-resources:
# check it with a dry run
snakemake -np --profile hpcc-profiles/slurm/alpine --configfile config/config.yaml --set-resources mark_duplicates:mem_mb=11220
# that dry run showed that the memory had been pumped up on the mark duplicates runs
# so, now, start it up for real.
snakemake --profile hpcc-profiles/slurm/alpine --configfile config/config.yaml --set-resources mark_duplicates:mem_mb=11220That run finished after a while, but three of the mark_duplicates jobs ran out of memory (and these are the biggest files, too). So, I am going to start them again and give them 8 cores worth of memory: 29920.
snakemake --profile hpcc-profiles/slurm/alpine --configfile config/config.yaml --set-resources mark_duplicates:mem_mb=29920Hey! Check this out…if you apply --set-resources on the command line,
it completely overwrites the values set with set-resources: in the
profile. That is kind of BS, I’ve gotta say.
But, good to know. The take-home message is: if you want to change the value of one resource for one rule, just go ahead and make that change in the profile, not on the command line. Lame! Sad! Bogus! (But I guess I can see why that is…multiple invocations of –set-resources completely overwrite previous ones…) As a consequence of this choice, however, I am going to have to restart this because the import_genomics_db rules reverted back to the 36 hour time limit, which exceeds the QoS. No worries. I soft-killed it and am waiting for the final make_gvcf rules to finish out.
When it was done I just started it up again with:
snakemake --profile hpcc-profiles/slurm/alpine --configfile config/config.yamlThat ran along fine. Until I got an error:
sacct failed to return the status for jobid 2355542
Maybe you need to use scontrol instead?
Failed to obtain job status. See above for error message.Checking it myself I see this:
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--% sacct -j 2355542 --parsable2
JobID|JobName|Partition|Account|AllocCPUS|State|ExitCode
2355542|smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0020,sg_or_chrom=NC_046222.1|amilan|csu-general|2|FAILED|1:0
2355542.batch|batch||csu-general|2|FAILED|1:0
2355542.extern|extern||csu-general|2|COMPLETED|0:0
# and here is what the slurm log says:
(base) [login11: genomics_db2vcf_scattered]--% head -n 200 *2355542*
==> genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0020,sg_or_chrom=NC_046222.1-2355542.err <==
Waiting at most 60 seconds for missing files.
Traceback (most recent call last):
File "<frozen runpy>", line 198, in _run_module_as_main
File "<frozen runpy>", line 88, in _run_code
File "/projects/eriq@colostate.edu/mambaforge/envs/snakemake-7.30.1/lib/python3.11/site-packages/snakemake/__main__.py", line 4, in <module>
main()
File "/projects/eriq@colostate.edu/mambaforge/envs/snakemake-7.30.1/lib/python3.11/site-packages/snakemake/__init__.py", line 3079, in main
wait_for_files([args.wait_for_files_file], latency_wait=args.latency_wait)
File "/projects/eriq@colostate.edu/mambaforge/envs/snakemake-7.30.1/lib/python3.11/site-packages/snakemake/io.py", line 878, in wait_for_files
raise IOError(
OSError: Missing files after 60 seconds. This might be due to filesystem latency. If that is the case, consider to increase the wait time with --latency-wait:
/scratch/alpine/eriq@colostate.edu/mega-non-model-wgs-snakeflow/.snakemake/tmp.wxmfac_p/snakejob.genomics_db2vcf_scattered.18166.sh.waitforfilesfile.txtThe main job does not exist. Whoa! This job did not create a log at all! It is like it died super early on. Perhaps the genomics db is corrupted.
In fact all of the genomics_db2vcf_scattered rule runs for sg_or_chrom=NC_046222.1 failed in exactly the same way. Perhaps the genomicsdbimport rule was done but the files had not all appeared, because of filesystem latency, but the receipts were there, and so all the genotypeGVCFs jobs got confused and failed as soon as they were launched (they were all launched with 2 seconds of one another).
These jobs are still running, even though snakemake aborted itself:
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) CPUS MIN_MEMORY TIME_LIMIT TIME_LEFT PRIORITY
2355521 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0008,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355522 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0037,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355523 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0030,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355524 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0023,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355525 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0051,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355526 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0020,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355527 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0048,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355528 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0041,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355529 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0006,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355530 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0034,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355531 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0027,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-c11-u34-4 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355532 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0055,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-c11-u34-4 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355533 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0045,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-c11-u34-4 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355534 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0017,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-c11-u34-4 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355535 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0009,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-c11-u34-4 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355536 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0003,sg_or_chrom=NC_046221.1 eriq@colos R 25:40 1 c3cpu-c11-u34-4 2 11000M 1-00:00:00 23:34:20 0.00000378931873
2355501 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0043,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a9-u5-2 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355502 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0038,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a9-u5-2 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355503 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0013,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a9-u5-2 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355504 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0036,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a5-u15-2 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355505 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0001,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a5-u15-2 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355506 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0029,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a5-u15-2 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355507 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0022,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a5-u15-2 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355508 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0050,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a5-u15-2 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355509 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0019,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355510 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0047,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355511 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0011,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355512 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0040,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355513 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0007,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355514 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0005,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355515 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0033,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355516 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0002,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355517 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0026,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355518 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0054,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355519 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0016,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355520 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0044,sg_or_chrom=NC_046221.1 eriq@colos R 26:10 1 c3cpu-a2-u3-1 2 11000M 1-00:00:00 23:33:50 0.00000378931873
2355481 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0024,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-c9-u1-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355482 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0052,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-c9-u1-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355483 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0031,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-c11-u3-3 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355484 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0014,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-c11-u3-3 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355485 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0042,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-c11-u3-4 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355486 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0035,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a9-u1-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355487 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0028,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a9-u1-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355488 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0056,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a7-u1-2 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355489 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0021,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a7-u1-2 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355490 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0049,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a7-u1-2 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355491 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0018,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a7-u1-2 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355492 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0046,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a5-u1-2 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355493 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0039,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a5-u1-2 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355494 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0010,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a2-u1-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355495 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0004,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a2-u1-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355496 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0032,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a2-u1-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355497 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0012,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-a2-u1-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355498 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0025,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-c9-u5-2 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355499 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0053,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-c11-u5-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355500 amilan smk-genomics_db2vcf_scattered-bqsr_round=0,scatter=scat_0015,sg_or_chrom=NC_046221.1 eriq@colos R 26:40 1 c3cpu-c11-u5-1 2 11000M 1-00:00:00 23:33:20 0.00000378931873
2355540 amilan smk-bung_filtered_vcfs_back_together-bqsr_round=0,sg_or_chrom=NC_046224.1 eriq@colos R 23:09 1 c3cpu-c9-u1-1 1 3740M 8:00:00 7:36:51 0.00000378908590
2355472 amilan smk-mark_dp0_as_missing-bqsr_round=0,sg_or_chrom=NC_046223.1 eriq@colos R 35:43 1 c3cpu-c9-u1-1 1 3740M 8:00:00 7:24:17 0.00000378908590I will check the logs on these when they are done to make sure that they finished correctly.
I did check that. They were all fine. I restarted it and then we ran into two out-of-memories on:
(base) [login11: mega-non-model-wgs-snakeflow]--% grep Error .snakemake/log/2023-07-08T173040.213753.snakemake.log
Error in rule genomics_db2vcf_scattered:
Error executing rule genomics_db2vcf_scattered on cluster (jobid: 19130, external: 2357623, jobscript: /scratch/alpine/eriq@colostate.edu/mega-non-model-wgs-snakeflow/.snakemake/tmp.ybs08jv_/snakejob.genomics_db2vcf_scattered.19130.sh). For error details see the cluster log and the log files of the involved rule(s).
Error in rule genomics_db2vcf_scattered:
Error executing rule genomics_db2vcf_scattered on cluster (jobid: 19128, external: 2357643, jobscript: /scratch/alpine/eriq@colostate.edu/mega-non-model-wgs-snakeflow/.snakemake/tmp.ybs08jv_/snakejob.genomics_db2vcf_scattered.19128.sh). For error details see the cluster log and the log files of the involved rule(s).
Predictably, these were scat006 and scat007 on scaff_group_001. We can check the logs to see how far they got.
They only got about half way through. Such BS. I am just going to give each sequence its own scatter group in those and let it work it out that way:
(base) [login11: config]--% pwd
/home/eriq@colostate.edu/scratch/mega-non-model-wgs-snakeflow/config
(base) [login11: config]--% awk 'BEGIN {OFS="\t"} $1=="scaff_group_001" && $2~/scat_000[67]/ {new2 = $2 "." ++n; print $1, new2, $3, $4, $5, $6; next} {print}' scatters_3000000.tsv > scatters_3000000_exploded.tsv
# and then we edit the config.yaml to use scatters_3000000_exploded.tsvAfter doing a dry run, it was clear that we need to set the modification date for scatters_3000000_exploded.tsv to earlier since that file is considered an input, and so it will want to rerun all the GenotypeGvcfs steps. No worries:
touch -d "last Monday" config/scatters_3000000_exploded.tsv
# and check:
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--% ls -l config/scatters_3000000_exploded.tsv
-rw-r--r--. 1 eriq@colostate.edu eriqgrp@colostate.edu 30371 Jul 3 00:00 config/scatters_3000000_exploded.tsv(Pretty rad that you can use “last Monday” to do that!)
And that finished up by Sunday noon. So 3-days total to get that processed. Not bad.
And check how many CPU hours that was:
sacct -S2023-07-03 -u eriq@colostate.edu -ojobid,start,end,alloccpu,cputime | column -t | awk '!/[a-z.]/ {n=split($5,a,/:/); h=a[1]; m=a[2]; s=a[3]; fh = h + m/60 + s/3600; cumul += $4 * fh;} END {print "TotalCoreHours:", cumul}'
TotalCoreHours: 8875.77OK!
There were about 70 birds that had been re-run, but the seq center put a Z in front of their names, so I didn’t realize they were the same birds. So, the units file will need to be re-made to show two units for each of those birds.
# doing this on my laptop
library(tidyverse)
old_units <- read_tsv("~/Documents/git-repos/mega-non-model-configs/SWTH-2023-07-05/config/OLD_units.tsv")
# learn a bit about these. There are 74 samples starting with a Z
old_units %>%
count(str_detect(sample, "^Z"))So, which of those are re-runs?
old_units %>%
mutate(noZ = str_remove(sample, "^Z")) %>%
count(noZ) %>%
count(n)So, all 74 of those are re-runs. Thus, we can just take the Z off the sample and sample_id fields and then put them together and amend the units column:
new_units <- old_units %>%
mutate(
sample = str_remove(sample, "^Z"),
sample_id = str_remove(sample_id, "^Z")
) %>%
arrange(sample, library) %>%
group_by(sample) %>%
mutate(unit = 1:n())But, I need to ask CH if these were completely new library preps or not, because that affects the duplicate marking stage. OK! They were completely new library preps, so new_units is correct as it it, with different library preps in there for the resequences.
So, now I can save that as the units file
write_tsv(new_units, file = "~/Documents/git-repos/mega-non-model-configs/SWTH-2023-07-05/config/units.tsv")Then, when I used that new units file with:
snakemake -np --profile hpcc-profiles/slurm/alpine --configfile config/config.yamlit said everything was done. I am a little surprised by that, because it seems like it should notice that the inputs are different to some of the rules. But perhaps not, since the file modification times are not any different on all those. So, there are various ways that I could conceive of dealing with that…
For fun, I could try touching the data files for the 74 “Z” fastqs. But the problem with that is that it will trigger new fastqc runs that are not necessary. But, that might be the cleanest way to do it anyway. I will try:
(base) [login11: SWTH_Novoseq_Plate2]--% pwd
/home/eriq@colostate.edu/scratch/mega-non-model-wgs-snakeflow/data/SWTH_Novoseq_Plate2
(base) [login11: SWTH_Novoseq_Plate2]--% touch Z*/*fq.gz
# that was two files for each of the 74 reruns:
(base) [login11: SWTH_Novoseq_Plate2]--% ls -l Z*/*fq.gz | wc
148 1332 20164Now, when I give it a dry run, I see:
Job stats:
job count min threads max threads
-------------------- ------- ------------- -------------
all 1 1 1
clip_overlaps 74 1 1
concat_gvcf_sections 74 1 1
fastqc_read1 74 1 1
fastqc_read2 74 1 1
make_gvcf_sections 3108 1 1
map_reads 148 4 4
mark_duplicates 74 1 1
multiqc_dir 1 1 1
samtools_stats 74 1 1
trim_reads_pe 148 1 1
total 3850 1 4Note that there are 41 “chromosomes” and 1 scaffold group. And 74 * 42 = 3108.
So, that looks to be right. It is going to have to re-trim everything, because I deleted the trimmed files, but that is not such a big deal.
Also, those fastqc’s do have to be redone because the naming of them have changed.
But, what is interesting here is that this has not triggered a re-run of the genomics data bases and beyond. I expect that is because of the way that I rigged things so that I could add individuals to the genomics data bases. At some point I should jettison that stuff. But for now I can just delete all the genomicsDBs. They have to be redone anyway!
(base) [login11: bqsr-round-0]--% pwd
/home/eriq@colostate.edu/scratch/mega-non-model-wgs-snakeflow/results/bqsr-round-0
(base) [login11: bqsr-round-0]--% rm -rf gdb_intervals gdb_accounting genomics_dbNow that I have done that, my dry run looks like:
Job stats:
job count min threads max threads
---------------------------------- ------- ------------- -------------
all 1 1 1
bcf_concat 1 1 1
bcf_concat_mafs 1 1 1
bcf_maf_section_summaries 42 1 1
bcf_section_summaries 126 1 1
bung_filtered_vcfs_back_together 42 1 1
clip_overlaps 74 1 1
combine_bcftools_stats 3 1 1
combine_maf_bcftools_stats 1 1 1
concat_gvcf_sections 74 1 1
fastqc_read1 74 1 1
fastqc_read2 74 1 1
gather_scattered_vcfs 42 1 1
genomics_db2vcf_scattered 494 2 2
genomics_db_import_chromosomes 41 2 2
genomics_db_import_scaffold_groups 1 2 2
hard_filter_indels 42 1 1
hard_filter_snps 42 1 1
maf_filter 42 1 1
make_gvcf_sections 3108 1 1
make_indel_vcf 42 1 1
make_snp_vcf 42 1 1
map_reads 148 4 4
mark_dp0_as_missing 42 1 1
mark_duplicates 74 1 1
multiqc_dir 1 1 1
samtools_stats 74 1 1
trim_reads_pe 148 1 1
total 4896 1 4That looks just about right! It also looks like I have removed all of
the protected() statements in the workflow, so this should not run
into any problems. Let’s load this bad boy up!
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--% snakemake -p --profile hpcc-profiles/slurm/alpine --configfile config/config.yamlBy morning that had ended with some failures. It was easy to find them:
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--% grep Error .snakemake/log/2023-07-27T160926.747195.snakemake.log | awk '/executing/ {err[$4] = sprintf("%s%s", err[$4], $10)} END {for(i in err) print i, err[i]}'
mark_duplicates 2516965,2517268,2517263,2517378,2517370,2518710And those were out of memory, and some that just failed.
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--% sacct -j 2516965,2517268,2517263,2517378,2517370,2518710
JobID JobName Partition Account AllocCPUS State ExitCode
------------ ---------- ---------- ---------- ---------- ---------- --------
2516965 smk-mark_+ amilan csu-gener+ 1 OUT_OF_ME+ 0:125
2516965.bat+ batch csu-gener+ 1 OUT_OF_ME+ 0:125
2516965.ext+ extern csu-gener+ 1 COMPLETED 0:0
2517263 smk-mark_+ amilan csu-gener+ 1 FAILED 1:0
2517263.bat+ batch csu-gener+ 1 FAILED 1:0
2517263.ext+ extern csu-gener+ 1 COMPLETED 0:0
2517268 smk-mark_+ amilan csu-gener+ 1 FAILED 1:0
2517268.bat+ batch csu-gener+ 1 FAILED 1:0
2517268.ext+ extern csu-gener+ 1 COMPLETED 0:0
2517370 smk-mark_+ amilan csu-gener+ 1 OUT_OF_ME+ 0:125
2517370.bat+ batch csu-gener+ 1 OUT_OF_ME+ 0:125
2517370.ext+ extern csu-gener+ 1 COMPLETED 0:0
2517378 smk-mark_+ amilan csu-gener+ 1 OUT_OF_ME+ 0:125
2517378.bat+ batch csu-gener+ 1 OUT_OF_ME+ 0:125
2517378.ext+ extern csu-gener+ 1 COMPLETED 0:0
2518710 smk-mark_+ amilan csu-gener+ 1 OUT_OF_ME+ 0:125
2518710.bat+ batch csu-gener+ 1 OUT_OF_ME+ 0:125
2518710.ext+ extern csu-gener+ 1 COMPLETED 0:0So, I will just change the profile to give mark_duplicates more memory and then I will restart.
Here is what remains:
Job stats:
job count min threads max threads
---------------------------------- ------- ------------- -------------
all 1 1 1
bcf_concat 1 1 1
bcf_concat_mafs 1 1 1
bcf_maf_section_summaries 42 1 1
bcf_section_summaries 126 1 1
bung_filtered_vcfs_back_together 42 1 1
clip_overlaps 6 1 1
combine_bcftools_stats 3 1 1
combine_maf_bcftools_stats 1 1 1
concat_gvcf_sections 6 1 1
gather_scattered_vcfs 42 1 1
genomics_db2vcf_scattered 494 2 2
genomics_db_import_chromosomes 41 2 2
genomics_db_import_scaffold_groups 1 2 2
hard_filter_indels 42 1 1
hard_filter_snps 42 1 1
maf_filter 42 1 1
make_gvcf_sections 252 1 1
make_indel_vcf 42 1 1
make_snp_vcf 42 1 1
mark_dp0_as_missing 42 1 1
mark_duplicates 6 1 1
multiqc_dir 1 1 1
samtools_stats 6 1 1
total 1324 1 2It seems to be working great.
It did as much as it could but it threw two errors on genomics_db2vcf rule runs. sacct says that both of the jobs finished, so it must just be a job latency thing.
It said:
sacct failed to return the status for jobid 2527084
Maybe you need to use scontrol instead?
Failed to obtain job status. See above for error message.
WorkflowError:
Failed to obtain job status. See above for error message.
File "/projects/eriq@colostate.edu/mambaforge/envs/snakemake-7.30.1/lib/python3.11/asyncio/runners.py", line 190, in run
File "/projects/eriq@colostate.edu/mambaforge/envs/snakemake-7.30.1/lib/python3.11/asyncio/runners.py", line 118, in run
File "/projects/eriq@colostate.edu/mambaforge/envs/snakemake-7.30.1/lib/python3.11/asyncio/base_events.py", line 653, in run_until_complete
Shutting down, this might take some time.
Exiting because a job execution failed. Look above for error message
Complete log: .snakemake/log/2023-07-28T105355.438230.snakemake.log(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--% grep Error .snakemake/log/2023-07-28T105355.438230.snakemake.log | awk '/executing/ {err[$4] = sprintf("%s%s", err[$4], $10)} END {for(i in err) print i, err[i]}'
genomics_db2vcf_scattered 2525063,2525064,
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--%
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--%
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--%
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--% sacct -j 2525063,2525064
JobID JobName Partition Account AllocCPUS State ExitCode
------------ ---------- ---------- ---------- ---------- ---------- --------
2525063 smk-genom+ amilan csu-gener+ 2 COMPLETED 0:0
2525063.bat+ batch csu-gener+ 2 COMPLETED 0:0
2525063.ext+ extern csu-gener+ 2 COMPLETED 0:0
2525064 smk-genom+ amilan csu-gener+ 2 COMPLETED 0:0
2525064.bat+ batch csu-gener+ 2 COMPLETED 0:0
2525064.ext+ extern csu-gener+ 2 COMPLETED 0:0I checked the log of one of those and GATK clearly reported that it had finished.
When I did a dry-run it showed 500 or so jobs to go. But there are still a few jobs running in SLURM:
(snakemake-7.30.1) [login11: mega-non-model-wgs-snakeflow]--% myjobs 40
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) CPUS MIN_MEMORY TIME_LIMIT TIME_LEFT PRIORITY
2524840 amilan smk-make_gvcf_sections-bqsr_round=0,samp eriq@colos R 3:31:13 1 c3cpu-c11-u15-2 1 3600M 1-00:00:00 20:28:47 0.00000394601375
2524820 amilan smk-make_gvcf_sections-bqsr_round=0,samp eriq@colos R 3:31:51 1 c3cpu-c11-u13-1 1 3600M 1-00:00:00 20:28:09 0.00000394601375
2524710 amilan smk-make_gvcf_sections-bqsr_round=0,samp eriq@colos R 3:36:02 1 c3cpu-c9-u3-3 1 3600M 1-00:00:00 20:23:58 0.00000393949449
2524716 amilan smk-make_gvcf_sections-bqsr_round=0,samp eriq@colos R 3:36:02 1 c3cpu-c11-u1-1 1 3600M 1-00:00:00 20:23:58 0.00000393949449
2524626 amilan smk-make_gvcf_sections-bqsr_round=0,samp eriq@colos R 3:39:07 1 c3cpu-c9-u1-4 1 3600M 1-00:00:00 20:20:53 0.00000393926166
2524629 amilan smk-make_gvcf_sections-bqsr_round=0,samp eriq@colos R 3:39:07 1 c3cpu-c9-u1-4 1 3600M 1-00:00:00 20:20:53 0.00000393926166
2525981 amilan smk-genomics_db_import_chromosomes-bqsr_ eriq@colos R 1:58:18 1 c3cpu-c9-u1-1 2 7480M 1-00:00:00 22:01:42 0.00000393856317
2525942 amilan smk-genomics_db_import_chromosomes-bqsr_ eriq@colos R 2:04:23 1 c3cpu-c9-u1-2 2 7480M 1-00:00:00 21:55:37 0.00000393856317
2525803 amilan smk-genomics_db_import_chromosomes-bqsr_ eriq@colos R 2:19:09 1 c3cpu-c9-u1-1 2 7480M 1-00:00:00 21:40:51 0.00000393856317
2527010 amilan smk-genomics_db_import_scaffold_groups-b eriq@colos R 1:01:49 1 c3cpu-c9-u1-1 2 11000M 1-00:00:00 22:58:11 0.00000393716619
2527016 amilan smk-genomics_db_import_chromosomes-bqsr_ eriq@colos R 1:00:18 1 c3cpu-c9-u1-2 2 7480M 1-00:00:00 22:59:42 0.00000393693335
2526759 amilan smk-genomics_db_import_chromosomes-bqsr_ eriq@colos R 1:28:07 1 c3cpu-c9-u1-1 2 7480M 1-00:00:00 22:31:53 0.00000393693335
So, that is a weird error. I am going to let the currently running jobs finish, and then check with sacct that they finished OK.
Wen those were done, I did a dry-run, it seemed OK, so I ran snakemake
and it all seemed OK until it started trying to concatenate the BCFs.
Some of the BCFs still had the old count of individuals in there—the
ones starting with Zs.
That was a bit of a mess. A couple of observations:
- I think it is a serious error when
sacctdoesn’t manage to return a status for a job. The sacct_status_robust.sh tries 5 times before it fails. I suspect there are times when the slurm controller is just messed up and failing to reply at some point. I might try adding more attempts (20, instead of 5) and also give it asleep 3after each attempt. - I am assuming that error was the problem. For some reason, after it occurred, re-runs of certain files were not triggered.
I checked the GATK logs, and I saw that all the genomics data bases had gotten properly regenerated, but there were about 8 or 9 chromosomes that had not gotten properly updated. So, I will track those down and remove them, and that should trigger a re-run for them.
Here is what I did:
- Remove all the vcf sections that were from before July 10:
rm $(ls -lrt results/bqsr-round-0/vcf_sections/*/* | awk '$7 < 10 {print $NF}')
rm $(ls -lrt results/bqsr-round-0/vcf_sections/*.gz | awk '$7<10 {print $NF}' )
rm $(ls -lrt results/bqsr-round-0/vcf_sect_miss_denoted/* | awk '$7<10 {print $NF}')After a dry run it looked to me like I should just re-do the bcf stats
rm -r results/bqsr-round-0/qc/bcftools_stats/{maf_sections,sections}
Now, the dry-run looks like this:
Job stats:
job count min threads max threads
-------------------------------- ------- ------------- -------------
all 1 1 1
bcf_concat 1 1 1
bcf_concat_mafs 1 1 1
bcf_maf_section_summaries 42 1 1
bcf_section_summaries 126 1 1
bung_filtered_vcfs_back_together 13 1 1
combine_bcftools_stats 3 1 1
combine_maf_bcftools_stats 1 1 1
gather_scattered_vcfs 13 1 1
genomics_db2vcf_scattered 387 2 2
hard_filter_indels 13 1 1
hard_filter_snps 13 1 1
maf_filter 13 1 1
make_indel_vcf 13 1 1
make_snp_vcf 13 1 1
mark_dp0_as_missing 13 1 1
total 666 1 2
That looks right. I will set that a-running after hacking the sacct-status-robust script.
That all ran without a hitch.
To copy it back to the sharepoint I did:
snakemake -n --profile hpcc-profiles/slurm/alpine send_to_gdrive2 --configfile config/config.yamlin order to get the code:
git log | head -n 150 > config/latest-git-commits.txt; mkdir -p results/qc_summaries/bqsr-round-{0..0}; for i in {0..0}; do cp -r results/bqsr-round-$i/qc/{multiqc.html,bcftools_stats/*.txt} results/qc_summaries/bqsr-round-$i/; done; for i in {0..0}; do tar -cvf results/bqsr-round-$i/qc.tar results/bqsr-round-$i/qc; gzip results/bqsr-round-$i/qc.tar; done; for i in {0..0}; do tar -cvf results/bqsr-round-$i/benchmarks.tar results/bqsr-round-$i/benchmarks; gzip results/bqsr-round-$i/benchmarks.tar; done; for i in {0..0}; do tar -cvf results/bqsr-round-$i/logs.tar results/bqsr-round-$i/logs; gzip results/bqsr-round-$i/logs.tar; done; rclone copy --dry-run --drive-stop-on-upload-limit . onedrive2:BGP_Share/Genetic_and_Environmental_Data/Species_genetic_data/SWTH-2 --include='config/**' --include='results/qc_summaries/**' --include='results/bqsr-round-0/{qc,benchmarks,logs}.tar.gz' --include='results/bqsr-round-0/bcf/**' --include='resources/**' --include='data/**' --include='results/bqsr-round-0/gvcf/*' --include='results/bqsr-round-0/mkdup/*' --include='results/bqsr-round-0/overlap_clipped/*'I am only going to copy the overlap_clipped and the bcf up there, and we also need to copy the DS_control direcctory: So I went with:
rclone copy --dry-run --drive-stop-on-upload-limit . \
onedrive2:BGP_Share/Genetic_and_Environmental_Data/Species_genetic_data/SWTH-2 \
--include='config/**' --include='results/qc_summaries/**' \
--include='results/bqsr-round-0/{qc,benchmarks,logs}.tar.gz' \
--include='results/bqsr-round-0/bcf/**' --include='resources/**' \
--include='results/bqsr-round-0/overlap_clipped/**' \
--include='results/bqsr-round-0/DS_control/**'