Hej hej, I am trying to run the ampliseq pipeline for a set of approximately 400 samples. I had a previous run that conducted the DADA2 steps and taxonomic assignment and now I am trying to resume the pipeline to construct a phylogenetic tree using via the pplace parameters. However, I am running into a CPU time limitation very early in the pipeline, although most of the tasks have been done in the previous run. [1f/4bc336] NFC…Eco_101241_2024_w09_shark) | 182 of 182, cached: 182 [1e/ead627] NFC…Eco_101241_2024_w09_shark) | 182 of 182, cached: 182 [1e/62a5f3] NFC…Eco_101241_2024_w09_shark) | 182 of 182, cached: 182 [- ] NFC…KFLOW:CUTADAPT_SUMMARY_STD - [- ] NFC…LOW:CUTADAPT_SUMMARY_MERGE - [- ] NFC…EPROCESSING:DADA2_QUALITY1 - [- ] NFC…DA2_PREPROCESSING:TRUNCLEN - [- ] NFC…PROCESSING:DADA2_FILTNTRIM - [- ] NFC…EPROCESSING:DADA2_QUALITY2 - [- ] NFC…MPLISEQAMPLISEQDADA2_ERR - [- ] NFC…QAMPLISEQDADA2_DENOISING - [- ] NFC…QAMPLISEQDADA2_RMCHIMERA - [- ] NFC…LISEQAMPLISEQDADA2_STATS - [- ] NFC…LISEQAMPLISEQDADA2_MERGE - [- ] NFC…QAMPLISEQMERGE_STATS_STD - [- ] NFC…_AMPLISEQAMPLISEQBARRNAP - [- ] NFC…EQAMPLISEQBARRNAPSUMMARY - [93/b93e7f] NFC…QAMPLISEQFORMAT_TAXONOMY | 1 of 1, cached: 1 ✔️ [- ] NFC…TAXONOMY_WF:DADA2_TAXONOMY - [- ] NFC…XONOMY_WF:DADA2_ADDSPECIES - Plus 20 more processes waiting for tasks… CPU time limit exceeded My params-file looks like this: { "input": "samplesheet_all.tsv", "FW_primer": "CCTACGGGNGGCWGCAG", "RV_primer": "GACTACHVGGGTATCTAATCC", "outdir": "results_all_new", "multiple_sequencing_runs": true, "dada_ref_taxonomy": "sbdi-gtdb", "pplace_tree": "/cfs/klemming/projects/snic/naiss2023-22-1217/ampliseq_analysis/SBDI_GTDB_v11/sbdi-gtdb-sativa.r10rs226-2.bac120.sprepr.brlenopt.newick", "pplace_aln": "/cfs/klemming/projects/snic/naiss2023-22-1217/ampliseq_analysis/SBDI_GTDB_v11/sbdi-gtdb-sativa.r10rs226-2.bac120.sprepr.alnfna", "pplace_taxonomy": "/cfs/klemming/projects/snic/naiss2023-22-1217/ampliseq_analysis/SBDI_GTDB_v11/sbdi-gtdb-sativa.r10rs226-2.bac120.sprepr.taxonomy.tsv", "pplace_model": "GTR+F+I+G4", "pplace_name": "bac120.r10rs226", "skip_qiime_downstream": true, "skip_barplot": true, "skip_abundance_tables": true, "skip_alpha_rarefaction": true, "skip_ancom": true } And I start the pipeline like this: nextflow run nf-core/ampliseq -r 2.16.1 -resume -name NicolaiD6 -profile pdc_kth -params-file nf-params.json --project naiss2025-22-1670 Thank you for your help!

A few comments and further questions: • Which version did you run originally? If this was not 2.16.1 your cache might be invalid. • I'm not sure what

CPU time limit exceeded

actually means. Are you perhaps running nextflow at the login node? If so, try allocating an interactive node, login to the node you get and run Nextflow from that (within screen). Example command:

salloc -n 7 -N 1 --mem 6000 -t 12:00:00 -A naiss2025-22-1670 -p shared

• If the only purpose of this run is phylogenetic placement, maybe it's easier for you to run #C04NLA7QHH9 directly from the fasta file you got from the original #CEA7TBJGJ run? (This would also require less resources at the login node since Nextflow won't have to figure out the cache.)

I used 2.16.1 for the original run as well and I am running nextflow within a screen. I got the same error though when I ran the whole ampliseq pipeline (DADA2, tax assignment and ppplace) from the beginning with a reduced raw dataset I will try to look into just running phyloplace! But I find the error message weird. I had one try where it "resumed" through the DADA2 and tax assignment and then wanted to start ppplace but aborted since I had a typo in my paramsfile. So, there was no CPU limitation error

Right, I don't think I've seen that either. My guess is that it relates to your login node session, i.e. you've been running too much in that process. Since you're running screen, each time you login you will continue in the same process, so if they have a limit on cpu usage per process you might have reached that. Starting a new screen session would start from scratch. But, note, this is me guessing.

I started a new screen session and I got the same error. So, I guess my next try will be to just run ppplace. Although I have quite confused, because that has never happened before in the last 1-2 years

I haven't either. I think #C04NLA7QHH9 is better suited since you already denoised your sequences, but if that fails similarly, try allocating an interactive node for the Nextflow job. Or, send also the Nextflow interactive job to Slurm. The disadvantage of the latter is that you don't see as clearly what's happening since stdout will be written to file.

How would you allocate an interactive node when you are in a screen session for example?

There are two ways. You either allocate the interactive node,

ssh

to that from outside the

screen

session and start

screen

in there, or you

ssh

to the node you get from inside your currently running

screen

session. I find the latter more practical.

Thanks!

In either case, after you're done and have logged out of the compute node you got, make sure the Slurm job for that is cancelled so it's not eating cpu hours. If it doesn't terminate by itself when you log out, just

scancel

it.