You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Mar 16, 2022. It is now read-only.
I had run "fc_run.py fc_run_ecoli_local.cfg"on the one dataset and some error occured.
[ERROR]Task Node(0-rawreads) failed with exit-code=256
[INFO]recently_satisfied: set([])
[INFO]Num satisfied in this iteration: 0
[INFO]Num still unsatisfied: 1
[ERROR]Some tasks are recently_done but not satisfied: set([Node(0-rawreads)])
[ERROR]ready: set([])
submitted: set([])
Traceback (most recent call last):
File "/falcon/FALCON-integrate/FALCON/src/py_scripts/fc_run.py", line 5, in <module>
main(sys.argv)
File "falcon/FALCON-integrate/FALCON/falcon_kit/mains/run1.py", line 504, in main
main1(argv[0], args.config, args.logger)
File "falcon/FALCON-integrate/FALCON/falcon_kit/mains/run1.py", line 150, in main1
input_fofn_plf=input_fofn_plf,
File "falcon/FALCON-integrate/FALCON/falcon_kit/mains/run1.py", line 216, in run
wf.refreshTargets([rdb_build_done])
File "falcon/FALCON-integrate/pypeFLOW/pypeflow/simple_pwatcher_bridge.py", line 273, in refreshTargets
self._refreshTargets(updateFreq, exitOnFailure)
File "falcon/FALCON-integrate/pypeFLOW/pypeflow/simple_pwatcher_bridge.py", line 339, in _refreshTargets
raise Exception(msg)
Exception: Some tasks are recently_done but not satisfied: set([Node(0-rawreads)])
fc_run_ecoli_local.cfg are as following:
> [General]
> job_type = local
>
> # list of fasta files
> input_fofn = falcon/FALCON-integrate/FALCON-examples/e.coli/input.fofn
>
> # input type, raw or pre-assembled reads (preads, error corrected reads)
> input_type = raw
> #input_type = preads
>
> # The length cutoff used for seed reads used for error correction.
> # "-1" indicates FALCON should calculate the cutoff using
> # the user-defined genome length and coverage cut off
> # otherwise, user can specify length cut off in bp (e.g. 2000)
> length_cutoff = -1
> genome_size = 4641652
> seed_coverage = 20
>
> # The length cutoff used for overalpping the preassembled reads
> length_cutoff_pr = 12000
>
> ## resource usage ##
> # need this line even if running local
> jobqueue = your_queue
> # grid settings for...
> # daligner step of raw reads
> sge_option_da = -pe smp 5 -q %(jobqueue)s
> # las-merging of raw reads
> sge_option_la = -pe smp 20 -q %(jobqueue)s
> # consensus calling for preads
> sge_option_cns = -pe smp 12 -q %(jobqueue)s
> # daligner on preads
> sge_option_pda = -pe smp 6 -q %(jobqueue)s
> # las-merging on preads
> sge_option_pla = -pe smp 16 -q %(jobqueue)s
> # final overlap/assembly
> sge_option_fc = -pe smp 24 -q %(jobqueue)s
>
> # job concurrency settings for...
> # all jobs
> default_concurrent_jobs = 30
> # preassembly
> pa_concurrent_jobs = 30
> # consensus calling of preads
> cns_concurrent_jobs = 30
> # overlap detection
> ovlp_concurrent_jobs = 30
>
> # daligner parameter options for...
> # https://dazzlerblog.wordpress.com/command-guides/daligner-command-reference-guide/
> # initial overlap of raw reads
> pa_HPCdaligner_option = -v -B4 -t16 -e.70 -l1000 -s1000
> # overlap of preads
> ovlp_HPCdaligner_option = -v -B4 -t32 -h60 -e.96 -l500 -s1000
>
> # parameters for creation of dazzler database of...
> # https://dazzlerblog.wordpress.com/command-guides/dazz_db-command-guide/
> # raw reads
> pa_DBsplit_option = -x500 -s50
> # preads
> ovlp_DBsplit_option = -x500 -s50
>
> # settings for consensus calling for preads
> falcon_sense_option = --output_multi --min_idt 0.70 --min_cov 4 --max_n_read 200 --n_core 6
>
> # setting for filtering of final overlap of preads
> overlap_filtering_setting = --max_diff 100 --max_cov 100 --min_cov 20 --bestn 10 --n_core 24
>
The text was updated successfully, but these errors were encountered:
I had run "fc_run.py fc_run_ecoli_local.cfg"on the one dataset and some error occured.
The text was updated successfully, but these errors were encountered: