-
Notifications
You must be signed in to change notification settings - Fork 4
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Error at first stage - Pipeline is not running #70
Comments
Thought I'd removed that problem - oh well...
|
Hi
The pipeline started to run, however, I got another error message: Starting pipeline...
|
WHat version of rubra did you download, and what cluster system are you using? |
It is rubra version 0.1.5 and I'm using SCG Informatics Cluster https://login.scg.stanford.edu/ |
Try the version in the branch named 'Slurm' - think this is the problem. |
Sorry, do you mean I have to ask them to install another version of rubra? |
Yes - the other version is for a qsub (torque) system, not SLurm |
Hi We installed Rubra-0.1.5 from the slurm branch as you suggested and the pipeline started to run, however, we got another error message: `Starting pipeline... Original exception:
No way to run job: Input file '/labs/jandr/Kesia/Asia/stdy5352/Nepal/SEAP/Red_out/temp/AL513382_1.fasta' does not exist
Any idea what the problem might be here? Kesia |
If you check the log folder, there are standard out and error files for each job - can you check to see if there is a particular error message from the makeDir step (this is the one that failed). BTW you probably want to get rid of the #'s from your sequence names - that character (and a few other special one's) can lead to errors in many bioinformatic programs (some handle them fine...). I usually replace it with an underscore. (That is not the problem here...) Just thought of something else really important - did you change the modules in the config file to those for your local machine? stageDefaults = { You will need to change these for distributed (queuing) installation You need to change the ones in bold for your local installation... |
I have changed the modules in the config file for the local installation and I also removed the #'s from sequence names but I'm still getting the same error message. Here is the message from the makeDir step file: #!/bin/bash srun --error=log/makeDir.%j.stderr --output=log/makeDir.%j.stdout --mem=4096 --job-name=makeDir --time=00:10:00 bash /oak/stanford/scg/lab_jandr/Kesia/Asia/stdy5352/Nepal/SEAP/RedDog-master/tmpM3C0B9 |
the srun command sets the output files for for standard out (stdout) and standard error (stderr) with the following... --error=log/makeDir.%j.stderr --output=log/makeDir.%j.stdout so is there anything reported in either of those two files? (the latest ones as you have run the pipe a few times...) |
These files were not created. The only files that were created in the log folder were: makeDir.sh |
try running makDir.sh via batch outside the pieline and see what happens |
Hi I had to add the following information to makeDir.sh file to run via batch outside the pipeline: #!/bin/bash Then I got this message: python: can't open file 'makeDir.py': [Errno 2] No such file or directory |
Can you send me your config file - if you don't want to post it here, you can send it directly to me via David.Edwards @ monash.edu (remove the spaces) |
Apologies for the delay in responding - just got feedback on a paper very close to submission and have had to repeat some of the analysis rather quickly.... There is a second way to run redog on a distributed system as a single job - albeit with lots of processors to get the job done. Means you can't really monitor progress directly, but it also means the job only has to enter the job queue once (instead of sending lots of jobs to the queue), so can be more efficient on a crowded system. have a look at the config file we currently use: The changes needed to run the pipeline like this are:
set and 3. just below Then launch the job as a single job, giving the pipeline a few hours for smaller runs - with bigger runs that can take longer than our seven day limit, I run the pipeline until it goes down, then just restart it - reddog then starts again from where it was interrupted... You will have to add all the module loading relevant for your environment, but our batch job script is as follows:
You may not need all these setting (partition is a local variable, as is qos and account which you may or may not need...) |
Hi David, Thank you so much for your email. I changed my config file following your suggestions and I could run the pipeline in two samples to test and worked. However, when I tried to run again with more samples the pipeline started and them failed. Now I'm trying to launch the job again and I'm getting this message Traceback (most recent call last): |
Hi David, I download all the RedDog files again and I used the https://github.com/katholt/RedDog/blob/master/RedDog_config_massive.py as the config file.... this time the pipeline started but failed at check bam stage. Here is the message in the log folder: |
Hi David,
Do you have any idea on what the problem is now? Best |
Hi David, I'm trying to run RedDog is a very large dataset >2000 genomes. Increase the walltime and memory to "walltime": "06:00:00" and "memInGB": 8, would be enough for the run? |
Hi Kesia, Sorry, paper out of way (for now) so can get on to this... You shouldn't need to up the mem to 8GB unless any of your read sets are particularly large - if I get a fail at the mapping (BAM) step, I often rerun just that stage with extra memory (see the manual/wiki for how to stop the pipeline after a particular step..). There is one other step that may require more than 4Gb latte rin the pipeline, but that also depends on the number of SNPs you need to process... (which you often don't know until you have run reddog...). So if it fails later when collating the SNP matrix (or indeed parsing the matrix), you may want to up the GB just for rerunning the failed stage(s) - the pipe finishes not long after that (by default you won't get a tree for that many isolates...), so you could runn it to the end with the higher GB, if required (hope that made sense). As for time, that does depend on the number of cores you can give the pipe - with 20 processors and ~2000 isolates I would expect reddog to take less than a day, but does depend on the read set sizes, number of SNPs etc. So set it for 24 hours (1 day), if using 20 procs (if you are using more [lucky you - I can only get 24 max for a single job] then you will get away with less time...). Remember if the pipe runs out of time before completing, you can just restart it and it will complete what is left. good luck! David |
Hi,
We are having trouble getting Red dog to run on our cluster. The pipeline does not start running and we keep getting the error below.
Thanks,
Kesia
Traceback (most recent call last):
File "/scg/apps/software/reddog/v1b11/bin/rubra", line 10, in
sys.exit(main())
File "/scg/apps/software/reddog/v1b11/lib/python2.7/site-packages/rubra/rubra.py", line 30, in main
options = getOptions(args)
File "/scg/apps/software/reddog/v1b11/lib/python2.7/site-packages/rubra/utils.py", line 210, in getOptions
imported = import(module)
File "RedDog_config_run.py", line 34
SyntaxError: Non-ASCII character '\xe2' in file RedDog_config_run.py on line 35, but no encoding declared; see http://python.org/dev/peps/pep-0263/ for details
The text was updated successfully, but these errors were encountered: