Running Manta in Docker

[lslochov]

Hi, I'm running SV calling in the Google Cloud Platform, using a Dockerized version of Manta I encapsulated in a CentOS image. I have run the Manta sample workflow in this Docker image on my local machine without any issues, but when I try to run an analysis in the Google Cloud, I can run the configure step, but running "runWorkflow.py" results in the following error:

Traceback (most recent call last): File "/mnt/data/input/gs/simons_bams/bams/manta_run/runWorkflow.py", line 181, in main(r"/mnt/data/output/gs/simons_bams/bams/manta_run/runWorkflow.py.config.pickle","manta",MantaWorkflow) File "/mnt/data/input/gs/simons_bams/bams/manta_run/runWorkflow.py", line 144, in main warningpath=os.path.join(flowOptions.runDir,"workflow.warning.log.txt") AttributeError: 'WorkflowOptions' object has no attribute 'runDir'

The Google Cloud works by taking data in storage buckets, and mounting them as external filesystems in Docker containers. So the sample workflow, being part of the Manta distribution, is part of the image's internal filesystem, but the cloud data is imported from outside the image—this might be affecting the workflow's configuration. Any help you can provide to get this running would be most appreciated!

[x-chen]

We haven't tried to run Manta on Google Cloud. If we heard any successful experience, will update this thread. Sorry for not being very helpful.

[ctsa]

Sounds like it might be having trouble unpickling the configuration file...

Do you a file named runWorkflow.py.config.pickle in the same directory as runworkflow.py? If so, does it contain a runDir key? Note this should be protocol 0 pickle file so a simple text search should work.

[lslochov]

Hi, yes there is a runWorkflow.py.config.pickle in the directory where I configured my Manta workflow, and it does have a runDir key.

[ctsa]

Thanks, in that case I see that on the config file read in here:

https://github.com/Illumina/manta/blob/master/src/python/lib/configureUtil.py#L71-L80

We've made a (probably) bad design choice of returning an empty hash instead of issuing an error if os.path.isfile fails for some reason on the pickle file.

Somewhat less likely is the primarySection ('manta') retrieval experiencing an error and returning the default config here:

https://github.com/Illumina/manta/blob/master/src/python/lib/configureUtil.py#L90

Some value printing around these locations might reveal more about what's going on. Your reported issue is at the very first point where flowOptions is being used after deserializing these options from the pickle file, so it seems almost certain that one of those points in the deserializing process is experiencing some unexpected behavior specific to GCP.

---

EDIT: Fixed both github links.

[WKabza]

Is this issue have been solved? I have similar issue but with running Manta using Docker on ssh