Install NCI gdc-client on RHEL 6 to download GDC data

how-to-gdc-client-install-rhel6.md

gdc-client on RHEL/CentOS 6

Linux 2.6.32-431.23.3.el6.x86_64

## Using conda 4.3.33

## remove unstable conda env
conda env remove --name rvlabpy2

## create gdc_client conda env with required dependencies
conda create --name gdc_client python=2.7 pip parcel pyopenssl cryptography libxml2 yaml

## activate gdc_client env
source activate gdc_client

## make accessible for all users
umask 0022

## clone GDC repository
cd /fastscratch/dump/gdc
git clone git@github.com:NCI-GDC/gdc-client.git
cd gdc-client

which python

/projects/verhaak-lab/verhaak_env/anaconda/v4.2.0/envs/gdc_client/bin/python

pwd
# fastscratch/dump/gdc/gdc-client

python setup.py install 2>&1 | tee -a install.log

• This will install gdc-client in your PATH

which gdc-client

gdc-client --help

• If you get an error like this during download, it needs a patch by replacing url with uri at ~ line 262 of gdc_client/download/client.py file. See https://github.com/NCI-GDC/gdc-client/issues/129#issuecomment-340134109

'DownloadStream' object has no attribute 'url'

        # gdc-client calls parcel's parallel_download,
        # which is where most of the downloading takes place
        file_id = stream.uri.split('/')[-1]
        super(GDCDownloadMixin, self).parallel_download(stream)

• Run setup again

cd fastscratch/dump/gdc/gdc-client
python setup.py install 2>&1 | tee -a install_patch1.log

## remove source
rm -rf fastscratch/dump/gdc/gdc-client

Test gdc-client download

OUTDIR=/fastscratch/dump/gdc/lgg
MANIFEST=/fastscratch/dump/gdc/lgg/controlled_manifest.tsv
TOKEN="${HOME}"/.myconfs/gdc_token.key
CMD=gdc-client

## single sample
# ${CMD} download --log-file="${OUTDIR}"/download.log -n 2 --token-file ${TOKEN} --dir ${OUTDIR} 6c849069-054f-42d6-94ef-194d8a4eb58a 

## batch download
${CMD} download --log-file="${OUTDIR}"/download.log -n 2 -t ${TOKEN} -d ${OUTDIR} -m ${MANIFEST}

Known issues

• There is an open issue of output directory containing full URL instead of UUID/out.bam.

END

thank you! used this today. the url line that needs changing was on line 303.