Skip to content

oicr-gsi/dashi

Repository files navigation

Dashi

Dashi is a quality control reporting system developed for genomic and transcriptomic data. It is built on top of gsi-qc-etl, OICR's private QC data integration system, and so is currently of limited utility to external parties. Please contact us for more information.

Requirements

Environment Variables

Create a .env file in the root directory of this repository:

Variable name Required? Description Example Default
GSI_QC_ETL_ROOT_DIRECTORY Yes One or more colon seperated directories where the QC-ETL caches are located. Records will be deduplicated, with records kept from the cache listed first /qcetl or /qcetl:/qcetl_archive
MONGO_URL This or MONGO_FILE URL to location of MongoDB which holds Pinery data mongodb://user:password@mongo_web_url:27017/db_name
MONGO_FILE This or MONGO_URL File location of hd5 file holding DataFrame dump /mongo_provenance.hd5
PINERY_URL Yes URL to location of Pinery web service root http://pinery-url:8080/pinery-ws-miso
MISO_URL Yes URL to location of MISO web service root http://miso.your.domain/
LOG_FILE_LOCATION Yes File path where logs should be written ~/logs/dashi.log ./dashi.log
BARCODES_STREXPAND Yes Tab-separated file listing 10X barcodes and 4 sequences for each ~/barcodes
LOG_TO_CONSOLE No Set to log to console as well as to log file specified above True do not log
USE_BLEEDING_EDGE_ETL No Set to install gsi-qc-etl@master instead of the release version of gsi-qc-etl in requirements.txt (Docker only) 1 use release version
EXCLUDE_SWAP_LIBS No File path to TSV file of library pairs to be excluded for swap view ./exclude_swap_lib.tsv
SAMPLES_FOR_PROJECTS No Indicate whether samples from ALL projects should be used, or only samples from ACTIVE projects. ALL ACTIVE
DISPLAY_USER_MESSAGE No A JSON file containing a dictionary of page names (key) and messages to display (value) ./user_messages.json

Setup on bare metal

  1. Install python3, pip.
  2. Setup new virtual environment.
  3. pip install -r requirements.txt.
  4. Ensure your .env file is populated as per Environment Variables above.
  5. flask run OR gunicorn --bind 0.0.0.0:5000 wsgi:app

Set up Docker container

The Docker container is a more straightforward way to launch Dashi for testing. The base container has most of the dependencies required for the app, with one crucial exception: gsi-qc-etl. This repository is private to OICR and so you need to pass in your SSH keys to permit download and installation.

Requirements

  • Docker 18.09.6+
  • Docker compose 1.23.1+
  1. Ensure your ssh key has been added to OICR's Bitbucket and you can access and clone gsi-qc-etl.
  2. Download the gsi-qc-etl cache data to $HOME/qcetl (or modify docker-compose.yml to point to the correct location). The current location for this on OICR's cluster is at /scratch2/groups/gsi/<development or production>/qcetl.
  3. Create a file at .mongopass with the password to the MongoDB database and make sure the location in docker-compose.yml is correct in secrets.
  4. Ensure your .env file is populated as per Environment Variables above.
  5. Ensure the file 'dashi.log' exists in the place specified by LOG_FILE_LOCATION. Create a blank file if necessary.
  6. Build the container with docker-compose build.
  7. Launch with docker-compose up. Note that this completes installation of gsi-qc-etl before launching the app. `

Then navigate to http://0.0.0.0:5000/.

Troubleshooting

1. docker-compose up fails with git@bitbucket.oicr.on.ca: Permission denied (publickey). fatal: Could not read from remote repository. Please make sure you have the correct access rights and the repository exists.

This is likely due to an error with binding your SSH key into the container. The SSH key is required to install the gsi-qc-etl dependency. Follow these steps for troubleshooting:

  1. Try to clone gsi-qc-etl: git clone ssh://git@bitbucket.oicr.on.ca/gsi/gsi-qc-etl.git. If this fails, check that you have correct permissions to the repository https://bitbucket.oicr.on.ca/projects/GSI/repos/gsi-qc-etl/browse
  2. Check that your SSH key is located in ~/.ssh/. If it is not, change the bind paths in docker-compose.yml for id_rsa, id_rsa.pub, and known_hosts.
  3. Check if your SSH key requires a passphrase. If it is prompting from inside the container, it will fail. Create a private file with your SSH key passphrase. Open docker-compose.yml and uncomment out the ssh_passphrase lines in the dashi service and in the secrets section, and point docker-compose to your passphrase file in the secrets section. Do a fresh build of the compose file, ie. docker-compose build --no-cache. When next launching the container, the start.sh script will automatically provide the passphrase upon request.

2. docker-compose up fails with ModuleNotFoundError: No module named 'gsiqcetl'

Likely gsi-qc-etl failed to download. Check the Troubleshooting tip #1.

**3. docker-compose up is not starting and reporting messages like : dashi | [2019-12-04 22:08:06 +0000] [54] [CRITICAL] WORKER TIMEOUT (pid:113) dashi | [2019-12-04 22:08:06 +0000] [113] [INFO] Worker exiting (pid: 113) dashi | [2019-12-04 22:08:07 +0000] [120] [INFO] Booting worker with pid: 120

Gunicorn and Docker dislike each other. Try using flask run instead.