Skip to content

Latest commit

 

History

History
90 lines (51 loc) · 2.33 KB

README.md

File metadata and controls

90 lines (51 loc) · 2.33 KB
Raw

One-time setup

Python

On Sherlock, run pyenv local dvv_seq in your cloned directory

bash_profile

Add the following to your $HOME/.bash_profile (using the appropriate path to your cloned directory):

export PYTHONPATH="/path/to/SingleCellSequencing:$PYTHONPATH"

To import the necessary shared libraries, you will need to execute the following each time after you log in to Sherlock (alternatively, you can add it as a line to your $HOME/.bash_profile):

module load dvv_seq

MongoDB

  • Create an account at mongolab.com

  • Sign in and get to the home screen

  • Next to "MongoDB Deployments" you'll see three buttons. Click the one that says "Create new".

  • For "Cloud provider", select "amazon web services". For "Location": "Amazon's US East (Virginia) Region (us-east-1)".

  • Under "Plan", choose "Single-node" and select "Sandbox" (...it's free).

  • For "Database name" write "single_cell_sequencing".

  • Click "Create new MongoDB deployment".

  • Back on the home screen, click on "single_cell_sequencing".

  • Click on the "Users" tab and then select "Add a database user".

  • Choose a username and password. Note that in fireworks, the password is stored in plaintext.

  • Note that the information shown at the top ("To connect using the shell") contains the database hostname and database port (the part after the colon is the port).

Config files for Fireworks

  • Run python initialize.py

  • Run lpad reset and choose Yes if prompted

Usage

Queue

To queue a set of directories to process, run:

python fw_queue.py /path/to/directory/containing/libraries

To see a list of help options, run:

python fw_queue.py --help

Run interactively

To run fireworks interactively (after having queued your jobs), run:

rlaunch rapidfire

This is useful for debugging, but you probably don't want to use it in production because it runs serially.

Do not do this on a login node (your tasks are too computationally expensive)

Run using the SLURM Scheduler

To run using the scheduler:

qlaunch -r rapidfire --nlaunches infinite --sleep 5

This command will run forever until you Ctrl-C to kill it once you see that all the output and analysis files have been generated.

qlaunch is relatively lightweight, so you can probably get away with running it on a login node.