Skip to content

int-brain-lab/ibl-aws

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ibl-aws

Utilities to run processing on AWS Cloud

cd ~/Documents/PYTHON/ibl-sorter/docker
sudo docker compose exec spikesorter /bin/bash
sudo docker compose exec spikesorter python /root/Documents/PYTHON/ibl-sorter/examples/run_ibl_recording.py 7ae3865a-d8f4-4b73-938e-ddaec33f8bc6 probe01 --cache_dir /mnt/s0 --scratch_dir /mnt/s0/scratch

Installing docker with Nvidia support from a blank EC2

  1. Install nvidia drivers
  2. Install docker with GPU support
  3. Build the docker image and install the container
  4. Mount elastic volume that will hold the spike sorting results
  5. Whitelist security group
  6. Setup ONE

Select an EC2 instance type with Ubuntu on it (P2 or GP3). Needs to have one system volume of 16 Gb (for the docker build) and an another volume of 800 Gb

  1. Install the nvidia drivers (~3mins) - maybe this can be skipped using a EC2 pytorch image
sudo apt-get update
sudo apt install -y ubuntu-drivers-common
sudo ubuntu-drivers install
sudo reboot

It will take a couple of minutes before the instance is available to login again

  1. Install Docker with GPU support using the bootstrap script provided (~2mins)
mkdir -p ~/Documents/PYTHON
cd ~/Documents/PYTHON
git clone -b aws https://github.com/int-brain-lab/ibl-sorter.git
cd ibl-sorter/docker
sudo ./setup_nvidia_container_toolkit.sh
  1. Format and mount the attached volume, here you want to check that the volume is indeed /dev/xvdb using df -h (few secs):
sudo mkfs -t xfs /dev/xvdb
sudo mkdir -p /mnt/s0
sudo mount /dev/xvdb /mnt/s0
sudo chown ubuntu:ubuntu -fR /mnt/s0
df -h
  1. Build the docker image referring to the instructions above and start the container

TODO link to ibl-sorter docker readme file

  1. Setup ONE from inside the container, make sure the cache directory is /mnt/s0/spikesorting, configure the base URL according to your needs, for internal re-runs it should be set to https://alyx.internationalbrainlab.org Here is the command to enter the container shell:
sudo docker compose exec spikesorter /bin/bash
  1. If you want to send the data to flatiron you'll have to setup the ~/.ssh/config file so as to reflect the sdsc SSH configuration.

About

Utilities to run processing on AWS Cloud

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages