Skip to content
Branch: master
Find file Copy path
Find file Copy path
1 contributor

Users who have contributed to this file

99 lines (67 sloc) 4.54 KB

AWS Marketplace Administration

The Graphistry marketplace instance is designed for secure and purely web-based use. However, command-line administration can be helpful. This document shares common marketplace tasks. See the main docs for general CLI use.


  1. Recommended configuration
  2. Solve GPU availability errors
  3. Log in
  4. Docker
  5. Install Python packages
  6. Install native packages

1. Recommended configuration

2. Solve GPU availability errors

Upon trying to launch, Amazon may fail with an error about no available GPUs for two reasons:

  • Lack of GPU availability in the current region. In this case, try another valid GPU type, or launching in another region. For example, Virginia => Oregon. Keeping the GPU close to your users is a good idea to minimize latency.

  • Insufficient account quota. In this case, the error should also contain a link to increase your quota. Request p3.2 (and above), and 1-2 for a primary region and 1-2 for a secondary region.

3. Log in

Log in using the key configured at AWS instance start and your instance's public IP/domain:

ssh -i my_key.pem ubuntu@MY_PUBLIC_IP_HERE

Many ssh clients may require you to first run chmod 400 my_key.pem or chmod 644 my_key.pem before running the above.

4. Docker

Graphistry leverages docker-compose and nvidia-docker2.

cd ~/graphistry
docker-compose ps


               Name                             Command                  State                        Ports                  
graphistry_celerybeat_1              /entrypoint bash /start-ce ...   Up             8080/tcp                                
graphistry_celeryworker_1            /entrypoint bash /start-ce ...   Up             8080/tcp                                
graphistry_forge-etl_1               /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_nexus_1                   /entrypoint /bin/sh -c bas ...   Up             8080/tcp                                
graphistry_nginx_1                   nginx -g daemon off;             Up   >443/tcp,>80/tcp
graphistry_notebook_1                /bin/sh -c graphistry_api_ ...   Up             8080/tcp                                
graphistry_postgres_1       postgres    Up             5432/tcp, 8080/tcp                      
graphistry_redis_1          redis ...   Up             6379/tcp, 8080/tcp                      
graphistry_streamgl-datasets_1       /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-gpu_1            /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-sessions_1       /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-svg-snapshot_1   /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-vgraph-etl_1     /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-viz_1            /tini -- /entrypoints/stre ...   Up             8080/tcp   

5. Install Python packages

If you see wheel errors, you may need to run pip3 install wheel and restart your Jupyter kernel.

6. Install native packages

By default, Jupyter users do not have sudo, restricting them to user-level installation like pip. For system-level actions, such as for installing golang and other tools, you can create interactive root user sessions in the Jupyter Docker container:


Note that sudo is unnecessary:

ubuntu@ip-172-31-0-38:~/graphistry$ docker exec -it -u root graphistry_notebook_1 bash
root@d4afa8b7ced5:/home/graphistry# apt update 
root@d4afa8b7ced5:/home/graphistry# apt install golang


ubuntu@ip-172-31-0-38:~/graphistry$ docker exec -it  graphistry_notebook_1 bash
graphistry@d4afa8b7ced5:~$ go version


go version go1.10.4 linux/amd64

See general installation for further information.

You can’t perform that action at this time.