Skip to content

Latest commit

 

History

History
73 lines (45 loc) · 3.73 KB

README.md

File metadata and controls

73 lines (45 loc) · 3.73 KB
Raw

SOLVCON in Google Compute Engine

This repository sets up the bash environment for running SOLVCON in Google Compute Engine (GCE). It provides tools for creating, provisioning, and accessing virtual-machine (VM) intances in GCE. Clone the repository to install:

$ git clone http://github.com/solvcon/solvcon-gce ~/opt/gce

To enable it, run source ~/opt/gce/etc/gcerc. Append the following line in .bashrc to enable it automatically:

if [ -f ~/opt/gce/etc/gcerc ]; then source ~/opt/gce/etc/gcerc; fi

Set up Google Cloud Platform SDK

solvcon-gce are essentially wrappers to Google Cloud Platform (GCP) SDK command-line tool. solvcon-gce will try to load the SDK from ~/opt/google-cloud-sdk. If it's not there and not in PATH, you can install it (to the assumed path) using the following script:

~/opt/gce/bin/admin/install-google-cloud-sdk.sh

Setup a Project

Before using solvcon-gce scripts, you need to sign up the GCE service and create a project, and initialize the cloud SDK by running gcloud init. See https://cloud.google.com/sdk/docs/. After a project is created, you also need to do the following setup in the Google Compute Platform Console:

  1. Enable "Compute Engine API".
  2. Add a project-wide SSH key. Accounts logged into the GCE instance using a project-side SSH key can run sudo in the instance.

If somehow the project ID isn't set up properly by gcloud init, do it by running gcloud config set project <project_id>.

(GCP offers a 60-day free-trial program, including $300 credits: https://cloud.google.com/free-trial/ .)

Cache Conda Packages

To save time from downloading conda packages from the Anaconda server, solvcon-gce needs to cache them in a Google Cloud Storage bucket. Before the cache is in place, Anaconda won't be available in the instance, and when starting the instance, solvcon-gce scripts would complain:

bash: /var/lib/conda/packages//Miniconda3-latest-Linux-x86_64.sh: No such file or directory
~/opt/gce/bin/admin/install-conda.sh: line 12: conda: command not found
bash: /var/lib/conda/packages//Miniconda2-latest-Linux-x86_64.sh: No such file or directory
~/opt/gce/bin/admin/install-conda.sh: line 18: conda: command not found

To populate the cache, run:

$ gce-prepare-conda-packages

The script will create a bucket based on the project ID (gs://<project_id>-conda-packages/). Before executing the above command, you need to create the bucket yourself. Please note that the bucket should be created in the same zone that the solvcon-gce tools assume, otherwise additional charges may incur. For now it is asia-east1.

You can use Google Compute Platform Console or the following command to create the bucket:

$ gsutil mb -c standard -l asia-east1 gs://<bucket-name>

Running gce-prepare-conda-packages in GCE would make uploading to the bucket much faster.

Use an Instance

Run gstart <instance_name> to create a GCE VM instance. It usually takes 2 minutes. gstart also runs the provisioning scripts.

After gstart finishes, run gssh <instance_name> to connect to the instance.

To remove the instance (and stops being charged), run gce-delete-instance <instance_name>.

Trouble Shooting

SSH Connection Refused

If your SSH connectioned is refused after issuing gstart, please make sure you have unlocked your key phrase of the project-wide SSH key.