Best Practices

Best Practices

  • Spin up a VM instance corresponding to your needs.

  • Always start using virtual machine/instance by applying latest security updates

  • When the instance is terminated or unavailable, any data stored inside it is unavailable or lost. Upload data that you need to be persistent to a S3 bucket, or attach a volume and copy it there.

  • Pull data you need immediate access to into your VM’s local storage. If your workflow needs a large reference dataset, it might be better to store this dataset on a web or FTP server in Collaboratory, instead of storing it inside each instance.

  • Execute analysis, review result, delete any unnecessary local data.

  • Push results and code you wish to keep to the S3-compatible object storage.

  • Terminate your VM and, subsequently, the local storage.

  • When editing the security group associated with your virtual machines/instances, only allow access from trusted IP addresses.

  • If you DO have to allow SSH access to your publicly accessible instances from (entire Internet), make sure that only SSH key-based authentication is allowed by editing SSH's config file and restarting the ssh service afterwards:

    • PasswordAuthentication no

    • ChallengeResponseAuthentication no

  • Do not share SSH keys or other credentials.

  • Terminate/delete unused resources (instances, volumes, snapshots).

  • Do not snapshot instances containing credentials and then share them with other users.

  • Do not copy or create large data sets on a virtual machine that you plan to snapshot later, because it will increase the snapshot size and take longer to start instances from it later, as well as taking more storage space.

  • Clean up confidential data (e.g. ssh keys, any saved credentials, bash history) before taking a snapshot of an instance, in case you will want to share that instance with other users later.

  • If planning to later snapshot an instance, use the SMALLEST possible flavor type that will allow you to install your applications because you can not start an instance of a smaller size than the original size used when the snapshot was taken.

  • The snapshot only captures and stores the contents of the root disk (/dev/sda), so do not store data or install applications on attached volumes if you need them to be part of the snapshot.

  • Do not use docker containers or upload Openstack images from untrusted sources.

  • Do not install software from untrusted sources.

  • Associate a floating IP only with a single instance and SSH into it before using SSH to connect to your other instances. In this way, you limit your security exposure.

  • Make use of cloud utilities for automating provisioning steps. Cloud-init is a package pre-installed in most modern Linux distributions that allows (among other things) execution of a script upon the initial boot of the instance. More information about using cloud-init and useful examples are available at

International Cancer Genome Consortium

© 2016 Cancer Genome Collaboratory. All rights reserved.