Preliminary backup implementation using rclone

[craig-willis]

Along with repo https://github.com/whole-tale/backup This PR fixes https://github.com/whole-tale/terraform_deployment/issues/11

Changes:

• Adds init-backup.sh script to create systemd service and timer for daily backups on fileserver
• Removes restore_url capability. Restore now requires manual execution from within the backup container
• Changes wt_mongo network to be attachable

Test case:

• Precondition: Assumes new deployment (I've been testing on Nebula in NDS project)
• Configure rclone, following instructions in README.md to configure for Box. (This is somewhat cumbersome, but so far the easiest way):

  cd terraform_deployment
  rclone --config rclone.conf config

  • Configure variables.tf as per usual, then run terraform apply
• ssh <fileserver> and confirm /etc/systemd/system/backup.service and backup.timer are present
• Restore from a previous dev backup then use the init-mongo.sh to replace the domain and globus credentials. This step requires restarting Girder.

  $ docker run --rm --network wt_mongo -v /mnt/home:/backup -v /home/core/rclone/:/conf wholetale/backup bash
  # restore.sh -c dev -d 20180416
  # init-mongo.sh <domain> <globus_client_id> <globus_client_secret>

• Confirm restore worked by logging into your instance
• Run a backup. Either wait for cron task to execute or manually run command docker run --rm --network wt_mongo -v /mnt/home:/backup -v /home/core/rclone/:/conf wholetale/backup backup.sh -c cluster_id where cluster_id is a unique name for your cluster. This will create in Box WT// containing tgz files for both home directories and mongodb.
• Confirm files are present via rclone (in backup container)

  rclone --config /conf/rclone.conf ls backup:WT/<cluster_id>/<date>

By default, it will only retain backups for 7 days.