Skip to content

Latest commit

 

History

History
201 lines (122 loc) · 7.09 KB

README.md

File metadata and controls

201 lines (122 loc) · 7.09 KB
Raw

cf-backup-manager

Cloud Foundry application to automate backup and restore of application back end data services.

See GSA/datagov-deploy#2768 for design history.

Usage

The backup-manager runs as a Cloud Foundry application with zero instances. Most backup/restore commands should be run via Cloud Foundry task.

$ cf run-task backup-manager --wait --name dashboard-restore --command 'restore mysql dashboard-db /backup-manager-v1/development/dashboard-db/dashboard-db-20211201022504-backup.gz'

Inspect the currently running tasks.

$ cf tasks backup-manager

Monitor the backup-manager logs during task run.

$ cf logs backup-manager | grep APP/TASK/dashboard-restore

If you need to SSH into the container, make sure to set the path so that the commands are available to you.

$ PATH=/usr/local/bin:$PATH

Backup and restore procedure

We recommend you run a backup regularly via a scheduled CI task.

$ cf run-task backup-manager --wait --name "backup" --command "backup mysql my-service"

$ cf run-task backup-manager  --name "backup" --command "backup psql my-service <backup_path>"

To restore, you should create a new service to ensure you're restoring to a clean state rather than restore into an existing service.

$ cf create-service mysql micro-mysql my-service-new
$ cf run-task backup-manager --wait --name "restore" --command "restore mysql my-service-new /path/to/mysql-backup.gz"

Then rename the old and new service.

$ cf rename-service my-service my-service-venerable
$ cf rename-service my-service-new my-service

Restart your application.

$ cf restart my-app

Supported services

  • mysql

Commands

list [-r] [path]

List the contents of the backup-manager-s3 bucket. Specify -r to list the path recursively.

$ cf run-task backup-manager --name "<run-name>" --command "list /"
$ cf logs backup-manager --recent

backup ['<db_flags>'] <service_type> <service_name> [backup_path]

Create a backup for the named service. You must specify the service type e.g. mysql. If you don't provide a backup path, then one will be generated in the form:

/backup-manager-v1/$space/$service_name/$service_name-$datetime-backup.gz

restore ['<db_flags>'] <service_type> <service_name> <backup_path>

Restore the named backup to the specified service. In most cases, you should restore to a new service instead of restoring to an existing service and then rename the new service to replace the old one.

Migrations

You can also use this application to migrate databases from outside of Cloud Foundry.

  1. Get a service key for the backup-manager-s3 service
  2. Use these credentials to export and upload your backup to the backup-manager-s3
  3. Restore the backup to your Cloud Foundry service

First, create a service key for the backup-manager-s3 service.

$ cf create-service-key backup-manager-s3 fcs-migration
$ cf service-key backup-manager-s3 fcs-migration

Use these credentials with the AWS CLI to backup to the backup-manager-s3 bucket. You can create a secure tmp file and set these variables:

# create this file with mktemp
# /tmp/tmp.abcef
export AWS_ACCESS_KEY_ID=
export AWS_SECRET_ACCESS_KEY=
export BUCKET_NAME=
export AWS_DEFAULT_REGION=us-gov-west-1

Here's an example using mysqldump to export and upload to our backup-manager-s3 bucket.

$ source /tmp/tmp.abcdef
$ time mysqldump --host "$DB_HOST" --password="$DB_PASSWORD" --user "$DB_USER" --no-create-db --verbose "$DB_NAME" | gzip | aws s3 cp - s3://${BUCKET_NAME}/migrations/mysql-dashboard-migration.sql.gz

Now that your backup is in the backup-manager-s3 bucket, you can restore it to your service.

$ cf run-task backup-manager --wait --name mysql-migration --command 'restore mysql dashboard-db /migrations/mysql-dashboard-migration.sql.gz'

Once complete, you should have your database migrated to Cloud Foundry.

Setup

management space diagram showing backup-manager's interaction with data services

  1. Create the backup-manager-s3 service where backups will be stored
  2. Share the backup-manager-s3 service with each space you want to backup
  3. Push the backup-manager app to each space you want to backup
  4. Bind the backup-manager-s3 service
  5. Bind any target services to be backed up

First, create the S3 service in your management space.

$ cf target -s management
$ cf create-service s3 basic backup-manager-s3

Share the S3 service with each space.

$ cf share-service backup-manager-s3 -s production

Push the application to each space that you want to backup. This helps to avoid conflicts with services that might not have unique names across spaces.

$ cf target -s production
$ cf push backup-manager -f manifest.yml --vars-file vars.production.yml

Bind the backup-manager-s3 service to the backup-manager app in each space.

$ cf target -s production
$ cf bind-service backup-manager backup-manager-s3

Bind any target services you want to backup to the backup-manager app.

$ cf target -s production
$ cf bind-service backup-manager dashboard-db

You're ready to make some backups!

Implementing additional services

Each service must implement this API:

  • service_get_credentials_env: parses the service credentials from VCAP_SERVICES and echos them to stdout to be evaluated by the shell.
  • service_backup: runs commands to backup the service to stdout.
  • service_restore: runs commands to restore the service from stdin.
Function name arguments stdin stdout
service_get_credentials_env 1: name of the service instance N/A Credentials in environment variable (env=val) form. These variables are consumed by service_backup and service_restore.
service_backup N/A N/A Pipes service backup data to stdout
service_restore N/A Restores service instance from data read from stdin N/A

Publishing the image

The Cloud Foundry application uses a docker image published to this repository. New images are published on any push to main via GitHub Actions.

Tests

The bats test framework is used to perform unit testing of individual backup/restore aspects. The tests are located in the test/ subdirectory. To run the tests,

git submodule init
git submodule update
make test

Contributing

See CONTRIBUTING for additional information.

Public domain

This project is in the worldwide public domain. As stated in CONTRIBUTING:

This project is in the public domain within the United States, and copyright and related rights in the work worldwide are waived through the CC0 1.0 Universal public domain dedication.

All contributions to this project will be released under the CC0 dedication. By submitting a pull request, you are agreeing to comply with this waiver of copyright interest.