Awesome

soba: backup hosted git repositories

• about
• configuration
• run using the binary
• run with Docker
• run on Synology NAS
• run on Kubernetes
• scheduling backups
• rotating backups
• notifications
• logging
• setting provider credentials
• additional options
• restoring backups

about

soba is tool for backing up private and public git repositories hosted on the most popular hosting providers. It generates a git bundle that stores a backup of each repository as a single file.

As unchanged git repositories create identical bundle files, new bundles will only be stored if changes to the repository have been made. This can be done by re-cloning each repository every time soba runs, or by comparing refs without cloning.

soba includes its own scheduler that triggers it to run every specified number of hours, or it can be run with other schedulers such as cron.

quick start

soba can run as a binary or using docker with the prebuilt image distributed with each release. For example, the following will create git bundles of all repositories in your GitHub user's account in the soba-backups directory:

$ mkdir soba-backups
$ docker run --rm -v ./soba-backups:/backups -e GITHUB_TOKEN=<token-here> -e GIT_BACKUP_DIR=/backups jonhadfield/soba:latest

latest updates

1.2.20 release 2024-10-08

• Add Telegram notifications

1.2.19 release 2024-09-02

• Let user define the request timeout

1.2.18 release 2024-08-24

• Increase clone timeout to allow for larger repos and slower connections

1.2.17 release 2024-07-29

• Performance improvement

1.2.16 release 2024-06-09

• Add Slack notifications

See full changelog here.

supported OSes

Tested on Windows 10, MacOS, and Linux (amd64). Not tested, but should also work on builds for: Linux (386, arm386 and arm64), FreeBSD, NetBSD, and OpenBSD.

supported providers

• Azure DevOps
• BitBucket
• Gitea
• GitHub
• GitLab

configuration

soba can be run from the command line or as a container. In both cases the only configuration required is an environment variable with the directory in which to create backups, and additional to define credentials for each the providers.

On Windows 10:

• search for 'environment variables' and choose 'Edit environment variables for your account'
• choose 'New...' under the top pane and enter the name/key and value for each of the settings

On Linux and MacOS you would set these using:

export GIT_BACKUP_DIR="/repo-backups/"

To set provider credentials see below.

run using the binary

Download the latest release here and then install:

install <soba binary> /usr/local/bin/soba

After setting GIT_BACKUP_DIR, set your provider token(s) as detailed here.

and then run:

soba

run with Docker

Using Docker enables you to run soba without anything else installed.

Docker requires you pass environment variables to the container using the '-e' option and that you mount your preferred backup directory. For example:

docker run --rm -t \
             -v <your backup dir>:/backup \
             -e GIT_BACKUP_DIR='/backup' \
             -e GITHUB_TOKEN='MYGITHUBTOKEN' \
             -e GITLAB_TOKEN='MYGITLABTOKEN' \
             ghcr.io/jonhadfield/soba

To hide credentials, you can instead use exported environment variables and specify using this syntax:

docker run --rm -t \
             -v <your backup dir>:/backup \
             -e GIT_BACKUP_DIR='/backup' \
             -e GITHUB_TOKEN=$GITHUB_TOKEN \
             -e GITLAB_TOKEN=$GITLAB_TOKEN \
             ghcr.io/jonhadfield/soba

run on Kubernetes

For instructions on how to run soba on Kubernetes, see here.

scheduling backups

Backups can be scheduled to run by setting an additional environment variable: GIT_BACKUP_INTERVAL. The value can be specified in hours (default) or minutes. For example, this will run the backup daily:

export GIT_BACKUP_INTERVAL=24h

and this will run the backup every 45 minutes:

export GIT_BACKUP_INTERVAL=45m

note:

• if you don't specify the trailing 'm' or 'h' then hours are assumed.
• the interval is added to the start of the last backup and not the time it finished, therefore ensure the interval is greater than the duration of a backup.

rotating backups

A new bundle is created every time a change is detected in the repository. To keep only the x most recent, use the following provider specific environment variables:
GITEA_BACKUPS=x
GITHUB_BACKUPS=x
GITLAB_BACKUPS=x
BITBUCKET_BACKUPS=x

setting the request timeout

By default, soba will wait up to five minutes for a response to complete. This could be anything from an API call to discover repositories to a clone of a large repository.
If you have a slow connection or very large repositories, you may want to increase this. To do so, set the environment variable GIT_REQUEST_TIMEOUT to the number of seconds you wish to wait. For example, to wait up to ten minutes:

export GIT_REQUEST_TIMEOUT=600

notifications

Telegram

(since release 1.2.20)
To send a Telegram message on completion, set the environment variables:
SOBA_TELEGRAM_BOT_TOKEN with the bot token
SOBA_TELEGRAM_CHAT_ID with the chat/group id

To get the bot token:

• send a message to @BotFather of /newbot
• submit a name, e.g. soba-notifier
• submit a username for the bot
• record bot token

To get the chat id:

• add the bot user to the group (get group info and click Add)
• run command:curl -s -X POST https://api.telegram.org/bot<YOUR_BOT_TOKEN>/getUpdates
• record the chat id in the response

Slack

(since release 1.2.16)
To send a Slack message on completion, set the environment variables:
SLACK_CHANNEL_ID with the channel id
SLACK_API_TOKEN with the token for the Slack app
For example:
$ export SLACK_CHANNEL_ID=C12345678
$ export SLACK_API_TOKEN=xoxb-***********-************-************************

note

• channel id can be in About section at bottom of the channel details
• the app needs to be added under Apps in the Integrations section of channel details
• use the token starting with xoxb- and not the one starting with xoxp-

webhooks

(since release 1.2.8)
To send a webhook on completion of a run: set the environment variable SOBA_WEBHOOK_URL with the url of the endpoint.
For example:
$ export SOBA_WEBHOOK_URL=https://api.example.com/webhook

webhook payload

The payload is a JSON document containing details of the backup run. The default format lists each repository and the success or failure of its backup. You can see an example here.
For a shorter format, with just stats on the success and failure counts, use the environment variable SOBA_WEBHOOK_FORMAT.
For example:
$ export SOBA_WEBHOOK_FORMAT=short
You can see a sample here.
The default format (if not specified) is long

NOTE: The long format webhook will contain a list of your repos and, if there's an error, may contain other details including URLs. Please keep this in mind when sending to endpoints that may be insecure.

ntfy

(since release 1.2.10)
ntfy is a popular service that enables push notifications for desktop and mobile apps.
To send a message on completion of a run: set the environment variable SOBA_NTFY_URL with the url of the endpoint.
For example:
$ export SOBA_NTFY_URL=https://ntfy.sh/example-topic

logging

persistence

Messages are written to stdout and can be persisted by directing to a file, e.g.
soba > soba.log

logging to /var/log/soba

create a user called soba: sudo adduser soba create a log directory: sudo mkdir /var/log/soba set user permissions: sudo chown soba /var/log/soba && sudo chmod 700 /var/log/soba switch to soba user: sudo su - soba run soba and direct output: soba > /var/log/soba/soba.log

rotation

Logrotate is a utility that comes with most Linux distributions and removes and/or compresses messages older than a certain number of hours or days.
This example assumes you persist the log file to /var/log/soba/soba.log create a file in /etc/logrotate.d/soba with the following content:

/var/log/soba/soba.log {
  rotate 7      # remove backups older than seven days
  daily         # process log file each day
  compress      # compress the backup
  copytruncate  # don't delete the file after backup, but instread truncate
}

Each day, this copy the latest logs to a new file that is then compressed. The existing log file is then truncated. Any backups older than seven days are then removed.

keep running after reboot

In case the computer is rebooted or the process ends for another reason, you can ensure it automatically restarts with a simple script and cron job.

script

For example:

#!/bin/bash -e
export GIT_BACKUP_DIR=/backup-dir
export GITHUB_TOKEN=xxxxxxx   # avoid hard-coding if possible
export GITHUB_BACKUPS=7
export GIT_BACKUP_INTERVAL=12
export GITHUB_COMPARE=refs
/usr/local/bin/soba

cron job

ensure the user running soba has an entry in /etc/cron.allow.

run crontab -e

add the following (assuming you have a user called soba with a script to run it called backup in their home directory):
* * * * * /usr/bin/flock -n /tmp/soba.lockfile /home/soba/backup >> /var/log/soba/soba.log 2>&1

note: A useful tool for testing cron jobs is crontab guru.

setting provider credentials

On Linux and MacOS you can set environment variables manually before each time you run soba:

export NAME='VALUE'

or by defining in a startup file for your shell so they are automatically set and available when you need them. For example, if using the bash shell and running soba as your user, add the relevant export statements to the following file:

/home/<your-user-id>/.bashrc

and run:

source /home/<your-user-id>/.bashrc

additional options

Azure DevOps

Returning Organisations' repositories (available since soba 1.2.11)

An organisation must be specified using environment variable AZURE_DEVOPS_ORGS in order for soba to discover the projects and their repos. Note: Only a single organisation is currently supported.

Repo/Bundle comparison method

Environment variable: AZURE_DEVOPS_COMPARE

See explanation below

BitBucket

Repo/Bundle comparison method

Environment variable: BITBUCKET_COMPARE

See explanation below

Gitea

Gitea instructions

Official documentation

The value for GITEA_APIURL needs to be in the format: https://[domain]/api/v1, where domain is something like gitea.example.com.

GITEA_TOKEN is the secret you need to generate using the API (see official documentation above), or via the web GUI:

• Login to Gitea
• Select your user icon in the top right-hand corner and choose Settings from the dropdown
• Select Applications
• Enter a Token Name, e.g. "soba backups"
• Select Public only or All depending on use-case
• Expand the Select permissions menu
• Select read:organization and read:repository.
• Click on Generate Token and the value will appear at the top of the page

Returning Organisations' repositories

Repositories in Gitea organisations are not backed up by default. To back these up, specify a comma separated list of organisations in the environment variable: GITEA_ORGS. To include "all" organisations, set to *.

Gitea Repo/Bundle comparison method

Environment variable: GITEA_COMPARE

See explanation below

GitHub

Returning Organisations' repositories

Repositories in GitHub organisations are not backed up by default. To back these up, specify a comma separated list of organisations in the environment variable: GITHUB_ORGS.

Skipping User repository backups

By default, all users' repositories will be backed up, even when specifying organisations. To skip user repositories set environment variable: GITHUB_SKIP_USER_REPOS to true.

Limit user repo backups to those owned by the user

By default, all repositories a user is affiliated with, e.g. a collaborator on, are included for backup. To limit these to only those owned by the user, set environment variable: GITHUB_LIMIT_USER_OWNED to true.

GitHub Repo/Bundle comparison method

Environment variable: GITHUB_COMPARE

See explanation below

GitLab

filtering Projects by access level (available since soba 1.1.3)

The way in which a user's GitLab Projects are returned. By default, every Project a user has at least Reporter access to will be returned. New environment variable GITLAB_PROJECT_MIN_ACCESS_LEVEL can be set to override this, by specifying the number matching the desired access level shown here and here:

GitLab Repo/Bundle comparison method

Environment variable: GITLAB_COMPARE

See explanation below

Comparing remote repository with local backup

By default, each repository will be cloned, bundled, and that bundle compared with the latest local bundle to check if it should be kept or discarded. When processing many large repositories, this can be a lengthy process. Alternatively, you can now compare the Git refs of the latest local bundle with the remote repository without having to clone. This is carried out using native commands git bundle list-heads <bundle file> and git ls-remote <remote repository>. This process is far quicker than cloning but should only be used if the following is understood: Comparing refs means comparing the tips of, and not the entire history of, the repository. This post on Stack Overflow goes into additional detail.

run on Synology NAS

The following was tested on DS916+

1. Create a directory on your NAS for backing up Git repositories to
2. Install Docker from the Synology Package Center
3. Open Docker and select 'Image'
4. Select 'Add' from the top menu and choose 'Add From URL'
5. In 'Repository URL' enter 'jonhadfield/soba', leave other options as default and click 'Add'
6. When it asks to 'Choose Tag' accept the default 'latest' by pressing 'Select'
7. Select image 'jonhadfield/soba:latest' from the list and click 'Launch' from the top menu
8. Set 'Container Name' to 'soba' and select 'Advanced Settings'
9. Check 'Enable auto-restart'
10. Under 'Volume' select 'Add folder' and choose the directory created in step 1. Set the 'Mount Path' to '/backup'
11. Under 'Network' check 'Use the same network as Docker Host'
12. Under 'Environment' click '+' to add the common configuration:
    • variable GIT_BACKUP_DIR Value /backup
    • variable GIT_BACKUP_INTERVAL Value (hours between backups)
13. Also under 'Environment' click '+' to add the relevant provider specific configuration:
    • variable BITBUCKET_USER Value
    • variable BITBUCKET_KEY Value
    • variable BITBUCKET_SECRET Value
    • variable BITBUCKET_BACKUPS Value (Number of backups to keep for each repo)
    • variable GITEA_APIURL Value
    • variable GITEA_TOKEN Value
    • variable GITEA_ORGS Value
    • variable GITEA_BACKUPS Value
    • variable GITHUB_TOKEN Value
    • variable GITHUB_ORGS Value (Optional - comma separated list of organisations)
    • variable GITHUB_SKIP_USER_REPOS Value (Optional - defaults to false)
    • variable GITHUB_LIMIT_USER_OWNED Value (Optional - defaults to false)
    • variable GITHUB_BACKUPS Value (Number of backups to keep for each repo)
    • variable GITLAB_TOKEN Value
    • variable GITLAB_BACKUPS Value (Number of backups to keep for each repo)
    • variable GITLAB_PROJECT_MIN_ACCESS_LEVEL Value (Optional - scope of repos to backup)
14. Click 'Apply'
15. Leave settings as default and select 'Next'
16. Check 'Run this container after the wizard is finished' and click 'Apply'

The container should launch in a few seconds. You can view progress by choosing 'Container' in the left-hand menu, select 'soba', choose 'details' and then click on 'Log'

restoring backups

A Git bundle is an archive of a Git repository. The simplest way to restore is to clone it like a remote repository.

git clone soba.20180708153107.bundle soba