72 Zeilen
2,7 KiB
Markdown
72 Zeilen
2,7 KiB
Markdown
|
# Cold-standby backup
|
||
|
|
|
||
|
|
mailcow offers an easy way to create a consistent copy of itself to be rsync'ed to a remote location without downtime.
|
||
|
|
|
||
|
|
This may also be used to transfer your mailcow to a new server.
|
||
|
|
|
||
|
|
## You should know
|
||
|
|
|
||
|
|
The provided script will work on default installations.
|
||
|
|
|
||
|
|
It may break when you use unsupported volume overrides. We don't support that and we will not include hacks to support that. Please run and maintain a fork if you plan to keep your changes.
|
||
|
|
|
||
|
|
The script will use **the same pathes** as your default mailcow installation. That is the mailcow base directory - for most users `/opt/mailcow-dockerized` - as well as the mountpoints.
|
||
|
|
|
||
|
|
To find the pathes of your source volumes we use `docker inspect` and read the destination directory of every volume related to your mailcow compose project. This means we will also transfer volumes you may have added in a override file. Local bind mounts may or may not work.
|
||
|
|
|
||
|
|
The use rsync with the `--delete` flag. The destination will be an exact copy of the source.
|
||
|
|
|
||
|
|
`mariabackup` is used to create a consistent copy of the SQL data directory.
|
||
|
|
|
||
|
|
After rsync'ing the data we will run `docker-compose pull` and remove old image tags from the destination.
|
||
|
|
|
||
|
|
Your source will not be changed at any time.
|
||
|
|
|
||
|
|
**You may want to make sure to use the same `/etc/docker/daemon.json` on the remote target.**
|
||
|
|
|
||
|
|
You should not run disk snapshots (e.g. via ZFS, LVM etc.) on the target at the very same time as this script is run.
|
||
|
|
|
||
|
|
Versioning is not part of this script, we rely on the destination (snapshots or backups). You may also want to use any other tool for that.
|
||
|
|
|
||
|
|
## Prepare
|
||
|
|
|
||
|
|
You will need a SSH-enabled destination and a keyfile to connect to said destination. The key should not be protected by a password for the script to work unattended.
|
||
|
|
|
||
|
|
In your mailcow base directory, e.g. `/opt/mailcow-dockerized` you will find a file `create_cold_standby.sh`.
|
||
|
|
|
||
|
|
Edit this file and change the exported variables:
|
||
|
|
|
||
|
|
```
|
||
|
|
export REMOTE_SSH_KEY=/path/to/keyfile
|
||
|
|
export REMOTE_SSH_PORT=22
|
||
|
|
export REMOTE_SSH_HOST=mailcow-backup.host.name
|
||
|
|
```
|
||
|
|
|
||
|
|
The key must be owned and readable by root only.
|
||
|
|
|
||
|
|
Both the source and destination require `rsync` >= v3.1.0.
|
||
|
|
The destination must have Docker and docker-compose **v1** available.
|
||
|
|
|
||
|
|
The script will detect errors automatically and exit.
|
||
|
|
|
||
|
|
You may want to test the connection by running `ssh mailcow-backup.host.name -p22 -i/path/to/keyfile`.
|
||
|
|
|
||
|
|
## Backup and refresh the cold-standby
|
||
|
|
|
||
|
|
Run the first backup, this may take a while depending on the connection:
|
||
|
|
|
||
|
|
```
|
||
|
|
bash /opt/mailcow-dockerized/create_cold_standby.sh
|
||
|
|
```
|
||
|
|
|
||
|
|
That was easy, wasn't it?
|
||
|
|
|
||
|
|
Updating your cold-standby is just as easy:
|
||
|
|
|
||
|
|
```
|
||
|
|
bash /opt/mailcow-dockerized/create_cold_standby.sh
|
||
|
|
```
|
||
|
|
|
||
|
|
It's the same command.
|
||
|
|
|