💾 Archived View for tdem.in › post › restic-with-systemd.gmi captured on 2023-03-20 at 17:29:17. Gemini links have been rewritten to link to archived content

View Raw

More Information

⬅️ Previous capture (2022-06-03)

-=-=-=-=-=-=-

Back to blog

Using Restic with systemd on Linux

2021-01-29 20:30

You can read the Russian version of this post here.

here

Restic, the simple backup program, is a fairly well-known piece of software. Designed to be simple to both use and script on any system, it doesn't include any OS-specific setup examples, which is precisely what this post describes.

Restic

So what we're trying to achieve here:

1. Automated backup runs daily at 12:00 AM (or any configurable time).

2. The backup includes only important configuration files and data stores.

3. The backup also includes all PostgreSQL databases, restorable with `psql -f`.

4. The backup expands to an unlimited number of repos on need.

This guide assumes we're backing up to a rest-server instance running at `192.168.1.200`. The configuration should still be trivial to adapt to pretty much any storage provider. This guide also assumes you have already initialized a rest-server repo with `restic init -r rest:http://192.168.1.200/your-repo/`.

rest-server

For this we will need two systemd services, two correspondent timers, and a single helper script.

Backing up files/directories

We'd rather not want to run restic as root for obvious reasons, so let's create a dedicated user:

# sudo useradd -m -N -s /usr/sbin/nologin restic

To backup files, we'll need this service[^1] along with its timer:

[^1]: A good catch by Geoffrey Brown: if you use instance names with dashes or other special symbols as documented in systemd.unit(5), you should be aware of escaping it performs on instance names; see "String Escaping for Inclusion in Unit Names" section for more details. You can also use the `%i` specifier which doesn't escape instance names.

systemd.unit(5)

`/etc/systemd/system/restic@.service`:

[Unit]
# this unit can be activated with a parameter, e.g. in
#   systemctl start restic@your-repo.service
# %I is "your-repo"
Description=Restic backup on %I
After=syslog.target
After=network-online.target

[Service]
Type=oneshot
User=restic
# runs restic backup on the files listed in /etc/restic/your-repo.files
ExecStart=/usr/local/bin/restic backup --files-from /etc/restic/%I.files
# source repo and password from /etc/restic/your-repo.env
EnvironmentFile=/etc/restic/%I.env
AmbientCapabilities=CAP_DAC_READ_SEARCH

[Install]
WantedBy=multi-user.target

`/etc/systemd/system/restic@.timer`:

[Unit]
# the timer, enabled as restic@your-repo.timer, will trigger
# restic@your-repo.service
Description=Run Restic at 12:00 AM

[Timer]
OnCalendar=*-*-* 0:00:00

[Install]
WantedBy=timers.target

The repo name is passed through an environment file in `/etc/restic`, read by systemd (systemd does this as root, and `/etc/restic` should really be only readable as root, so set permissions accordingly):

`/etc/restic/your-repo.env`

RESTIC_PASSWORD=your_repo_password
RESTIC_REPOSITORY=rest:http://192.168.1.200/your-repo/

We also have to supply restic with a file/directory list to back up:

`/etc/restic/your-repo.files`

/var/lib/docker
/etc/postgresql
/etc/restic
...

Backing up databases

This one is just a little less trivial.

Restic supports backing up data provided through stdin, so we can feed it with the output of `pg_dumpall`. The only limitation is that systemd runs whatever you specify in `ExecStart` with `execve(3)`, and, to use output redirection, we'll need a separate bash script:

`/usr/local/bin/pgdump.sh`:

#!/usr/bin/env bash

set -euo pipefail

/usr/bin/sudo -u postgres /usr/bin/pg_dumpall --clean \
    | gzip --rsyncable \
    | /usr/local/bin/restic backup --host $1 --stdin \
        --stdin-filename postgres-$1.sql.gz

To run `pg_dumpall` as `postgres`, we'll need the following sudo rule in `/etc/sudoers`:

restic ALL=(postgres) NOPASSWD:/usr/bin/pg_dumpall --clean

The unit file is as trivial as this:

`/etc/systemd/system/restic-pg@.service`:

[Unit]
Description=Restic PostgreSQL backup on %I
After=syslog.target
After=network-online.target
After=postgresql.service
Requires=postgresql.service

[Service]
Type=oneshot
User=restic
ExecStart=/usr/local/bin/pgdump.sh %I
EnvironmentFile=/etc/restic/%I.env

[Install]
WantedBy=multi-user.target

The timer isn't really any different from the one we've already seen above:

`/etc/systemd/system/restic-pg@.timer`

[Unit]
Description=Run Restic on PostgreSQL at 12:00 AM

[Timer]
OnCalendar=*-*-* 0:00:00

[Install]
WantedBy=timers.target

Finishing up

Start the timers and enable their autostart at boot. Remember that `your-repo` is used to expand file paths in `/etc/restic`:

# systemctl enable --now restic@your-repo.timer
# systemctl enable --now restic-pg@your-repo.timer

Test whether the whole backup system is working:

# systemctl start restic@your-repo.service
# systemctl start restic-pg@your-repo.service

These unit files allow backing up to an unlimited number of repos as long as the relevant configuration is provided through `/etc/restic/repo-name.{env,files}`.

Links / sources

Recipe to backing up PostgreSQL in a Docker container used for the backup script was originally found on restic forum. The excellent systemd docs (systemd.service, systemd.timer) also helped a lot.

Recipe to backing up PostgreSQL in a Docker container

systemd.service

systemd.timer