103 lines
3.5 KiB
Markdown
103 lines
3.5 KiB
Markdown
# Ansible Role: strfry
|
|
|
|
This Ansible Role builds and installs [strfry](https://github.com/hoytech/strfry), and also sets up [strfry-policies](https://gitlab.com/soapbox-pub/strfry-policies). It is intended to be composed with a separate role to handle the web proxy configuration.
|
|
|
|
Tested on:
|
|
* Archlinux
|
|
* Debian 11
|
|
* Ubuntu 22.04
|
|
|
|
## Requirements
|
|
|
|
None.
|
|
|
|
## Role Variables
|
|
|
|
```yaml
|
|
strfry_version: beta # git repository branch or release tag
|
|
strfry_make_jobs: "{{ ansible_processor_cores }}" # number of CPUs to build with
|
|
strfry_skip_config: no
|
|
strfry_policies_enabled: yes
|
|
```
|
|
|
|
See the role [defaults](defaults/main.yml).
|
|
|
|
If you are not using the `beta` branch/version, you should override the template with your own by enabling `strfry_skip_config` and managing the configuration manually.
|
|
|
|
```yaml
|
|
strfry_skip_config: yes
|
|
```
|
|
|
|
For more configuration info, see the relevant upstream [configuration example](https://github.com/hoytech/strfry/blob/beta/strfry.conf) for your branch/version.
|
|
|
|
## Example Playbook
|
|
|
|
```yaml
|
|
- hosts: strfry
|
|
roles:
|
|
- role: bleetube.strfry
|
|
- role: nginxinc.nginx_core.nginx
|
|
become: yes
|
|
tasks:
|
|
- import_tasks: nginx_conf.yml
|
|
become: yes
|
|
```
|
|
|
|
A sample [nginx configuration](docs/examples/nginx_conf.yml) is provided.
|
|
|
|
For a fully functional production example that includes hosting multiple relays, see this [homelab stack](https://github.com/bleetube/satstack).
|
|
|
|
## Upgrades
|
|
|
|
Occasionally there are upgrades that require rebuilding the database. You need to `export` before upgrading, and then `import` with the new binary. The role might do the export step, but the import needs to be done manually. Don't rely on the role for the backup. Here's a simple example:
|
|
|
|
```shell
|
|
# Before upgrade
|
|
doas -u strfry strfry export > /tmp/backup.jsonl
|
|
# After upgrade
|
|
systemctl stop strfry
|
|
mv strfry-db/data.mdb strfry-db/backup.mdb
|
|
cat /tmp/backup.jsonl | doas -u strfry strfry import
|
|
doas -u strfry strfry compact strfry-db/compact.mdb
|
|
mv strfry-db/compact.mdb strfry-db/data.mdb
|
|
systemctl start strfry
|
|
```
|
|
|
|
This is by no means the cleanest way to upgrade, but you get the idea. It's possible to perform the import in a separate process (I think you'd just use a different config file) and then sync the two databases before performing a zero downtime restart.
|
|
|
|
## Troubleshooting
|
|
|
|
* If an upgrade fails to build, it could be due to previously built objects. A simple workaround is to delete the strfry source folder `~/src/strfry` and let it try to build from scratch.
|
|
|
|
* If `make` fails, try building on a single core:
|
|
|
|
```shell
|
|
ansible-playbook playbooks/strfry/main.yml -e 'strfry_make_jobs=1'
|
|
```
|
|
|
|
* Reading your logs:
|
|
|
|
```shell
|
|
systemctl status strfry
|
|
journalctl -fu strfry
|
|
```
|
|
|
|
## Maintenance
|
|
|
|
* You should periodically run `compact` on your strfry database.
|
|
|
|
```shell
|
|
systemctl stop strfry
|
|
doas -u strfry strfry compact strfry-db/compact.mdb
|
|
mv strfry-db/compact.mdb strfry-db/data.mdb
|
|
systemctl start strfry
|
|
```
|
|
|
|
* You can prune events from the database, reducing it's size will reduce the overall compute load on the relay. Make a backup beforehand. Here is a simple example of deleting events that are more than 90 days old:
|
|
|
|
```shell
|
|
doas -u strfry strfry export > /tmp/backup.jsonl
|
|
doas -u strfry strfry delete --age=$((90 * 24 * 60 * 60))
|
|
```
|
|
For a more advanced pruning strategy, you can implement an export/import process to remove certain kinds of events more aggresively. See [bleetube/strfry-prune](https://github.com/bleetube/strfry-prune) for an example.
|
|
|