I'm running Sidekiq 6.4.1 on an Ubuntu 20.04 system with Ruby 3.1.1 installed via Snap for a Rails 7 application.
Using this systemd template, I noticed something odd when deploying with Capistrano. When my cap script ran `sudo systemctl restart sidekiq`, the Sidekiq process would restart, briefly go active, then go inactive and stop running. I would need to SSH in and manually do a `systemctl restart sidekiq` in order to get the process back up and running.
With some experimentation, I found that changing the `Restart` configuration from `on-failure` to `always` seems to do the trick. Now, when the Sidekiq process goes inactive, it properly restarts and stays restarted.
I'm not 100% sure of the "why" behind what's going on, but this change seems to be effective and I have not noticed any side effects.
Documentation of previous behavior
Prior to deploy (process running)
```
root@stage-worker01:~# systemctl status sidekiq
● sidekiq.service - sidekiq
Loaded: loaded (/lib/systemd/system/sidekiq.service; enabled; vendor prese>
Active: active (running) since Thu 2022-03-24 16:34:32 UTC; 51s ago
Main PID: 118856 (bundle)
Tasks: 30 (limit: 1120)
Memory: 291.2M
CGroup: /system.slice/sidekiq.service
└─118856 sidekiq 6.4.1 staging [0 of 20 busy]
```
Immediately post deploy (process restarting)
```
root@stage-worker01:~# systemctl status sidekiq
● sidekiq.service - sidekiq
Loaded: loaded (/lib/systemd/system/sidekiq.service; enabled; vendor prese>
Active: activating (start) since Thu 2022-03-24 16:38:35 UTC; 9s ago
Main PID: 119572 (bundle)
Tasks: 0 (limit: 1120)
Memory: 43.4M
CGroup: /system.slice/sidekiq.service
‣ 119572 /var/www/sox/staging/shared/bundle/ruby/3.1.0/bin/sidekiq>
```
After deploy (process started)
```
Mar 24 16:38:35 stage-worker01.suggestionox.com systemd[1]: Starting sidekiq...
root@stage-worker01:~# systemctl status sidekiq
● sidekiq.service - sidekiq
Loaded: loaded (/lib/systemd/system/sidekiq.service; enabled; vendor prese>
Active: active (running) since Thu 2022-03-24 16:38:45 UTC; 3s ago
Main PID: 119572 (bundle)
Tasks: 0 (limit: 1120)
Memory: 43.4M
CGroup: /system.slice/sidekiq.service
‣ 119572 sidekiq 6.4.1 staging [0 of 20 busy]
Mar 24 16:38:35 stage-worker01.suggestionox.com systemd[1]: Starting sidekiq...
Mar 24 16:38:45 stage-worker01.suggestionox.com systemd[1]: Started sidekiq.
```
Failure (process inactive)
```
root@stage-worker01:~# systemctl status sidekiq
● sidekiq.service - sidekiq
Loaded: loaded (/lib/systemd/system/sidekiq.service; enabled; vendor prese>
Active: inactive (dead) since Thu 2022-03-24 16:38:56 UTC; 2s ago
Process: 119572 ExecStart=/usr/bin/env bundle exec sidekiq -e staging (code>
Main PID: 119572 (code=exited, status=0/SUCCESS)
Mar 24 16:38:35 stage-worker01.suggestionox.com systemd[1]: Starting sidekiq...
Mar 24 16:38:45 stage-worker01.suggestionox.com systemd[1]: Started sidekiq.
Mar 24 16:38:56 stage-worker01.suggestionox.com systemd[1]: sidekiq.service: Su>
```
You recommend the use of capistrano-sidekiq. This uses a user service (which is likely a good idea for most users from a security perspective). In order to get it to work, changes need to be made to the Sidekiq.service file (and location)
RVM has option to run commands in specific directory, and if this directory has `.ruby-version` (with optional `.ruby-gemset`) or `.rvmrc` files in it, this version is automatically used.
Thanks to this, you no longer need to update systemd service configuration on Ruby version change.
