systemd: Neither a valid executable name nor an absolute path error because of ExecStart

[steverob]

Capistrano::Puma::Systemd is generating a systemd file for Puma with the below ExecStart value which uses the $HOME env var:

ExecStart=$HOME/.rbenv/bin/rbenv exec bundle exec puma -C /home/ubuntu/app/shared/puma.rb

This is causing the following error:

systemd/system/puma.service:9: Neither a valid executable name nor an absolute path: $HOME/.rbenv/bin/rbenv
service: Unit configuration has fatal error, unit will not be started.

Going through systemd's manpage, it does say we need to use an absolute path for ExecStart: (https://man7.org/linux/man-pages/man5/systemd.service.5.html)

For each of the specified commands, the first argument must be either an absolute path...

Systemd version:

ubuntu@ip-172-31-86-176:/etc/systemd$ systemctl --version
systemd 245 (245.4-4ubuntu3.2)
+PAM +AUDIT +SELINUX +IMA +APPARMOR +SMACK +SYSVINIT +UTMP +LIBCRYPTSETUP +GCRYPT +GNUTLS +ACL +XZ +LZ4 +SECCOMP +BLKID +ELFUTILS +KMOD +IDN2 -IDN +PCRE2 default-hierarchy=hybrid

Ubuntu version: Linux ip-172-31-86-176 5.4.0-1029-aws #30-Ubuntu SMP Tue Oct 20 10:06:38 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux

[m4c3]

Try putting set :rbenv_custom_path, "/home/ubuntu/.rbenv" into your deploy.rb

[ccastillop]

I changed set :rbenv_prefix, "RBENV_ROOT=#{fetch(:rbenv_path)} RBENV_VERSION=#{fetch(:rbenv_ruby)} #{fetch(:rbenv_path)}/bin/rbenv exec" to set :rbenv_prefix, "#{fetch(:rbenv_path)}/bin/rbenv exec"

It seems adding ENV to ExecStart just does not work

[seuros]

@ccastillop It is possible to set env variables now with 5.0.3

[TalkingQuickly]

Fix for me on Ubuntu 20.04 was to use Environment in the unit file to set env vars rather than setting them inline in ExecStart e.g:

[Service]
Type=simple
User=deploy
WorkingDirectory=/home/deploy/apps/rsra_rails6_example_production/current
Environment=RBENV_ROOT=/usr/local/rbenv
Environment=RBENV_VERSION=3.0.0
ExecStart=/usr/local/rbenv/bin/rbenv exec bundle exec puma -C /home/deploy/apps/rsra_rails6_example_production/shared/puma.rb

[TalkingQuickly]

My solution above has a bunch of holes in it. Looked into it a bit further and in my case it's due to how the Capistrano rbenv plugin changes the return of SSHKit.config.command_map[:bundle] so that it includes environment variables at the start. I ended up with this definition; https://github.com/TalkingQuickly/rdra_rails6_example/blob/master/config/deploy/templates/puma.service.erb

It's not generalisable, e.g. it has some rbenv specific stuff in it including hard coding the rbenv path to the default ubuntu system install one. But may be useful as a starting point for others.

[chriswnl]

Adding
append :puma_service_unit_env_vars, "RBENV_VERSION=#{fetch(:rbenv_ruby)}", "RBENV_ROOT=/home/deployer/.rbenv"
to deploy.rb fixed it for me.

[sagar-ranglani]

Still not working for me, I tried solutions provided by @chriswnl, @ccastillop and @m4c3.
Using Capistrano 3.16.0, Puma 5.3.1 & capistrano3-puma 5.0.3

When I directly change ExeStart in the .service file, I can run Puma when I provide the full path.
I wish to do it using the deploy.rb file.

Any help will be appreciated. Thank you! :-)

[hoppergee]

@sagar-ranglani

Still not working for me, I tried solutions provided by @chriswnl, @ccastillop and @m4c3.
Using Capistrano 3.16.0, Puma 5.3.1 & capistrano3-puma 5.0.3

When I directly change ExeStart in the .service file, I can run Puma when I provide the full path.
I wish to do it using the deploy.rb file.

Any help will be appreciated. Thank you! :-)

You can fix this issue in deploy.rb file through something like this:

set :rbenv_custom_path, "/home/ubuntu/.rbenv"

Ref: https://github.com/capistrano/rbenv

[rizqirizqi]

I'm also stuck with a similar error. What if I use rvm instead of rbenv?

[rizqirizqi]

Hi, I finally solved my issue by modifying the systemd unit file /etc/systemd/system/puma_rasylva-web_production.service manually on my server. However, I think it's a temporary solution, it may be better if the template can generate a more accurate systemd unit file.

GENERATED BY TEMPLATE (lib/capistrano/templates/puma.service.erb):

[Unit]
Description=Puma HTTP Server for rasylva-web (production)
After=network.target

[Service]
Type=simple
User=
WorkingDirectory=/home/rasylva/apps/rasylva-web/current
ExecStart=~/.rvm/bin/rvm 3.0.1 do bundle exec puma -C /home/rasylva/apps/rasylva-web/shared/puma.rb
ExecReload=/bin/kill -TSTP $MAINPID
StandardOutput=append:/home/rasylva/apps/rasylva-web/current/log/puma.error.log
StandardError=append:/home/rasylva/apps/rasylva-web/current/log/puma.access.log
Restart=always
RestartSec=1
SyslogIdentifier=puma

[Install]
WantedBy=multi-user.target

MODIFIED:

[Unit]
Description=Puma HTTP Server for rasylva-web (production)
After=network.target

[Service]
Type=simple
User=rasylva
Environment=RAILS_ENV=production PATH=/usr/local/bin:/home/rasylva/.rvm/rubies/ruby-3.0.1/bin:/home/rasylva/.nvm/versions/node/v14.17.0/bin
WorkingDirectory=/home/rasylva/apps/rasylva-web/current
ExecStart=/home/rasylva/apps/rasylva-web/current/sbin/puma -C /home/rasylva/apps/rasylva-web/shared/puma.rb /home/rasylva/apps/rasylva-web/current/config.ru > /home/rasylva/apps/rasylva-web/current/log/puma.access.log 2>/home/rasylva/apps/rasylva-web/current/log/puma.error.log
ExecReload=/bin/kill -TSTP $MAINPID
Restart=always
RestartSec=1
SyslogIdentifier=puma

[Install]
WantedBy=multi-user.target

Here are some notes of what makes it works and I think it may help others and improve capistrano-puma template:

1. I need to specify User manually because I don't specify puma_systemctl_user config (I just upgraded from the capistrano3-puma v3.1.1 and puma v3.11 to capistrano3-puma v5.0.4 and puma 5.3.0, and the previous config doesn't have puma_systemctl_user)
2. I need to add the rvm Environment so that it can run ruby command
3. I need to add nvm Environment so that it can run node related command
4. The config.ru file is needed, otherwise, it will show ERROR: No application configured, nothing to run
5. I cannot use StandardOutput and StandardError because my systemd version somehow doesn't recognize the append command, it shows an error: Failed to parse output specifier, ignoring: append:/home/rasylva/apps/rasylva-web/current/log/puma.error.log. The alternative way is using stdout and stderr redirection: <THE_PUMA_COMMAND> > /home/rasylva/apps/rasylva-web/current/log/puma.access.log 2>/home/rasylva/apps/rasylva-web/current/log/puma.error.log

For debugging the systemd, it may also be helpful for others:
https://containersolutions.github.io/runbooks/posts/linux/debug-systemd-service-units/

[pedromschmitt]

Thx to the @rizqirizqi I could fix the same problem using RVM too.

I just added the path to RVM instead of using the ~ to solve it. I didn't need to add environment or other things

In my case changing the ~/.rvm to /home/deploy on line 9.

[github-actions]

Stale issue message

[virusman]

I don't manually editing the configs is the proper solution to this. Is there a way to fix it in the gem?