[BUG FIX] Linux installers

[sam1el]

Description

This PR addresses the following issues
#201
#188

These 2 were addressed by the same change.
The current code is as follows.

def hab_command
    cmd = ["bash #{Chef::Config[:file_cache_path]}/hab-install.sh", "-v #{hab_version}"]
    cmd.join(' ')
  end
end

This logic does not check for platform or kernel version. This causes the install resource to default to the non-kernel2 version.

The new code verifies kernel release and assigns the correct values accordingly

def hab_command
    cmd = if node['kernel']['release'].to_f < 3.0
            ["bash #{Chef::Config[:file_cache_path]}/hab-install.sh", "-v #{hab_version} -t x86_64-linux-kernel2"]
          else
            ["bash #{Chef::Config[:file_cache_path]}/hab-install.sh", "-v #{hab_version}"]
          end
    cmd.join(' ')
  end
end

This is also a problem for package verification due to to the vast sprawl in what kernels are used by the supported linux distros. With the difference in version numbers vs kernel release between similar platform types. I feel like platform_version isn't an accurate check. In my testing I've seen the following.

• Currently the provider_hab_package library does the following. The problem with this is, Red hat 6.10 is using kernel2. Suse was released with kernel 2 up to version 11.2. Centos 6 can be on kernel 3 etc.

def platform_target
          if platform_family?('windows')
            'target=x86_64-windows'
          elsif platform_family?('rhel') && node['platform_version'].to_i < 6
            'target=x86_64-linux-kernel2'
          elsif platform_family?('suse') && node['platform_version'].to_i < 6
            'target=x86_64-linux-kernel2'
          else
            ''
          end
        end

• Changing this library to use kernel release allows for the correct target version to be passed to the hab_package resource.

def platform_target
          if platform_family?('windows')
            'target=x86_64-windows'
          elsif node['kernel']['release'].to_f < 3.0
            'target=x86_64-linux-kernel2'
          else
            ''
          end
        end

• These changes together are now installing the correct version of habitat which was partially responsible for package and service load failures.
• All resources appear to be functioning with the expected idempotence.

#47

Since creating hab-sup.service is done by the hab_sup resource which, also calls the install resource, I took a look at the code being used in the library resource_hab_sup_systemd and it was missing the ExecStop which is all that is required to add the systemctl restart I added the line along with a few other functions.

systemd_unit 'hab-sup.service' do
          content(Unit: {
                    Description: 'The Habitat Supervisor',
                  },
                  Service: {
                    LimitNOFILE: (new_resource.limit_no_files if new_resource.limit_no_files),
                    Environment: service_environment,
                    ExecStart: "/bin/hab sup run #{exec_start_options}",
                    ExecStop: '/bin/hab sup term',
                    Restart: 'on-failure',
                    RestartSec: '10',
                    KillMode: 'process',
                  }.compact,
                  Install: {
                    WantedBy: 'default.target',
                  })
          action :create
        end

Issues Resolved

• Issue Linux install script broken and hap-sup loaders not looking for correct init type for kernel2 systems(sysvinit vs upstart) #201 is now corrected and the correct hab versions are being install on linux
• Issue platform_target method is fairly broken #188 changes from using a platform_target to a kernel _release target which allows a more accurate detection of package versions.
• Issue hab-sup lacks restart action on systemd #47 hab-sup.service in systemd now provides a restart function.

Check List

• All tests pass. See https://github.com/chef-cookbooks/community_cookbook_documentation/blob/master/TESTING.MD
• New functionality includes testing.
• New functionality has been documented in the README if applicable
• All commits have been signed for the Developer Certificate of Origin. See https://github.com/chef-cookbooks/community_cookbook_documentation/blob/master/CONTRIBUTING.MD

[tas50]

Thanks. This was some hot nonesense here

[sam1el]

It was causing some serious crazy to happen

[tas50]

You could easily just .to_i < 3 here as well. No need to treat it as a float

[sam1el]

You're right. It think at first I was accidentally calling version rather than release and didnt understand why I was failing and forgot to change it when i caught my original mistake. Will change that up and push it again

[tas50]

Suggested change

	elsif node['kernel']['release'].to_i < 3.0
	elsif node['kernel']['release'].to_i < 3

[tas50]

Suggested change

	cmd = if node['kernel']['release'].to_i < 3.0
	cmd = if node['kernel']['release'].to_i < 3

[sam1el]

I can't believe I did that on both of those. They are fixed. Thanks pointing it out.