SLE 11 does not ship systemd and then using systemctl poweroff does not
work. Therefore we fall back to using /sbin/shutdown for machines without
systemd.
This fixes https://github.com/hashicorp/vagrant/issues/12487
We have started seeing occasional shutdown failures on openSUSE Tumbleweed with
Virtualbox inside a qemu virtual machine, where `shutdown -h now` would return
nil. While the machine is successfully turned off, the command fails and vagrant
reports an error.
This commit changes the shutdown command to launch in the background which
also triggers a shutdown, but always succeeds.
This adds a check to the `#wait_for_reboot` method on the linux
guest reboot capability to determine if the a reboot is still
in process. This prevents the reboot process from being initiated
and the `#ready?` check on the guest being called before the
system shutdown process has shutdown the communicator process.
Keep the default duration as a constant and fetch the custom environment
variable at run time with a fallback to the default. Set the sleep duration
into a constant and add tests covering the expected behaviors when the
default duration is in use as well as the override value. Also match the
environment variable up with the constant just for consistency.
Previously the maximum amount of time Vagrant would poll for whether a
machine has successfully reboot was hard coded to 120 seconds. This
change introduces the VAGRANT_MAX_REBOOT_RETRY_TIMEOUT environment
variable to allow this attribute to be configurable.
Add RSpec tests of the maximum retry logic. Since the maximum retries
are configured as a constant, we'd need to reload the class and that's
fairly ugly to do in RSpec.
Fixes#11695
Leaving out smbclient from the install will cause a cyclical
dependency error.
```
Was getting error
Stderr from the command:
warning: dependency cycle detected:
warning: smbclient will be installed before its cifs-utils dependency
error: failed to commit transaction (conflicting files)
```
This corrects the `mkdir` command used by rsync on Windows to make sure
the destination directory exists before starting to sync. The old form
was correct on Linux but not on Windows, and it was just a coincidence
that the `-p` argument appeared to be work.
