Fix wait_for_task backoff behavior in vmware modules

[albers]

SUMMARY

I stepped into this when I tried vmware_guest_powerstate to turn off a running VM in VMware by setting state=powered-off. The instance went down almost immediately, but I had to wait another minute for the module task to return.

In #39596, a timer with an exponential backoff algorithm was added:

sleep_time = min(2 ** failure_counter + randint(1, 1000), max_backoff)

The intent was obviously to wait for [1,2,4,8,16,32,64,64...] seconds, with some fuzzing added to the seconds. But as it is implemented, the added randint(1, 1000) will be greater than max_backoff (64) most of the time, resulting in a 93,6% chance that the first wait duration will be 64 seconds.

I assume that the random value was intended to add a slight fuzzing in the range 0..1000 milliseconds.

This PR adjusts the code to that assumption.

ISSUE TYPE

• Bugfix Pull Request

COMPONENT NAME

lib/ansible/module_utils/vmware.py

ANSIBLE VERSION

ansible 2.6.3
  config file = /etc/ansible/ansible.cfg
  configured module search path = [u'/home/user/.ansible/plugins/modules', u'/usr/share/ansible/plugins/modules']
  ansible python module location = /usr/lib/python2.7/dist-packages/ansible
  executable location = /usr/bin/ansible
  python version = 2.7.12 (default, Dec  4 2017, 14:50:18) [GCC 5.4.0 20160609]

ADDITIONAL INFORMATION

sample invocation, credentials specified in environment variables

hacking/test-module -m lib/ansible/modules/cloud/vmware/vmware_guest_powerstate.py -a 'state=powered-off name=some-running-vm'

[ansibot]

cc @Akasurde @dav1x @warthog9
click here for bot help

[ansibot]

The test ansible-test sanity --test pep8 [explain] failed with 1 error:

lib/ansible/module_utils/vmware.py:61:69: E226 missing whitespace around arithmetic operator

click here for bot help

[Akasurde]

LGTM

[albers]

@Akasurde Thanks for merging so quickly and also for fixing the whitespace issue.

By the way, this fix explains the flakyness reported in #45223: Most of the time, the VMs had 64 seconds time to come up, which would be sufficient for the IP address to become available. A shorter timeout might not be sufficient.

I'm looking forward to some noticable faster executions now.