The retry logic for VMPooler has two issues. 1) Waiting longer than
VMPooler takes to refill the pool to retry introduces excessive delays,
and 2) waiting for a deterministic amount of time means that processes
started at the same time stay in sync causing peak loads.
This change caps the wait time at 15 seconds plus a random wait time,
addressing both issues.
* (MAINT) Update specs to remove VsphereHelper
VsphereHelper is no longer relevant since vmpooler doesn't use vSphere/vCloud any more.
* (MAINT) Drop conflicting, unused dependency on beaker-abs
* (MAINT) Conditionally pin fakefs '< 0.14.0' for Ruby < '2.3.0'
* (MAINT) Remove more VsphereHelper references
And minor whitespace cleanup
* (MAINT) Drop unused beaker-abs dependency
Could have sworn I fixed that... maybe in a different branch?
Previously, beaker-vmpooler defined its
connection preferences as string options, but
beaker provides defaults as symbols. This ends
with beaker saying that it doesn't support methods
that it does due to the mismatch. This will get
rid of those messages, as the provided methods
will be symbols, matching what beaker expects
Before, the host configuration was hard-coded. Added an option
to specify an ENV var in order to use a different string or a
pregenerated hostfile.
This also allows us to use abs in production, so that has been added
to the Gemfile
