Closed theballdredge closed 8 years ago
What about changing the delay growth to just add 5 secs each time rather than doubling? That gets you closer to what you're looking for, but also keeps the relative simplicity of the current delay+retry logic.
definitely closer. the delay at the point the machine is ready is going to be in the 80-90s range between checks, vice 512.
might be nice to be able to specify the duration too (eg add X seconds), time permitting.
The main reason it's in here is I've seen the vRA API lose its mind if it's used too much, requiring a restart of the software.
I think the right call here is to add 5 minutes each time up to a max of a 30-sec delay. I think that's pretty reasonable.
capping the delay sounds like an excellent compromise.
aside - which versions of vRA have you seen misbehaving?
PR #10 opened for this - please follow that PR for additional updates.
I believe it was 6.2, or 6.2.1, that I was seeing the API scalability issues.
kitchen-vra v1.3.0 has been released with this change - thanks again for raising your concerns to us.
when creating machines using vRA our system has some post processing that takes a few minutes to complete, and often runs into larger timeout values. specifically, the timeout frequently runs past the 256s timeout and often triggers the 512s delay.
it would be really nice if one could set a base delay before delay scaling kicks in.