Is it normal for pulumi-aws ec2 resources to get spun up/down serially, even though `pulumi up`/`down` is being called with

--parallel 100

. I simply have a list of ec2 instances that I'm spinning up and down, and I've notices that the bigger the list, the longer these operations take. Have I accidentally introduced some kind of data dependency between my machines that forces Pulumi to process them one-by-one?

Unfortunately, until https://github.com/pulumi/pulumi/issues/1063 is implemented, Python programs cannot be parallelized; so they always run serially. I think we have a reasonable idea of the design there -- it's "just" a matter of work (which we can prioritize higher to unblock you for sure). Are all of your programs in Python, or do you also use Java/TypeScript?

All in python, but that's only because we run Python apps with wallaroo. For consistency. I guess there's not reason not to do it in typescript if that means true parallelism

In general, TypeScript is a bit further along than Python -- in this regard and in the richness of higher level libraries like https://github.com/pulumi/pulumi-aws-infra, but we are committed to supporting them both at the same level long-term, it's just a moment in time thing.

Hmm, there's one thing that I forgot about. I generally wanted to keep the number of dependencies low -- if we keep the stack definition in Python, that means the entire wallaroo deployment system is just a

pip install -r requirements.txt

away

extra vote for python3x support as well!

@brave-angle-33257 come join #python! I’

at Wallaroo, we've currently got python3 support on our plate. python2 will retire soon: https://pythonclock.org/

Starting with the 0.16 SDK (which should be released soon!) we will delete clusters in parallel too!

That's awesome news!