RemoteStateReference extremely slow

[cheribral]

I am experimenting with the remote state to see whether we can use pulumi along side terraform. I tried a very simple python program which reads a 300k state file and exports one nested map.

Expected behavior

I would expect pulumi up to return rather quickly.

Current behavior

It takes on the order of 4-5 minutes to finish running and print the outputs.

Steps to reproduce

import pulumi
import pulumi_terraform as terraform

tf = terraform.state.RemoteStateReference('tf', backend_type='local', args=terraform.state.LocalBackendArgs(path='resources/tfstate/stage1.tfstate'))
pulumi.export('test', tf.get_output('sides'))

This is using:

• pulumi v3.2.0
• python 3.6 (centos 7)
• pulumi-terraform v5.1.0

[cheribral]

I've done a bit more work with pulumi now, and I'm not sure this problem is entirely specific to pulumi_terraform. This seems to be a problem I can also cause with general stack references. It must be doing something special, because normally it should be very quick to load a json file, but this can use a lot of CPU and take a long time relatively speaking.

I am guessing (heavy emphasis on "guess") this is because these remote state calls are trying to manage secrets and dependencies properly, converting everything to outputs etc. If this is the case, would it make more sense to somehow skip all this processing for terraform related inputs since they were never tied into the pulumi machinery to begin with?