pulumi: pulumi {preview/up/...} is slow with medium sized TS stack

I have a medium sized stack (181 total resources) with an AKS cluster, a few supporting azure resources, nginx-ingress, cert-manager, and a few other kubernetes resources.

---

I took a trace of a no-op up:

pulumi up --skip-preview --tracing=file:./up.trace

If you want to take a look, I can send you the details (update url and trace file [170MB]) per email or chat. (note to self: it’s update # 27)

A few things I noticed myself: there are a lot of calls to api/decryptValue, each about 150ms to 200ms.

The pulumi trace alone is 1m 9s:

Also, apparently the time logging is wrong/incomplete. In the console output it says Duration: 35s (I didn’t stop the time myself), but in the trace file its over a minute.

Inside there, it’s mostly /pulumirpc.ResourceMonitor/RegisterResource and /pulumirpc.ResourceMonitor/RegisterResource with 1s to 6s each.

---

I’m on windows 10 x64 with pulumi 3.9.0. The environment is Typescript / node. I’m mostly using “@pulumi/azure-native” and “@pulumi/kubernetes”.

---

Edit: I ran pulumi up --skip-preview (no-op) again and stopped the time with my phone. It took 1m 4s in reality, the cli output said 32s.

---

Edit2: pulumi refresh --target i_do_not_exist takes about 30 seconds: