I'm attempting to debug why my pulumi up commands ...
# azure
m
I'm attempting to debug why my pulumi up commands sometimes hang for minutes. I'm using dotnet and the azure blob persistence for state files. One thing I've noticed is a repeating error while an update is running:
Copy code
error copying .pulumi/stacks/<stackname>.json.gz to .pulumi/stacks/<stackname>.json.gz.bak: blob (key ".pulumi/stacks/<stackname>.json.gz -> .pulumi/stacks/<stackname>.json.gz.bak") (code=NotFound): PUT https://<storageaccount>.blob.core.windows.net/<containername>/.pulumi/stacks/<containername>/<stackname>.json.gz.bak
The actual names of these blobs do not have the
.gz
extension in the storage account. The names are:
Copy code
.pulumi/stacks/<stackname>.json

and

.pulumi/stacks/<stackname>.json.bak
That said, the correctly named blobs are being updated upon conclusion of the updates. At the end:
Copy code
Saved stack <stackname> checkpoint to: .pulumi/stacks/<stackname>.json (backup=.pulumi/stacks/<stackname>.json.bak)
It just appears that pulumi is doing a lot of incremental 404s against azure w/ the wrong blob names.
Just found this issue that is 2 years old. https://github.com/pulumi/pulumi/issues/10960 When running an update to create a half dozen resources on a stack with 400 resources, I output logs during the run. Resource creation took maybe 2 minutes, but the stack run took 9 minutes. The logs included the error response
RESPONSE 404: 404 The specified blob does not exist
a total of 852 times! What gives? Including @echoing-dinner-19531 as a commentor on the issue.