Why does redirecting stderr make my App Engine deployment failing consistently?

Viewed 119

I am deploying a trivial App Engine Standard Environment app. (Literally the shortest possible, a Python 3 "hello world".) I am using Macbook with zshell.

If I redirect standard error to file, I get an error (below) every time.

  gcloud app deploy -q  2>>err.log 

If I omit the redirection, it succeeds every time.

There is no difference between using > or >>. Redirecting with a pipe, e.g. to grep, does not cause the problem.

So this is a "solution" (by sending output through a passthrough grep) that does what I need and does not trigger the problem, but this is very roundabout.

  gcloud app deploy -q 2>&1 >/dev/null |egrep "." >> err.txt 

Note that I use -q, so waiting for my Y for approval is not the issue.

The error is this. (Identifiers were anonymized.)

..................failed.
ERROR: (gcloud.app.deploy) Error Response: [9] Cloud build BUILD_ID status: FAILURE
Build error details: Failed to download at least one file. Cannot continue.

Full build logs: https://console.cloud.google.com/cloud-build/builds/BUILD_ID?project=PROJECT_ID

Looking at the logs, I see this.

starting build "BUILD_ID"
FETCHSOURCE
BUILD
Starting Step #0 - "fetcher"
Step #0 - "fetcher": Already have image (with digest): gcr.io/cloud-builders/gcs-fetcher
Step #0 - "fetcher": Fetching manifest gs://staging.joshua-playground.appspot.com/ae/BUILD_ID/manifest.json.
Step #0 - "fetcher": Processing 728 files.
Step #0 - "fetcher": Failed to fetch gs://staging.my-project.appspot.com/BUILD_ID, will no longer retry: fetching "gs://staging.my-project.appspot.com/BUILD_ID" with timeout 1h0m0s to temp file "/workspace/.download/staging.joshua-playground.appspot.com-BUILD_ID": err SHA mismatch, got "SHA_VALUE", want "SHA_VALUE"
Step #0 - "fetcher": Failed to download at least one file. Cannot continue.
Finished Step #0 - "fetcher"
ERROR
ERROR: build step 0 "gcr.io/cloud-builders/gcs-fetcher" failed: step exited with non-zero status: 1
1 Answers

Google has confirmed this issue. Please track it here.

Related