error building image: dial tcp i/o timeout occurs with v0.10.0

[drpmma]

Actual behavior
The error: error building image: Get https://xxx/v2/xxx: dial tcp xx:443: i/o timeout occurs when running build process with v0.10.0. However, the same build process works totally fine in v0.9.0.
And the flags --skip-tls-verify-pull and --insecure-pull have no effect on this problem.

Expected behavior
Build and push success.

To Reproduce

sudo docker  run --rm -v /root/.docker/config.json:/kaniko/.docker/config.json -v `pwd`:/build-test gcr.io/kaniko-project/executor:latest --dockerfile=Dockerfile --context=dir:///build-test --destination=x:destination --verbosity=debug 

Additional Information

• Kaniko Image (fully qualified with digest)
  gcr.io/kaniko-project/executor:latest (v0.10.0)
  Digest: sha256:78d44ec4e9cb5545d7f85c1924695c89503ded86a59f92c7ae658afa3cff5400

• log

DEBU[0000] Getting source context from dir:///build-test
DEBU[0000] Build context located at /build-test
DEBU[0000] Copying file /build-test/Dockerfile to /kaniko/Dockerfile
INFO[0000] Resolved base name x:y to x:y
INFO[0000] Resolved base name x:y to x:y
INFO[0000] Downloading base image x:y
INFO[0000] Error while retrieving image from cache: getting file info: stat /cache/sha256:xx: no such file or directory
INFO[0000] Downloading base image x:y
error building image: Get https://xxx/v2/xxx: dial tcp x:443: i/o timeout

[drpmma]

I checked the relevant code about retrieving a remote image. And I doubt the error occurs in v0.10.0 because of the changes in vendor "github.com/google/go-containerregistry/pkg/v1/remote" between v0.9.0 and v0.10.0.
Not all the images get i/o timeout error. Retrieving the images in docker hub works fine while the images in my case, downloaded from the company private registry, get the i/o timeout error.

[OliverNocon]

Might even be related to #645.
We had similar issues and the solution from #648 resolved the DNS timeout issues for us ...

[drpmma]

Might even be related to #645.
We had similar issues and the solution from #648 resolved the DNS timeout issues for us ...

Thanks for your reply. The problem has no relationship with golang image version. In fact, it occurs due to the commit dbc4da of go-containerregistry/pkg/v1/remote/transport/bearer.go. It overwrote the redirect scheme, causing failure when requesting an https registry but redirecting to another http downloading URL.

And it is now fixed by commit ec69c4, so it should be fine when updating to the new version of go-containerregistry.

[willinwu]

@drpmma , I get the same error while i use the private registry.
It's strange that I only get error if the vm ( which the kaniko is running on )can't access the internet.
I upgrade the golang to 1.12 and upgrade the kaniko to 0.14.0.
But it doesn't help.
Can you give me some advise?