4

With a private registry, I cannot pull images because certains image layers fail to download. The pattern seems to be that this happens with larger, 50MB sized layers. It shouldn't be a registry neither network problem as colleagues with other systems manage to get these images.

6bbedd9b76a4: Pulling fs layer
fc19d60a83f1: Download complete
de413bb911fd: Download complete
2879a7ad3144: Download complete
668604fde02e: Download complete
f257ac562635: Download complete
07486ea9391a: Download complete
c131a1091dea: Download complete
e4e87f3caf2a: Downloading
6ef05c7ca3e3: Retrying in 1 second
85b7b45e23a1: Download complete
077ff1474b65: Retrying in 1 second
f85147a0e9d7: Waiting

enter image description here

Other errors on new retries with same image:

2c1171c58a199d2ec9daee729d88146dfb91caba4788: local error: tls: bad record MAC

error pulling image configuration: tls: received record with version 5ea7 when expecting version 303

UPD. Strange enough, the very first image layer which always gets stuck before it downloads anything stucks rarely at "downloading... 987B".

UPD2. Same behaviour with Docker EE client (Docker-Client/17.03.0-ee-1)

UPD3. Filed this as a bug to the Moby project.

2
  • 1
    What Docker version are you using? Maybe related: github.com/nats-io/go-nats/issues/269
    – friism
    Jul 12 '17 at 17:29
  • 1
    @friism interesting; so if I would increase TLS timeout how can I do that on Windows 10?! - Docker version 17.06.0-ce, build 02c1d87 Jul 12 '17 at 17:32
1

You may wish to try disabling and uninstalling any antivirus software you have installed and see if it resolves the issue (I of course don't recommend this in the long term). I have seen two different instances of antivirus interrupting SSL conections on Windows 10 recently. Chrome would report a similar error to "bad record MAC" and the issue was generalized to any SSL website or connection and any browser. Unstalling or changing A/V software resolved the issue.

1
  • 1
    Thx - I disabled all a/v services but w/o effect :/ Jul 12 '17 at 21:03
1

This has been a Docker bug caused by jumbo Ethernet frames while using VPN which has now been fixed. Thanks Docker team!

https://github.com/moby/moby/issues/34080

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.