[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: CI: Issues with networking on some platforms



Hi Paul,

On 2023-11-04 13:40:08 +0100, Paul Gevers wrote:
On 04-11-2023 12:01, Peter Wienemann wrote:
I have rewritten the autopkgtest of the Charliecloud package [0] and uploaded the changes on November 1, 2023. The salsa CI seems to be happy [1] but the jobs on ci.debian.net fail on some platforms [2]. Checking the logs it seems that the failed jobs were caused by a lack of network connectivity.

All failures on amd64 are on ci-worker13, which is a very powerful host [1]:
CPU(s):                          64
Total online memory:     256G

I rather suspect that something in your test can't handle that, as I'm not aware of other tests having network issues from that host.

thanks for your feedback. I slightly modified the tests to get more meaningful output. The outcome is interesting:

Now image building is only tested with Charliecloud's own builder (ch-image) and docker. Building with ch-image works without issues while building with docker fails. To be more precise: The Dockerfile line

dnf install -y --setopt=install_weak_deps=false openssh-clients && dnf clean all

yields the output

--------------------------------------------------------------------------
430s Errors during downloading metadata for repository 'baseos':
430s - Curl error (28): Timeout was reached for https://mirrors.almalinux.org/mirrorlist/8/baseos [Operation timed out after 30000 milliseconds with 0 out of 0 bytes received] 430s Error: Failed to download metadata for repo 'baseos': Cannot prepare internal mirrorlist: Curl error (28): Timeout was reached for https://mirrors.almalinux.org/mirrorlist/8/baseos [Operation timed out after 30000 milliseconds with 0 out of 0 bytes received]
--------------------------------------------------------------------------

when trying to build with docker while it works for ch-image. The docker build was tried five times with 60 seconds delay between each attempt. The result is always a timeout. So there is obviously some networking issue. What is weird is that

1. it only happens for the docker builder and
2. other nodes like the salsa CI runners do not show this behaviour.

Does anyone have an explanation for this phenomenon?

Best regards,

Peter


Reply to: