AzDO Networking issue impacting multiple builds

[missymessa]

Issue for tracking the intermittent, inconsistent networking errors we're encountering in our builds.

https://portal.microsofticm.com/imp/v3/incidents/details/292951370/home

{
   "errorMessage" : "net/http: request canceled while waiting for connection"
}

Report

Build	Definition	Step Name	Console log
17054	dotnet/roslyn	Initialize containers	Log

Summary

24-Hour Hit Count	7-Day Hit Count	1-Month Count
0	0	1

[ilyas1974]

I have created Azure support ticket 2203110010001681 to try and help address this.

[AraHaan]

Is there perhaps plans for a public version of the link above so we have a general idea on what the cause of the issue might be?

[AraHaan]

good news it seems my pr is now unblocked by this: https://github.com/dotnet/arcade/pull/8604

[markwilkie]

Hi @AraHaan - I don't think there's a way to provide public access to our internal issue tracking system. :( In this case, the title kinda says it all. We're having intermittent issues w/ connectivity and the Azure folks are trying to get to the bottom of it.

Great to see your PRs seems to have worked!

@adiaaida and/or @mmitche - would one of you mind taking a look at the PR? (looks good to me)

[michellemcdaniel]

I don't have context on the file being changed, so hopefully Matt can look?

[ilonatommy]

I am adding some new cases affected by this error.

1. Installer Build and Test coreclr Linux_arm64 Release, Pipelines - Run 20220314.2. error:

docker run -v /mnt/vss/_work/1/s:/root/runtime -w=/root/runtime -e VSS_NUGET_URI_PREFIXES -e VSS_NUGET_ACCESSTOKEN mcr.microsoft.com/dotnet-buildtools/prereqs:rhel-7-rpmpkg-c982313-20174116044113 ./build.sh --ci --subset packs.installers /p:BuildRpmPackage=true /p:Configuration=Release /p:TargetOS=Linux /p:TargetArchitecture=arm64 /p:RuntimeFlavor=coreclr /p:RuntimeArtifactsPath=/root/runtime/artifacts/transport/coreclr /p:RuntimeConfiguration=release /p:LibrariesConfiguration=Release /bl:artifacts/log/Release/msbuild.rpm.installers.binlog 
Unable to find image 'mcr.microsoft.com/dotnet-buildtools/prereqs:rhel-7-rpmpkg-c982313-20174116044113' locally 
docker: Error response from daemon: Get "https://mcr.microsoft.com/v2/": net/http: request canceled while waiting for connection

2. Installer Build and Test coreclr Linux_musl_x64 Release, Pipelines - Run 20220314.22, Mono Product Build Linux x64 debug Run 20220316.68 error:

docker: error pulling image configuration: Get "https://westus2.data.mcr.microsoft.com/01031d61e1024861afee5d512651eb9f-h36fskt2ei//docker/registry/v2/blobs/sha256/d3/d3358c58cff96d0874e70d9ef680e5c69a452079d7d651f9e441c48b62a95144/data?se=2022-03-14T18%3A52%3A56Z&sig=7CM6Q6E1lL%2F07ifd%2FR1VVO%2BRlBbCH%2FiCs8V%2Fki%2BvxXE%3D&sp=r&spr=https&sr=b&sv=2016-05-31&regid=01031d61e1024861afee5d512651eb9f": dial tcp 131.253.33.219:443: i/o timeout. 

3. Build Android arm Release AllSubsets_Mono, Pipelines - Run 20220314.1, Build Browser wasm Linux Release LibraryTests_EAT, Pipelines - Run 20220315.4, Build Linux x64 Release AllSubsets_Mono_LLVMJIT Run 20220316.68. error:

Error response from daemon: Get "[https://mcr.microsoft.com/v2/"](https://mcr.microsoft.com/v2/%22): net/http: request canceled while waiting for connection (Client.Timeout exceeded while awaiting headers)
##[error]Docker pull failed with exit code 1

[dougbu]

Are we tracking the numerous MCR download failures here or somewhere else❔ That problem seems to be getting worse instead of better e.g.

• https://dev.azure.com/dnceng/internal/_build/results?buildId=1664939&view=results
• https://dev.azure.com/dnceng/internal/_build/results?buildId=1665084&view=results
• https://dev.azure.com/dnceng/internal/_build/results?buildId=1665221&view=results
• https://dev.azure.com/dnceng/internal/_build/results?buildId=1665434&view=results

[ilyas1974]

Per information from @agocke , there is a problem with the CDN behind MCR.

[garath]

Per information from, there is a problem with the CDN behind MCR.

Tracking in 295259702

[ilyas1974]

After doing some additional research, we could not find any recent instances of this error. Should it occur again, we will open another issue with the MCR team

[AraHaan]

@ilyas1974 https://github.com/dotnet/roslyn/pull/56162/checks?check_run_id=5776024591

[AlitzelMendez]

reopening as new instance of this issues are happening such as:

• https://dev.azure.com/dnceng/public/_build/results?buildId=1752769&view=logs&j=174348cd-3455-59d8-c7f7-32c969b0807d&t=eda8acff-ca02-497c-b15e-87884972117b&l=29

[markwilkie]

Do y'all think this should be a known build error, or marked as critical? My impression is that the hit count is low enough that it doesn't meet the critical bar....but not sure if we have a bar yet.... :) Thoughts @ilyas1974 ?

[ilyas1974]

We set a bar of 200 jobs impacted before we engage a partner team (metric review back in May).

[ilyas1974]

As there have not been any instances of this issue for the last 7 days, I am closing this issue.