Unable to connect to azure.archive.ubuntu.com flaky failure on ubuntu-latest

[traversaro]

Describe the bug
On some jobs (apparently in a non-deterministic way) commands such as apt-get install fail with the following error:

Get:1 http://ppa.launchpad.net/ubuntu-toolchain-r/test/ubuntu bionic/main amd64 gdb amd64 8.2-0ubuntu1~18.04 [3024 kB]
Get:2 http://ppa.launchpad.net/ubuntu-toolchain-r/test/ubuntu bionic/main amd64 gdbserver amd64 8.2-0ubuntu1~18.04 [292 kB]
Get:3 http://ppa.launchpad.net/ondrej/php/ubuntu bionic/main amd64 libxml2-dev amd64 2.9.10+dfsg-2+ubuntu18.04.1+deb.sury.org+1 [821 kB]
Get:4 http://ppa.launchpad.net/ondrej/php/ubuntu bionic/main amd64 libxml2 amd64 2.9.10+dfsg-2+ubuntu18.04.1+deb.sury.org+1 [726 kB]
Get:5 http://ppa.launchpad.net/ondrej/php/ubuntu bionic/main amd64 libgraphicsmagick-q16-3 amd64 1.3.30+hg15796-1+ubuntu18.04.1+deb.sury.org+2 [1181 kB]
Get:6 http://ppa.launchpad.net/ondrej/php/ubuntu bionic/main amd64 libgraphicsmagick++-q16-12 amd64 1.3.30+hg15796-1+ubuntu18.04.1+deb.sury.org+2 [144 kB]
Err:7 http://azure.archive.ubuntu.com/ubuntu bionic/main amd64 libdouble-conversion1 amd64 2.0.1-4ubuntu1
  Could not connect to azure.archive.ubuntu.com:80 (52.177.174.250), connection timed out
Ign:8 http://azure.archive.ubuntu.com/ubuntu bionic-updates/main amd64 libqt5core5a amd64 5.9.5+dfsg-0ubuntu2.5
Ign:9 http://azure.archive.ubuntu.com/ubuntu bionic-updates/main amd64 libwayland-server0 amd64 1.16.0-1ubuntu1.1~18.04.3
Err:10 http://azure.archive.ubuntu.com/ubuntu bionic-updates/main amd64 libgbm1 amd64 19.2.8-0ubuntu0~18.04.3
  Unable to connect to azure.archive.ubuntu.com:http:
Err:11 http://azure.archive.ubuntu.com/ubuntu bionic-updates/main amd64 libxcb-xfixes0 amd64 1.13-2~ubuntu18.04
  Unable to connect to azure.archive.ubuntu.com:http:
Err:12 http://azure.archive.ubuntu.com/ubuntu bionic-updates/main amd64 libegl-mesa0 amd64 19.2.8-0ubuntu0~18.04.3
  Unable to connect to azure.archive.ubuntu.com:http:
Err:13 http://azure.archive.ubuntu.com/ubuntu bionic-updates/main amd64 libegl1 amd64 1.0.0-2ubuntu2.3

For an example of such failure, check https://github.com/robotology/idyntree/pull/668/checks?check_run_id=560084383 . I would not be too surprised about CI jobs failing for networks problems, but this specific issue seems to be extremely frequent. I did not collect precise data on how much frequently this happens, but my impression that it happens in the order of magnitude of 1 out of 10 builds.

This issue is already being discussed on GitHub Community Forum at https://github.community/t5/GitHub-Actions/Install-dependencies-in-Ubuntu-flakes-several-times-a-wekk/td-p/51785 , but the GitHub support suggested me to open an issue here as well.

Area for Triage:

Question, Bug, or Feature?:
Bug

Virtual environments affected

• macOS 10.15
• Ubuntu 16.04 LTS
• Ubuntu 18.04 LTS
• Windows Server 2016 R2
• Windows Server 2019

Expected behavior
I would expect commands such as apt-get install to work correctly.

Actual behavior
It is not clear what is triggering this behavior, but sometimes apt-get install fails.

[MSP-Greg]

I added -o Acquire::Retries=3, and today it worked, then failed for a while, then starting working again.

It may not be a GitHub issue, but they (and especially MSFT) have the resources to mirror azure.archive.ubuntu.com so Actions CI is reliable...

[boegel]

We've been seeing issues like this fairly frequently since Sunday March 29th...

[boegel]

Using -o Acquire::Retries=3 is not helping for us, see for example https://github.com/easybuilders/easybuild-easyblocks/pull/2012/checks?check_run_id=563883264

[MCOfficer]

I mentioned this on the github community post, but you can probably use apt-spy2 to find a working mirror before using apt.

[MSP-Greg]

@boegel

Sorry, I didn't mean to imply that using -o Acquire::Retries=3 was a fix. What I meant was that if one is using it and apt-get still times out, there is really a problem.

[boegel]

I mentioned this on the github community post, but you can probably use apt-spy2 to find a working mirror before using apt.

That adds a couple of minutes to each run, but it seems to work well to circumvent the problems, thanks a lot!

For other, see for example easybuilders/easybuild-easyconfigs#10341

[znarf]

On our side, we get 503 errors on almost all runs past few hours:

E: Failed to fetch http://azure.archive.ubuntu.com/ubuntu/pool/universe/f/fonts-arphic-bkai00mp/fonts-arphic-bkai00mp_2.10-17_all.deb  503  Service Unavailable [IP: 52.177.174.250 80]

[deivid-rodriguez]

I added this to my workflow before apt-get install as a workaround:

sudo sed -i 's/azure\.//' /etc/apt/sources.list
sudo apt-get update

[Betree]

We used a different approach to work around this issue:

steps:
  - name: Hack sources.list
    run: sudo sed -i 's|http://azure.archive.ubuntu.com/ubuntu/|http://mirror.arizona.edu/ubuntu/|g' /etc/apt/sources.list
  - ...

You can pick any mirror from https://launchpad.net/ubuntu/+archivemirrors, ideally one in the US since Github Actions servers are in the US.

[ddelange]

hard-coding a specific mirror will most likely overload their server if too many people pick it. poor universities 😅

[Spenhouet]

Adding apt-spy2 did not do the trick for me.

The following packages have unmet dependencies:
 libegl-mesa0 : Depends: libgbm1 (= 22.0.5-0ubuntu0.1) but 22.0.5-0ubuntu0.3 is to be installed
                Depends: libglapi-mesa (= 22.0.5-0ubuntu0.1) but 22.0.5-0ubuntu0.3 is to be installed
E: Unable to correct problems, you have held broken packages.
Failed to install browsers
Error: Installation process exited with code: 100

[ddelange]

switching to launchpad (my snippet above) fixed that one for us ^

[ClaudiaKolesnikova]

Adding apt-spy2 did not do the trick for me.

The following packages have unmet dependencies:
 libegl-mesa0 : Depends: libgbm1 (= 22.0.5-0ubuntu0.1) but 22.0.5-0ubuntu0.3 is to be installed
                Depends: libglapi-mesa (= 22.0.5-0ubuntu0.1) but 22.0.5-0ubuntu0.3 is to be installed
E: Unable to correct problems, you have held broken packages.
Failed to install browsers
Error: Installation process exited with code: 100

hi, I have the same error, did you fix that?

[Spenhouet]

sudo gem install apt-spy2
sudo apt-spy2 fix --commit --launchpad --country=US
sudo apt-get update

@ClaudiaKolesnikova Yes, I did this, suggested by @ddelange.

[aethernet]

Same issue for me, and simply running an apt-get update before installing anything fixed it.

[hloeung]

Same issue for me, and simply running an apt-get update before installing anything fixed it.

Not an "issue", it's what you're meant to do.

When images used are built, they're bundled with a package index. Packages are constantly updated with various security fixes and such and thus indexes can be out-of-date. apt-get update is required to pull in the latest package indexes.