Hi Everybody,

Quick feedback on the recent security release.

While we successfully managed to publish everything in time, I must admit that 
it was more painful than I expected.

So what happened?

Yesterday, as part of the release procedure we trigger the maven release job 
but we quickly got hit by timeout errors between the Jenkins agent and the 
controller. As we faced the same issue on infra.ci last week, we had an idea on 
how to mitigate that issue and  with the help of Gareth and Damien we quickly 
apply the same mitigation that we applied to infra.ci which was switching from 
using a Jenkins tunnel  to a WebSocket connection as defined in this PR 
<https://github.com/jenkins-infra/charts/pull/1030/files> so we could finish 
the first release part


Today we met again with Daniel Beck to finalize the release, so we build and 
publish packages but this time we got hit by two additional issues.
First, we had WebSockets connection timeout error that we solved by increasing 
that value from 30sec to 60sec as defined in this PR 
<https://github.com/jenkins-infra/charts/pull/1032/files>

The second issue that hit us was that we couldn't upload windows artifacts from 
the Jenkins agent to pkg.jenkins.io using the ssh-agent. The problem seems to 
be related to the latest ssh-agent plugin version which deleted non-exec based 
agent factories according to the changelog 
<https://github.com/jenkinsci/ssh-agent-plugin/releases/tag/ssh-agent-1.22> :p.

Thanks to Gareth recent work on building a custom Jenkins image with our 
plugins   as I could quickly pin down the previous version in this PR 
<https://github.com/jenkins-infra/docker-jenkins-lts/pull/232> 

So we first focus on publishing packages for non-Windows distributions, then I 
published the windows ones.

In this process, I also identified additional issues.
1) Our monitoring didn't detect that the latest Jenkins version wasn't 
available from get,jenkins.io, which seems to be regression.
2) I had to manually trigger mirrrobits mirror scan to enable them, I don't 
know yet why it didn't automatically scan mirrors.

So what's next? 
Next week we'll update the AKS cluster used by the release environment hoping 
that it will solve all the network issues we currently have.
Regarding the ssh-agent issue with the windows containers, I am looking for 
someone with better windows skills than mine :) to spare me some debugging 
time. 


Cheers,

Olivier

-- 
You received this message because you are subscribed to the Google Groups 
"Jenkins Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jenkinsci-dev/1782f6e3-09bf-4dce-85fc-c15cd7372de0%40www.fastmail.com.

Reply via email to