Closed gavanderhoorn closed 6 years ago
Docker repo appears to be added in a similar fashion, but there the Exec[apt_update]
does appear to get executed before it tries to install docker
.
I had a similar problem, which took me some time to figure out what was going on, because I didn't know the puppet. log
file existed.
I supposed it might be failing because it wasn't the latest version in the repository, so my workaround went through checking the existing jenkins
versions with apt-cache policy jenkins
, and setting the new version in the buildfarm deployment configuration file.
I wonder if removing the jenkins version on the config file and just installing the last available version would be a proper fix for this. I guess this could work, at least as long as jenkins is backwards compatible.
Thanks for the comment @inigomartinez.
Did you re-run reconfigure.bash
on the same host after you made the change? Because if you did, I'm wondering if it was the version-spec that you changed that got it to work, or the fact that the re-run of puppet
succeeded as the deb
repository was now available.
2.60.3
should be available from the deb
repository that the Jenkins class installs, so I'm rather confused.
Yes, I ran again the reconfigure.bash
script and it finally installed jenkins0. The last version at the time was 2.73.3
.
I've also been wondering why 2.60.3
wasn't installed, given the fact that is present (along many other versions) in the repository.
[0]: I wasn't aware of #160, so I'll recheck our instance to avoid any problem it might cause.
Also potentially related is #70
@tfoote: yes :) that's why I wrote that this somewhat of a duplicate of #70.
Sorry @gavanderhoorn I didn't scroll back far enough to see you'd already pointed to it. I was responding to @inigomartinez that the package is in the repo, but apt hasn't updated which is likely why it's didn't install the first time.
This is somewhat of a duplicate of #70, but seeing as that was reported against the Trusty version of the buildfarm, I figured I post a new issue to not potentially conflate things.
I seem to be running into the problem reported by the OP in #70 (and myself in https://github.com/ros-infrastructure/buildfarm_deployment/issues/70#issuecomment-301836295) again.
Using
buildfarm_deployment
at 3da95258c079452de52083cc071167ec90bdb15f on a fresh Ubuntu Server 16.04amd64
VM.After invoking
reconfigure.bash master
thepuppet
run fails eventually with this in the log:after this, just about everything is skipped (as expected) as installation of
jenkins
failed.At this point, manually running
apt-get install jenkins=2.60.3
does install it. Runningreconfigure.bash master
a second time also seems to work (again: as I expect: the repository and key have at that point been imported in the previous run, so it should now not be a problem anymore). But because of #160 I'm not sure that is a viable work-around (although it only mentions the "jenkins configuration", which at this point probably doesn't exist yet)./etc/apt/sources.list.d/jenkins.list
containsdeb http://pkg.jenkins-ci.org/debian-stable binary/
, which is correct as far as I can tell.I also notice a
Scheduling refresh of Class[Apt::Update]
in the log output, but could it be that it hasn't been run yet before the installation ofjenkins
is attempted? In that case, this would be https://github.com/jenkinsci/puppet-jenkins/issues/453 again (wrong order in schedule).