toggle quoted messageShow quoted text
I have mixed feelings on this, Just back in January I migrated a Client from 3.6 to 5 ... On one hand support can be maintained at a cost to the maintainers, mean while pinning versions and revving major versions is perfectly acceptable in my book... I have now worked at 3 different shops using puppet to varying degrees of success, but every one of them have pinned version of the external modules they were using (after learning the hard way first).
At a bare minimum, mirroring the PE LTS EOL dates would be recommended.
On Wed, Mar 21, 2018 at 2:50 AM, Tim Meusel <tim@...>
Puppet 5 is available since a long time. Besides many bug fixes it also
has support for hiera 5 configs as data-in-modules. Some time ago in the
IRC channel we agreed on Puppet 4.7.1 as our lowest supported Puppet
version and at least Stdlib 4.13.1. A few of our modules ship
data-in-modules with hiera 5. This got introduced in Puppet 4.8(?) but
had some bugs. The agreement, and recommendation by Puppet, was to
require at least Puppet 4.10.X in those cases.
I would like to get feedback from the community here. Do you think it is
still needed to support puppet 4 at all? Is 4.10 even to high as the
My point of view is that upgrades, since the AIO packages are available,
are really easy and everybody should be able to regularly update their
nodes. Dropping Puppet 4 would allow us to reduce the test matrix and
speed CI runs up. Also we would be safe with data-in-modules/hiera5. We
also need to keep in mind that not everybody is required to update their
used modules to the latest version, just because they are available.
The trigger for this was
https://github.com/camptocamp/puppet-systemd/pull/73. What are your
opinions on this topic?