thoughts on puppet 2.7 vs 3.0

[Jason Antman]

I currently have ~30 hosts managed by puppet, running on 2.6.14. I'm
just starting a project to build a new puppetmaster in a new datacenter,
which will be our single master. Within a short time, we'll have upwards
of 200 hosts managed by Puppet, and about 25% of these will be "legacy"
OSes (RHEL4 and SLES9) for which I'm building puppet and all of its
dependencies. I'd also like to consider, if not definitely use, puppetDB.

Our change control and QA processes make it very difficult to upgrade
non-application software like Puppet, so there's good odds that whatever
the new master and clients run will be staying for quite some time
(probably years...).

I'd *wanted* to get this rolled out within the next few weeks.

For anyone in the know, given this situation, would you recommend
building on puppet 2.7.18, or trying to draw things out as much as
possible and wait for a stable 3.x release? Does anyone have any general
idea of when one might be forthcoming (weeks? months? next year?)

Thanks for any input,
Jason Antman

[Sandra Schlichting]

Please see this post.

https://groups.google.com/forum/#!topic/puppet-users/8AwzWv2-Qeo

[Jason Antman]

I saw that, thanks. I'm looking for more of an opinion on what route other people would take...

[Jakov Sosic]

On 08/29/2012 06:01 PM, Jason Antman wrote:
> I saw that, thanks. I'm looking for more of an opinion on what route
> other people would take...

Considering the long term lock-in, I would try to postpone deployment
till 3.0 stable gets out.

PS. This new addon of parametrized classes getting info from hiera
auto-magicly really sounds great :)

[Christopher Wood]

In your position I would consider it well worth waiting for puppet 3.

On Wed, Aug 29, 2012 at 09:01:05AM -0700, Jason Antman wrote:
> I saw that, thanks. I'm looking for more of an opinion on what route other
> people would take...
>
> On Wednesday, August 29, 2012 10:57:40 AM UTC-4, Sandra Schlichting wrote:
>
> Please see this post.

> [1]https://groups.google.com/forum/#!topic/puppet-users/8AwzWv2-Qeo
>
> --
> You received this message because you are subscribed to the Google Groups
> "Puppet Users" group.
> To view this discussion on the web visit
> [2]https://groups.google.com/d/msg/puppet-users/-/q5AdWIPVEw8J.
> To post to this group, send email to puppet...@googlegroups.com.
> To unsubscribe from this group, send email to
> puppet-users...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/puppet-users?hl=en.
>
> References
>
> Visible links
> 1. https://groups.google.com/forum/#%21topic/puppet-users/8AwzWv2-Qeo
> 2. https://groups.google.com/d/msg/puppet-users/-/q5AdWIPVEw8J

[Jason Antman]

Chris & Jakov,

That's my tentative feeling right now, but given the way things work
around here, and that puppet is really a lower-level project (i.e.
experimentation, failure, and bugs aren't tolerated well because we only
have management support for the end result, not puppet itself) I'll
probably have to wait for community adoption of 3.0 and then see how it
goes. A slightly complicating factor is that I need RHEL4 and SLES9... I
just built ruby-1.8.7 and deps and haven't tested puppet 2.7.18 on them
yet, so the 2.0 switch might be a bit of a pain...

Thanks for the input.

-Jason

[Jakov Sosic]

On 08/29/2012 07:34 PM, Jason Antman wrote:
> but given the way things work
> around here, and that puppet is really a lower-level project (i.e.
> experimentation, failure, and bugs aren't tolerated well because we only
> have management support for the end result, not puppet itself) I'll

I don't see why the fuss around it, even if puppet itself breaks, it
shouldn't affect anything else on the system.

Also, you should consider having dev/test environments if your
management is so itchy :)

[John Warburton]

On 30 August 2012 00:52, Jason Antman <ja...@jasonantman.com> wrote:

For anyone in the know, given this situation, would you recommend building on puppet 2.7.18, or trying to draw things out as much as possible and wait for a stable 3.x release? Does anyone have any general idea of when one might be forthcoming (weeks? months? next year?)

I'm in a similar position with 1000 hosts, 100 odd modules, 2K+ resources managed by puppet on 2.6.x. Given the deprecation warnings, and changes in 2.7, then 3.x, I'm using 2.7 both as a stepping stone to 3 (code cleanup) and added functionality (puppet DB)

Seeing how 2.7 wasn't suitable *for me* until 2.7.10 due to some specific bugs, I think you should move to 2.7 unless you have a simple setup and prepared to wait for 3


John

[Peter Brown]

From what I have seen there is a big difference between 2.6+ and 3.0+.
I would be setting up a completely separate environment with 2.7 to
test if your code works.
2.7 apparently let you know which things will stop working in 3.0 so
it will be a better idea to use that in the dev environment and iron
out the obvious things first and then switch to 3.0+
There will likely be a bunch of things you will need to fix with your
code before you switch your environment to 3.0 so by the time it's
"stable" enough you will have your code ported and ready to roll out.

I am currently running 2.7+ and plan on setting up my dev environment
to 3.0 once i finish a few more important modules i need to write for
a big project. (yes i write most of my modules from scratch)

Just my opinion but it seems like the best way to make sure everything
works and avoid breaking your production environment and spending way
too long fixing it.
Hope that helps.

Pete.

> --
> You received this message because you are subscribed to the Google Groups
> "Puppet Users" group.

[Jason Antman]

Agreed on both. If puppet itself breaks, yes, it shouldn't affect
anything else on the system. But if I spend the time building puppet
X.X.X for the ~50 legacy hosts on unsupported systems, roll it out, and
find out that it's got a bug that effects us (like... the daemon locking
up) *after* the rollout is "complete", my head might be on a platter.
And dev/test environments, yes, we have them (albeit small), but I'm
battling a "get it done but don't spend too much time on it" directive.

[Jason Antman]

I think that's my new theory. For now, I think I'll be building and
testing a 2.7.(18|19) master and clients. The master will be
bootstrappable though puppet itself (I hope...). Assuming all goes well,
when I'm ready to roll out I'll re-assess the situation with 3.0 and
consider, if things look good, starting the build/test/validate loop
with 3.0.

Thanks for all the input.
-Jason

[Eric Sorenson]

For what it's worth, we on the puppetlabs open-source team are highly motivated to work with and fix bugs for people who are willing to beta-test the RCs of 3.0 - especially if it's something you're headed to production with. On the flip side, if there's something that doesn't work for you in 2.7, it's going to get prioritized below anybody else's 3.0 bugs. 

But I totally understand if the effort/reward equation doesn't work out for you.

-=Eric

[Ryan Coleman]

On Wed, Aug 29, 2012 at 7:52 AM, Jason Antman <ja...@jasonantman.com> wrote:
> For anyone in the know, given this situation, would you recommend building
> on puppet 2.7.18, or trying to draw things out as much as possible and wait
> for a stable 3.x release?

I'm clearly biased but data-code separation is _much_ nicer in 3.0. If
you think you'll be on that major release for years, that element
alone is worth waiting for.