Puppet Community

Here's the list of facts and which ones are legacy.  <https://www.puppet.com/docs/puppet/8/core_facts.html>

Facter.value(:fqdn)  if i have this in code what should it be changed to ?

I think
```Facter.value(:networking)["fqdn"]```

IIRC isolation-related things were about non-namespaced Ruby functions.. haven’t heard about Puppet DSL.. maybe <#C0W1X7ZAL|> is a bit better match..

<@U11QGB2PN> Why are you asking? There are naturally many differences so there must be something specific you are thinking…

From the language standpoint, a call is a call, and it looks up the name of a function in what is known as the “loaders” - loaders load a function being a 3x, 4x or Puppet DSL based function different ways. What is returned has an API that makes it callable. So, one could say that the language is agnostic in respect of how the function is implemented.

If you are interested in how that works - you can take a look at this slide presentation. <https://docs.google.com/presentation/d/1k_XSovyZM-yN5V8Agi0LfkV52bXryQF_MuEDZAbJDew/edit?usp=sharing>

Not knowing what you are looking for - but if it is isolation related, anything based on Puppet DSL should be perfectly fine, 4x functions should be fine if the rules are followed and 3x functions should be suspected and carefully reviewed since it is way to easy for an implementor of those to make mistakes.

We have an issue where two nodes flip configs. Just two. But the Puppet DSL function is used at other systems without any issues. We have verified that hiera returns the correct data. But the compiler takes different sets of data.

Doubt it is DSL based functions screwing things up for you. 

3d party hiera driver with some kind of cashing? Use of ruby functions that cause things to leak.
Are the facts consistent between runs?

there is probably a package_provider fact you can use

<https://github.com/puppetlabs/puppetlabs-stdlib/blob/main/lib/facter/package_provider.rb|https://github.com/puppetlabs/puppetlabs-stdlib/blob/main/lib/facter/package_provider.rb>

IIRC if you’d like to use the pluginsync you should write some ruby code in one of your modules

that's how it reads to me that using a module and the facter dir in the module will allow it to get picked up by pluginsync

ah..
<https://www.puppet.com/docs/puppet/8/external_facts.html#executable-fact-locations>

as I see pluginsync works for external facts as well :thinking_face:

I'd already written it in Ruby (well modified someone elsess)

then easier to go with custom fact I’d say: <https://www.puppet.com/docs/puppet/8/plugins_in_modules.html#module_plugin_types>
<https://www.puppet.com/docs/puppet/8/custom_facts.html#custom_facts>

that (version 7) was the doc I'd been reading, which seems to suggest the easiest way is to include that fact in a module and push the module out

it's a really basic requirement, I've written a custom fact to get the ipv4 address of the fqdn so that I can reference that with machine that have multiple nic's so pushing that custom fact to the puppet server seems an easy approach

(of course if a module is the best/correct way to do this, I'm happy to do it, it just seemed overkill)

Hi folks, regarding the Puppet Platform release 7.25.0 - as of today, there are no changelogs for the puppetserver 7.12.0 - is this something that is being worked on?

Most likely an error. Release is not listed in jira either <https://tickets.puppetlabs.com/projects/SERVER?selectedItem=com.atlassian.jira.jira-projects-plugin%3Arelease-page&amp;status=released-unreleased>

Maybe I'm misunderstanding some nuance, <@UQ70FTGMC> but something like:
```fact('networking.interfaces.eth0.ip')```

Ah! Now I remember why I started the Puppet VM yesterday! :smile:

I was sure there was some fqdn-related IP address.. though not 100% is it Puppet or SaltStack..

I *guess* `networking.ip` should be the FQDN one..

<@UQ70FTGMC> does the above work for you?

well.. maybe not the FQDN one but default route one

that does return an IP but there is no way to ensure that is the FQDN of the host

especially on devices that are acting eg: as routers or bridges between networks

ah.. so you want to know the IP FQDN is resolving into?

correct, on multi-homed devices I want to expose certain services on the FQDN IP only,

more so as some hosts will have internet connected/routed IP’s

and some will bridge networks, so I need to make sure it’s just returning the FQDN of the host ip

well.. I believe networking.ip is based on the default route..

if it doesn’t work for you then you definitely need another fact

routers/gateways may have specific routing.. so it might not work for you reliably

you also run into problems if you have a split horizon DNS or source based/policy routing or multiple routing tables

networking.ip is fine in 85% of the cases. but always keep your setup in mind

oddly it was a policy based router that first tripped me up

then when I looked at it, there was a dual homed device that I found had assumed the public IP address to run consul on and ignored the internal device, despite the fqdn being the default route, but because there was a second router for 0.0.0.0 which was ‘internet’ traffic it took over

so I ended up with 2 use cases, where the fqdn would have just made life easier

then found a load of use cases on the internet of people running consul for service discovery binding to the wrong interface (because their setup was inconsistent, not a consul module problem)

it felt like an easy way to just make life simple

For VMs with multiple interfaces, do you have a convention for which interface corresponds to FQDN? We typically assign `eth0` as the application interface and `eth1` as the management interface. HTH.

so the device name is not consistent across hardware platforms

(it was one of the things I read on the internet of people complaining about their use of a consul module, it was their environmental problem and inconsistency - but the expectation was that the module would just work it out)

my own problem was triggered by a policy based router having different default routes depending on the source

if you have a single default route I recommend a networkd config thst renames that interface to uplink

I think that would capture a lot of problem usecases…..I may just deploy a module with the fact for fqdn for my policy based routers as that won’t work there

(and fix the random host where the internet route is 0.0.0.0 - that’s my environmental problem)

fantastic response and thoughts, thank you

yep, we did NIC renaming too while ago.. greatly simplifies things

yeah, it’s the routers that screw up, just tested it, all usecases great on networking.ip except for the routers

damn, that would have been a quick and easy win

I’m certain there isn’t but want to make sure I’ve not missed something before raising a PR

You have a space between Class and left bracket. 

This means that you have two expressions, one being just the type Class, and the second an array . The error is telling you that a class body cannot end with a value - the value is produced and isn’t used since a class body does not return a value like a function does.

```class foo() {
Class 2
}```
Has the same effect, first the type `Class` and then this value is thrown away without any effect, then the value `2` is also jsut produced and thrown away.

interesting, I thought class foo () { was right from reading the style guide

To use `[]` as an operator on the Left value it must be adjacent to it. Otherwise it ends up being a literal array.

again let me digest that, I think I've missunderstood it

`Class[foo]` is a reference to the class named `"foo"`. But `Class [foo]` is two values after each other.

There is a space in the gist you posted a link to

it is commented out, but I thought it was this line that was giving you trouble

that line just references a class without doing anything with it

it is like saying
```class foo {
  2
}```


you must do something with that reference

I was expecting that to dictate the execution order

right, included, but that line I believed told it to execute the install class first

you are not telling puppet what you want it to do with that class

let me find an example, as I read the developers guide then picked a basic module to try to work through as an example

eg: <https://github.com/puppetlabs/puppetlabs-ntp/blob/main/manifests/init.pp>

classes are included first from some kind of classification of a node, then those classes include other classes in the catalog (and include resources).

Yeah, but you commented out the line that did something with that reference

the `ntp` module has this at the end:
```Class['ntp::install']
  -&gt; Class['ntp::config']
  ~&gt; Class['ntp::service']```
and that tells puppet that the first class should come before the second, and the second before the third

When you comment out `-&gt; Class[…]` the only thing left is a reference to a class. And that has no effect what so ever, and that is what the error message is telling you.

i.e. the gist shows this
```{
  Class['cfg_sysadmin::install']
  #-&gt; Class ['cfg_sysadmin::config']
  }```
which is just a reference to a class when line 10 is commented out. And line 10 has an error in that there should be no space there:
```{
  Class['cfg_sysadmin::install']
    -&gt; Class['cfg_sysadmin::config']
}```
Should work, given that the second class exists and that you want them to be in that order.

I think I see what you’re saying, but I’m not how a class within a module is included from a node classification, I thought the module was included by classification, which automatically included init.pp which then dictated what classes should be used from the module

typically a module has one class that is defined in its init.pp and that class is named after the module. There can be many classes in a module that can be included from outside the module. Nothin inside a module is really private.

And top level classification typically include that (top) class of a module

ok - that does make sense, albeit different than what I thought I knew

in other words - `init.pp` is just the file that puppet searches for the definition of the class with the same name as the module.

I think I need to step back through some of the docs I’ve read with the context you’ve just given, thank you

Maybe this bloc post of mine will help you <http://puppet-on-the-edge.blogspot.com/2014/04/getting-your-puppet-ducks-in-row.html>

That post was written for Puppet before puppet 4. It is still 99% correct…

certainly was helpful, I've got some knowledge gaps I didn't think I had

<@U111Z4JR1> apologies, for messaging you direct again in this thread, I’m happy to ask the wider group, but I thought as you’d given me context from your medium post, you maybe able to fill a gap I had on this topic

ok, no worries - did you have a question?

<https://gist.github.com/ikonia/3794db6e485a1e699e73adb248a75663>

This is a really dumb question... I need to create a mount, but need the directory to exist before I mount it.  Problem is I need to use another class that requires the mount, but it also manages the directory.  I don't have control over the code for the other class.  Anyone have any suggestions to avoid the duplicate declaration error?

I might be missing something, but could you do this in your class?
```mount { '/mnt/foo':
   ensure =&gt; present,
   &lt;mount_params_go_here&gt;
   require =&gt; File['/mnt/foo'],
}```

• I need to create Directory /xyz
• Mount volume on /xyz
• Run another class, that also manages the file resources for /xyz

The other class also puts stuff in that directory, so I need the mount to happen before it does that, but I can't create the directory because the other class manages it

And the other class is code written by Puppet so I can't modify it

If you can identify the resources in that class that depend on the mount, you might be able to collect them in your class and add dependencies to them:
```Mount['/mnt/foo'] -&gt; Some::Other::Resource&lt;| |&gt;```
But this is something that IME is tricky to do in Puppet. You might be better off trying to find a way to do it in systemd (or equivalent for your OS).

systemd for instance has `RequiresMountsFor` that you can use if you can package this up as a unit - <https://www.freedesktop.org/software/systemd/man/systemd.unit.html#RequiresMountsFor=>

Use Puppet to manage the unit, and then systemd to manage the mount.

I wonder if I could get away with doing an exec

That would be another option, but would be a bit hackier. I've done this myself so can't judge, though

The systemd route would be more robust, especially if this is going to be permanent

I just don't have a lot of experience with systemd outside of starting and stopping services

Ideally that other class would have an option - manage_xyz, which you can turn off

Yeah, It's Puppet Comply, and doesn't seem to have such an option

Submit a ticket, meanwhile - mounttab + exec

This message contains interactive elements.

I think you can combine most of them into `lvm::logical_volume`

<https://github.com/puppetlabs/puppetlabs-lvm/blob/main/manifests/logical_volume.pp#L154>

yeah just lvm::logical_volume -&gt; class

Usually when I'm dealing with LVM I'm just copying code co-workers wrote and update it for me

Well, you didn't say it's lvm at first :slightly_smiling_face:

I guess I was vague when I said volume :smile:

this means you made the certificate and then changed the node's name.

generally speaking, the easiest way to fix this is to remove certificates and start the cert process over

Does anyone have any suggestion on how to fix this?

I have a setup running open source currently, should i install something new to get it to work?

you need a puppetdb, and puppetdb-terminus on your puppetserver

Awesome… will check and give it a shot.

I checked in with the teams and no this is not currently on the road map or prioritized.

If we can get details of who requires this we can make a case if there is a big enough requirement

how can we express interest so it gets prioritized?

every of your Debian 11 users because they want to upgrade their servers

for those with access to the case update with their comment. For those without its good to hear the size of estate and other such details so we can represent scale of demand.

its kinda sad that we users have to go trough this process every time an OS has a major release

do product managers dont know that people like to upgrade/reinstall their servers with a modern operating system

if I'm honest and open I think there's a degree of catchup being played right now with the likes of Ubuntu 22.04 and Rhel 9 taking precedence but I would hope as we catch up we will be more on top of it in future.

I appreciate that's all jam tomorrow.

but I will ping up to our product group. We can't afford to neglect long term

Weighing in on wanting bookworm packages (and ARM versions of that too)

Voted on the ticket as i cannot add a comment. <https://puppet.atlassian.net/jira/software/c/projects/PA/issues/PA-4995>

I hope votes are considered :slightly_smiling_face: Bookworm is not that critical for me as Ubuntu 22.04/arm64 but I voted too :slightly_smiling_face:

if there is any way we as community or I personally can help with speeding up arm64 releases please let me know <@UDDJQQZ38> :pray::skin-tone-2:

I don’t get why Debian is not considered as important as EL/Ubuntu and provided in parallel.

I understand it from the business perspective with limited resources.. no magic here

though difference between Ubuntu and Debian builds are not that big I believe.. anyway, I wish the release team to align all the processes and expectations and catch up with latest OS releases soon!

This should have been automated a long time ago. There is something like configmanagement… :sarc:

I have no idea what is the problem with releases.. but there should be some reason why they didn’t release latest OS support yet.. I don’t think they are happy with this too :man-shrugging::skin-tone-2:

It is automated, but it takes having specific backend systems to test against. There has to be enough usage from the community for us to commit the resources. Unfortunately, what we see is that something like 85+% of our deployments are on RHEL and Ubuntu, with Debian being a very small percentage.

how do you expect debian usage to grow when you treat it like a second class and releases take ages/dont happen at all

Because i suspect the real % of Debian is hidden to you. And not providing packages creates a vicious circle…

containers is another option (you can run an official Puppetserver container on any linux distro).. but there are no official containers now :slightly_smiling_face:

<@U1VFFBZV5> <@UNNE9882X> <@U1PC2RAH0> these are all very good points. Its something I want to understand better

How do we track and monitor usuage in open source community better? (Its something I'd like to understand in the forge and tooling better too)

It's always tricky to get good numbers. You can only measure that for stuff you actually offer and many people disable analytics. So instead of putting a huge effort into analytics I think it makes more sense to make the packaging workflow easier and more open. That will enable the community to contribute

when adding a new OS major version is a nobrainer, it doesn't matter how many people use it.

I want to repeat my claim from PuppetConf 2017: Make your internal pipelines public again (and make it easier to contribute). They were public in the past

We *need* ARM packages for bullseye &amp; bookworm. We can write a tool to package them internally, or we could indeed contribute to a puppet/perforce tooling to do the same. 

maybe ask Lennart from the module directly? Maybe he has some customers with old puppet versions too? i don't know if he is here at slack. :thonk:

See the `icinga-icinga2` module at <https://forge.puppet.com/icinga/icinga2?src=slack&amp;channel=puppet>

:fry_dancing: _Content and Tooling_ is about to start up in <#CFD8Z9A4T|>

:meeting: _:teacher:Puppet Core Team_ is about to start up in <#CFD8Z9A4T|>

I’d recommend you to play with `augtool` and some test XML file to understand it better.

I found few interesting links while googling for “augeas modify xml”

augtool, was my first try.  But the last()+ section is the problem.
`"set foo[last()+1]/#attribute/value 'bar'"`
will this line every run.
The old <https://www.puppet.com/docs/puppet/5.5/resources_augeas.html|documentation> of puppet will shown an solution with "[. =]"
```[. = '/foo']```
But this will not work.
my try was:
`"set foo[. = 'foo']/#attribute/value 'bar'"`
But this will also result in an repeated line every run.