On Thursday, May 19, 2016 at 4:39:03 AM UTC-5, James Denton wrote:
>
> Hi Greg
>
> Thanks for the reply, its strange that this is fairly intermittent (works
> for a couple of hours then fails with the error). I have other puppet
> classes that also inherit from a base class but these do not fail. The
> differance between this class and the others that work is that the class
> show above declares parameters before inheriting the base class.
>
> "Worth double-checking that the ENC data doesn't change when the error
> occurs (if you can catch it)"
>
> How can I do this (if possible)?
>
> Thanks!
>
> On Tuesday, May 17, 2016 at 5:47:37 PM UTC+1, Greg Sutcliffe wrote:
>>
>> It's almost certainly a problem within your puppet code. Foreman only
>> sends a list of classes (and parameters, together called the ENC data) to
>> Puppet, which then deals with compiling the catalog, eg:
>>
>> classes:
>> apache:
>> ssh:
>> loglevel: debug
>> syslog:
>>
>> And so forth. That's generally pretty static; in which case the issue is
>> likely that what Puppet is evaluating for the ENC data supplied by Foreman
>> is changing.
>>
>> Worth double-checking that the ENC data doesn't change when the error
>> occurs (if you can catch it), at which point you can rule out Foreman
>> entirely.
>>
>> Greg
>>
>
I am experiencing very similar issues.
In my case I am using an approved module from puppet forge, saz-rsyslog.
And I am not writing any declarations of any kind. Everything is done in
The Foreman.
I have had a detailed discussion on the puppet-users mail list about my
issue:
Issue with module from forge - Duplicate declaration Class[Rsyslog]
https://groups.google.com/d/topic/puppet-users/0JHrMGuo8YQ/discussion
We came to a point in this discussion thread that first points at an issue
with the saz-rsyslog module, but also *may* point at a possible issue with
The Foreman.
Here is the issue I submitted with the saz-rsyslog forge module:
Duplicate Declaration Class[Rsyslog] with The Foreman as ENC during catalog
compilation #237
https://github.com/saz/puppet-rsyslog/issues/237
The *possible* issue with The Foreman is that under certain and uncommon
circumstances The Foreman can sometimes build a catalog in which subclasses
can be declared before their parent class. With a some kind of unrelated
fact/parameter change in The Foreman, or after just waiting, this issue
will resolve itself.
In this thread you are discussing an intentional declaration.
But in my issue (see my puppet-user thread) the declaration is
unintentional.
And we think that under uncommon circumstances we cannot determine, The
Foreman as ENC causes the rsyslog::client subclass to be declared before
the rsyslog class.
Not only is it unintentional, but my specific environment (1 primary and 3
secondary masters puppet servers with The Foreman Proxy) shows that it can
happen on one secondary-node, and not at all on another secondary-master.
And then hours later this scenario can flip. And at the same time a third
secondary-master puppet server under The Foreman Proxy server does not ever
experience the issue.
The issue is never experienced with the primary-master puppet server, to
which is also the primary The Foreman server.
So perhaps it points to an issue with a The Foreman Proxy server? ... I do
not know.
Having subclasses declared before the parent class seems to not be an issue
with many other puppet modules. So that points at an issue with the
saz-rsyslog module - to which I submitted an issue with the puppet-module
author.
However, looking back at The Foreman, I wonder if it is not the intention
that The Foreman as ENC would under these undetermined circumstances
declare a subclass before the parent class, and then 8-12 hours later
change to have the parent class declared before the subclass. Something is
causing that subclass to be declared first in the catalog, and that cause
may be independent to The Foreman alone - after tracing out the results I
have experienced.
Keep in mind, if we know this is the issue, I have installed a The Foreman
Proxy server with secondary-master puppet servers that NEVER have this
issue. Would this suggest that the issue is with my configuration, and not
internally to The Foreman? And if so, where would I look to track down this
issue?
If this is not a configuration issue, is this an allowed or intended action
of The Foreman as I have described - seemingly randomly ordering
declaration of classes and subclasses?
I appreciate any leads to help track down this issue I have been
experiencing for many months.
-RG
--
You received this message because you are subscribed to the Google Groups
"Foreman users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to foreman-users+unsubscr...@googlegroups.com.
To post to this group, send email to foreman-users@googlegroups.com.
Visit this group at
