[ClusterLabs] Antw: Re: Antw: Re: bug in crm shell (SLES11 SP4)? required parameters not checked

[Ulrich Windl]

>>> Lars Marowsky-Bree <lmb at suse.com> schrieb am 11.05.2016 um 16:13 in Nachricht
<20160511141305.GI381 at suse.de>:
> On 2016-05-11T08:02:56, Ulrich Windl <Ulrich.Windl at rz.uni-regensburg.de> wrote:
> 
>> > $ crm help Checks
>> > $ crm options help check-frequency
>> > $ crm options help check-mode
>> > 
>> > If none of these settings match Ulrich's preferences, maybe he
>> > could pledge his case to introduce more.
>> 
>> Why do we need this? IMHO a Boolean (check yes/no) is sufficient.
> 
> No, you may not want to see those warnings for every intermediate step -
> the configuration may be complete at the end of all steps, and that's
> when it checks by default.

I understand that there are several levels of validity checks, and it makes sense to do a final check before commit. That does not mean to skip earlier checks, however.
Obviously leaving out a required parameter when defining a primitive is an error that can (and should) be checked as early as possible (if you want error checking).

> 
>> Or did I miss a use case for adding invalid configuration items?
> 
> Yes. That crmsh thinks something is "invalid" doesn't mean that it
> really is invalid within the resource agent (not all metadata is always
> correct, alas).

So crm shell may tag valid configurations invalid while not tagging invalid configurations as such?

Regards,
Ulrich