Do block until ready for cs_property flush

[bogdando]

No description provided.

[roidelapluie]

we should really check only ONCE for all the flushes .... not one time per type

[bogdando]

@roidelapluie makes sense, thank you

[roidelapluie]

But how could we implement that? cc @ffrank @igalic

[igalic]

@roidelapluie from my limited knowledge, it's practically impossible…
@ffrank should know more (i.e.: why ;)

but the opsview t&p's suffer from a similar issue, making them restart opsview/nagios for every added configuration item…

[ffrank]

This issue is rooted pretty deeply iirc, and we're indeed not the only ones facing it.

I don't think that there is currently a life-cycle hook with a semantics of "do this once all resources of type/with provider X are finished" or even "do this after all resources are synchronized". But this is the kind of situation where we would really like to have that.

I'll do some digging when I get a chance. Would anyone ping me again? I just burned the github notification ;-)

[roidelapluie]

Ping @ffrank

[ffrank]

Thanks @roidelapluie. And guess what, I just found a thing: post_resource_eval could be just adequate.

It is called by the transaction code, so this documentation is not necessarily outdated. I suggest giving that hook a shot.

It's a public API, but please note that as of Puppet 4.3, there does not seem to be a core provider that actually uses it. A quick walk through the git history does not indicate that anything ever did. YMMV ;-)

[bogdando]

Folks, I lost a track of this discussion. What about the suggested change? The block until ready actually is not a resource consuming operation, AFAICT. And it "grants" your changes a safe commit window to make it into the CIB. Thus, it is better to have it than not

[ffrank]

I guess we could merge this and open an issue in order to move the flush functionality to the provider hook instead, later.

In light of this, I also figure we could even neglect test code. Thoughts @igalic, @roidelapluie?

[roidelapluie]

👍

[ffrank]

Tracking this in #173 and merging this.

Thanks for the contribution @bogdando !

[igalic]

\o/