On Tue, Jul 21, 2015 at 12:40 PM, Branan Riley <bra...@puppetlabs.com> wrote:
> <This is email 2/2 inspired by PUP-4813> > > With puppet 4, we made a change to how prefetch failures are handled > by the transaction[1]. This is actually a pretty good change, as it > prevents puppet from some nasty misbehaviors. > > I still don't think it's the "correct" behavior, though. Aborting the > entire catalog because one resource type is broken doesn't seem like > the best behavior, especially when it comes to reporting[2] > > I feel like the correct behavior is something like this: > > * If an exception escapes a prefetch, mark all resources of that type > in the catalog as failed and do not try to apply them > Background for a question: Wrt exceptions in prefetch, as of 4.0 there is a distinction between the expected/supported exceptions (there are two: Puppet::MissingCommand and LoadError) and everything else. I'm not sure this is *super* well documented, but it is mentioned as a breaking change in the 4.0 release notes: https://docs.puppetlabs.com/puppet/4.0/reference/release_notes.html#stuff-a-lot-of-people-will-notice and the rescue in question is here: https://github.com/puppetlabs/puppet/blob/810624fe6f88d520339173c0e811c5ede4009fac/lib/puppet/transaction.rb#L322 So with that in mind, is the proposal here about the handling of the two supported exceptions or the handling of other exceptions? For the two supported exceptions, this seems like a reasonable behavior change and could also speed up catalog application in this failure case. For *other* exceptions, I'm not sure. I walked away from PUP-3656 thinking that those two exceptions were now part of the contract we're specifying for providers. If so, a provider that throws any *other* exception has a bug which should be fixed, and we could be making matters worse by trying to soldier on. I.e. this seems like a situation where it's better to fail hard and the user should file a bug, upgrade/downgrade the module, etc. Btw (and perhaps I should have asked this from the get-go before opining above): what was the exception in PUP-4813 that precipitated this thread? > * Make it possible to fail resources from *within* a prefetch. This > lets a sufficiently smart prefetch handle the case when only some of > its resources are hosed (parsedfile resources with multiple targets, > for example, might be only partially unfetchable) > How would this work? A new hook for providers to implement? If so, it'd be nice to flesh out a couple examples use cases so that we can think through the API with something concrete. > If there's agreement on this I'll file the appropriate tickets. If > there are other ideas of how we should handle prefetch failures I'd > also love to hear them. > Thanks for bringing these issues up! Sometimes innocuous looking tickets bloom into huge "how did this ever work?" conversations and this is kinda one of those :) Kylo > > Thanks, > Branan Riley > Puppet Labs Software Engineer > > [1] PUP-3656 > [2] See https://groups.google.com/forum/#!topic/puppet-dev/QMTJli2oabc > > -- > You received this message because you are subscribed to the Google Groups > "Puppet Developers" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to puppet-dev+unsubscr...@googlegroups.com. > To view this discussion on the web visit > https://groups.google.com/d/msgid/puppet-dev/CADWDnrmDGO%3DR__fc0YxB4ywcGAbcOnASMKBw%2BPdPusGoG_nbog%40mail.gmail.com > <https://groups.google.com/d/msgid/puppet-dev/CADWDnrmDGO%3DR__fc0YxB4ywcGAbcOnASMKBw%2BPdPusGoG_nbog%40mail.gmail.com?utm_medium=email&utm_source=footer> > . > For more options, visit https://groups.google.com/d/optout. > -- Kylo Ginsberg | k...@puppetlabs.com | irc: kylo | twitter: @kylog *PuppetConf 2015 <http://2015.puppetconf.com/> is coming to Portland, Oregon! Join us October 5-9.* *Register now to take advantage of the Early Bird discount <https://www.eventbrite.com/e/puppetconf-2015-october-5-9-tickets-13115894995?discount=EarlyBird> * *—**save $249!* -- You received this message because you are subscribed to the Google Groups "Puppet Developers" group. To unsubscribe from this group and stop receiving emails from it, send an email to puppet-dev+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/puppet-dev/CALsUZFHGNJH1fc1yy%3D8-J-5frY6JqRUHc%2BOe7w30%2B062K%2Bm%2Bjg%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
