Likewise A. I was part of the initial discussion earlier yesterday, and to me this felt like the most pragmatic thing to do under the circumstances.
On Tue, Sep 17, 2013 at 10:51 AM, Andy Parker <[email protected]> wrote: > On Tue, Sep 17, 2013 at 7:26 AM, Rob Reynolds <[email protected]> wrote: > >> So far it sounds like option A, but I think there was a thought started >> here that you could just order everything so that the reboot occurs last. >> Then if there is an issue along the way, the next run (provided it doesn't >> also error) will cause the reboot to happen. >> >> Which could technically meet both approaches. But perhaps we could add a >> `continue_on_catalog_error => true` as default. Then someone can make that >> decision themselves. >> >> Thoughts? >> >> > I think go for A, and leave out the property. Option A is the normal > puppet semantics (the resource didn't fail, other unrelated things did). If > it turns out that there are actual cases that users reach where they need > to stop the reboot because unrelated problems happen, then we can add in > the feature, but I think we should take a bit of an opinionated stance at > first and modify that as real uses are found. > > >> >> On Tue, Sep 17, 2013 at 2:10 AM, Josh Cooper <[email protected]> wrote: >> >>> >>> >>> >>> On Mon, Sep 16, 2013 at 4:19 PM, badgerious <[email protected]>wrote: >>> >>>> If we support this functionality and there is a failure during the >>>>> catalog run after a reboot at the end has been requested, what would be >>>>> your expectation for the system. >>>>> >>>> >>>>> Would you expect it to: >>>>> >>>> >>>>> A) still reboot >>>>> >>>> B) not reboot >>>>> >>>> C) something else (please comment) >>>> >>>> >>>> I vote A. As I understand it, skipping the reboot means throwing it >>>> into the oblivion; there's no way for the next puppet run to pick the >>>> reboot up because the event triggering it probably won't occur again. >>>> >>> >>> If a notify/subscribe relationship causes a reboot to be scheduled, but >>> is not performed due some other resource failing, then the event will be >>> lost. This is a general problem with events in puppet actually[1]. >>> >>> If puppet detected that a reboot is pending, e.g. >>> HKLM\SYSTEM\CurrentControlSet\Control\Session >>> Manager\PendingFileRenameOperations, and a reboot is scheduled, but not >>> performed for same reasons as above, then puppet should reschedule the >>> reboot the next time it runs (assuming the catalog remains the same). >>> >>> >>>> I've had plenty of small, inconsequential failures and wouldn't want >>>> them affecting other (reboot requiring) things. >>>> >>>> It may be in a state that won't boot cleanly due to the failing half >>>>> run catalog. >>>> >>>> >>>> I would think that a sequence of steps that may result in an unbootable >>>> system should be ordered such that the last item is the reboot, which would >>>> get around that issue. >>>> >>>> Eric >>>> >>>>> -- >>>> 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 [email protected]. >>>> To post to this group, send email to [email protected]. >>>> Visit this group at http://groups.google.com/group/puppet-dev. >>>> For more options, visit https://groups.google.com/groups/opt_out. >>>> >>> >>> So it sounds like there is agreement on option A? >>> >>> Josh >>> >>> [1] http://projects.puppetlabs.com/issues/3806 >>> >>> -- >>> Josh Cooper >>> Developer, Puppet Labs >>> >>> -- >>> 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 [email protected]. >>> To post to this group, send email to [email protected]. >>> Visit this group at http://groups.google.com/group/puppet-dev. >>> For more options, visit https://groups.google.com/groups/opt_out. >>> >> >> >> >> -- >> Rob Reynolds >> Developer, Puppet Labs >> >> Join us at PuppetConf 2014, September 23-24 in San Francisco >> >> -- >> 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 [email protected]. >> To post to this group, send email to [email protected]. >> Visit this group at http://groups.google.com/group/puppet-dev. >> For more options, visit https://groups.google.com/groups/opt_out. >> > > > > -- > Andrew Parker > [email protected] > Freenode: zaphod42 > Twitter: @aparker42 > Software Developer > > *Join us at PuppetConf 2014, September 23-24 in San Francisco* > > -- > 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 [email protected]. > To post to this group, send email to [email protected]. > Visit this group at http://groups.google.com/group/puppet-dev. > For more options, visit https://groups.google.com/groups/opt_out. > -- -- Ethan Brown [email protected] Software Engineer *Join us at PuppetConf 2014, September 23-24 in San Francisco* -- 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 [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/puppet-dev. For more options, visit https://groups.google.com/groups/opt_out.
