On Oct 22, 2009, at 9:50 AM, LdvT wrote: > > On Oct 20, 11:59 pm, Luke Kanies <l...@madstop.com> wrote: >> On Oct 19, 2009, at 7:56 AM, LdvT wrote: >> >> >> >>> I wouldn't mind having failure detection. >> >>> Say somehow a bad copy of a .conf file is distributed; and the >>> service >>> fails to start. >>> If that can be detected and used as a condition, puppet could than >>> call up another manifest to drop in a default/backup copy of >>> the .conf >>> file and restart the service almost immediately. >>> Toss in a notify function on that manifest, and it would be a very >>> handy feature for making changes on diverse production systems. >> >> My goal has always been to support this, but I've never quite gotten >> there, and it's never seemed to matter enough to people that. >> >> At the least, you need every resource to be reversible (including >> execs), and you need a more consistent internal event log than we >> currently maintain, but I think ordering issues become really >> complicated too -- if you installed a package then started a service, >> you have to do them in the inverse order, but if you modified a file >> and restarted the service then you have to do them in the same order. >> >> Basically, you need a much richer language for talking about >> dependencies, and I think that's the real complexity. >> >> -- >> Risk! Risk anything! Care no more for the opinions of others, for >> those >> voices. Do the hardest thing on earth for you. Act for yourself. Face >> the truth. -- Katherine Mansfield >> --------------------------------------------------------------------- >> Luke Kanies |http://reductivelabs.com|http://madstop.com > > As nice as it might be to have completely reversible resources, I > don't think we need quite that much. > Instead, I would like something like the notify parameter, but > specifically triggered on failure only. > i.e.: > > exec {"generic install" > command => "mkdir -p /opt/program ; tar -xzf /tmp/program- > $kernel.tar.gz -C /opt/program", > creates => "/opt/program", > onfail => Exec["opps"], > } > > exec => { "opps": > command => "rm -rf /opt/program", > refreshonly => true, > } > > Basically, just making it easy to create and call custom undo scripts > or functions. > Even though the creates paramater is bad, the exec would try again > next run because /opt/program gets removed on failure. > And than "opps" will only run after a failure of "generic install". > With the way notify works, an undo could start triggering other > resources as well; although the potential undo-redo cycle might be > problematic.
Ah; another interesting approach. Unfortunately our event system isn't quite flexible enough to handle it. We're revamping it for rowlf, so it should be extensible enough to support it after that. Can you enter that as a feature request? -- A person's maturity consists in having found again the seriousness one had as a child, at play. --Friedrich Nietzsche --------------------------------------------------------------------- Luke Kanies | http://reductivelabs.com | http://madstop.com --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Puppet Users" group. To post to this group, send email to puppet-users@googlegroups.com To unsubscribe from this group, send email to puppet-users+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/puppet-users?hl=en -~----------~----~----~----~------~----~------~--~---
