Activity log for bug #1441117

Date Who What changed Old value New value Message
2015-04-07 11:24:34 Dmitry Tantsur bug added bug
2015-04-07 11:25:32 Dmitry Tantsur description <lucasagomes> I mean, if someone downstream have a custom processing_hooks I think they want to not change it when discoverd changes <lucasagomes> that prevents discoverd downstream from running something they don't want to <lucasagomes> or... lemme rephrase, not that they don't want to <dtantsur> lucasagomes, not necessary, e.g. we only need to insert a couple of plugins <dtantsur> actually the default set of plugins is something you very rarely want to change, if ever <lucasagomes> but they don't want to run more or less hooks when they update discoverd if they have a custom processing_hooks, they might just want to run exactly the same <lucasagomes> right <dtantsur> lucasagomes, they can still use just 'processing_hooks' and override everything like before. But I'd like us (RH) to work with default discoverd + plugins <lucasagomes> dtantsur, I find it complicated to have 2 lists of hooks that will run <dtantsur> another option would be to have 'processing_hooks' that acts as append, and also some knd of blacklist <lucasagomes> right, I don't know. I kinda like the whitelist more <lucasagomes> as is right now <lucasagomes> it feels safer if you want to have control over ur enviroment, so that new things won't run unless you explicity tell it to run <lucasagomes> kinda liek enabled_drivers for ironic <dtantsur> lucasagomes, enabled_drivers in Ironic is a completely different beast, e.g. they all are independent <lucasagomes> and if one wants to do the default + plugins... perhaps they should script it? * lucasagomes thinks <lucasagomes> dtantsur, right, anyway it's a white list <dtantsur> lucasagomes, e.g. if I ever split essential plugin into 2, people will get broken. that can't happen with Ironic. but well, I'm starting to agree that I don't have compelling enough reason for this option... <lucasagomes> dtantsur, the good thing about this option is that you still can have the old behavior with processing_hooks only <lucasagomes> if u just leave the append_processing_hooks empty <dtantsur> yep, it's pure opt-in <lucasagomes> so yeah, I don't have a strong opnion about it either. It just looks a bit odd to have 2 options to enable hooks <lucasagomes> dtantsur, plus the other that the hooks are specified matters right? <dtantsur> lucasagomes, yep <lucasagomes> so sometimes one wnat to keep the default hooks, but he has a custom hook that should run before everything else <lucasagomes> and that option won't actually help him <lucasagomes> because it just appends to the end of the list <lucasagomes> devananda, g'night <dtantsur> lucasagomes, actually I was thinking about prepend_processing_hooks too :D <lucasagomes> heh <dtantsur> but yeah, maybe I'm overengineering things... <lucasagomes> dtantsur, yeah maybe we should brainstorm see what would be a good way to do it <lucasagomes> I think that, if one wants to change the hooks etc he could script it <lucasagomes> to apply the configuration with the hooks and order he wants <lucasagomes> I mean, if someone downstream have a custom processing_hooks I think they want to not change it when discoverd changes <lucasagomes> that prevents discoverd downstream from running something they don't want to <lucasagomes> or... lemme rephrase, not that they don't want to <dtantsur> lucasagomes, not necessary, e.g. we only need to insert a couple of plugins <dtantsur> actually the default set of plugins is something you very rarely want to change, if ever <lucasagomes> but they don't want to run more or less hooks when they update discoverd if they have a custom processing_hooks, they might just want to run exactly the same <lucasagomes> right <dtantsur> lucasagomes, they can still use just 'processing_hooks' and override everything like before. But I'd like us (RH) to work with default discoverd + plugins <lucasagomes> dtantsur, I find it complicated to have 2 lists of hooks that will run <dtantsur> another option would be to have 'processing_hooks' that acts as append, and also some knd of blacklist <lucasagomes> right, I don't know. I kinda like the whitelist more <lucasagomes> as is right now <lucasagomes> it feels safer if you want to have control over ur enviroment, so that new things won't run unless you explicity tell it to run <lucasagomes> kinda liek enabled_drivers for ironic <dtantsur> lucasagomes, enabled_drivers in Ironic is a completely different beast, e.g. they all are independent <lucasagomes> and if one wants to do the default + plugins... perhaps they should script it? * lucasagomes thinks <lucasagomes> dtantsur, right, anyway it's a white list <dtantsur> lucasagomes, e.g. if I ever split essential plugin into 2, people will get broken. that can't happen with Ironic. but well, I'm starting to agree that I don't have compelling enough reason for this option... <lucasagomes> dtantsur, the good thing about this option is that you still can have the old behavior with processing_hooks only <lucasagomes> if u just leave the append_processing_hooks empty <dtantsur> yep, it's pure opt-in <lucasagomes> so yeah, I don't have a strong opnion about it either. It just looks a bit odd to have 2 options to enable hooks <lucasagomes> dtantsur, plus the other that the hooks are specified matters right? <dtantsur> lucasagomes, yep <lucasagomes> so sometimes one wnat to keep the default hooks, but he has a custom hook that should run before everything else <lucasagomes> and that option won't actually help him <lucasagomes> because it just appends to the end of the list <lucasagomes> devananda, g'night <dtantsur> lucasagomes, actually I was thinking about prepend_processing_hooks too :D <lucasagomes> heh <dtantsur> but yeah, maybe I'm overengineering things... <lucasagomes> dtantsur, yeah maybe we should brainstorm see what would be a good way to do it <lucasagomes> I think that, if one wants to change the hooks etc he could script it <lucasagomes> to apply the configuration with the hooks and order he wants One of ideas: https://review.openstack.org/#/c/170535/
2015-07-07 16:20:31 OpenStack Infra ironic-inspector: status Confirmed In Progress
2015-07-07 16:20:31 OpenStack Infra ironic-inspector: assignee Sam Betts (sambetts)
2015-07-08 08:07:07 Dmitry Tantsur ironic-inspector: milestone 2.0.0
2015-07-08 09:51:28 OpenStack Infra ironic-inspector: status In Progress Fix Committed
2015-07-09 14:47:30 Dmitry Tantsur ironic-inspector: status Fix Committed Fix Released