14:02:30 <sigmavirus24> #startmeeting glance drivers
14:02:31 <openstack> Meeting started Tue Jun 30 14:02:30 2015 UTC and is due to finish in 60 minutes.  The chair is sigmavirus24. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:02:32 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:02:35 <openstack> The meeting name has been set to 'glance_drivers'
14:02:39 <flaper87> o.
14:02:40 <flaper87> o/
14:02:48 <sigmavirus24> #link https://etherpad.openstack.org/p/glance-drivers-meeting-agenda
14:03:35 <sigmavirus24> Doesnt' look like there's much more other than to re-review the specs from last week
14:04:09 <sigmavirus24> johnthetubaguy: it looks from your comment that https://review.openstack.org/#/c/192926/ is acceptable to Nova?
14:05:05 <sigmavirus24> To reiterate for jokke_ https://etherpad.openstack.org/p/glance-drivers-meeting-agenda
14:05:27 * flaper87 just put his tasks spec in the prio list
14:05:30 <flaper87> :P
14:05:47 <sigmavirus24> #topic Change-since filter
14:05:49 <sigmavirus24> #link https://etherpad.openstack.org/p/glance-drivers-meeting-agenda
14:05:51 <sigmavirus24> d'oh
14:05:57 <sigmavirus24> #link https://review.openstack.org/#/c/192926/
14:06:02 <sigmavirus24> (real one that time)
14:06:35 <flaper87> The overall spec sounds good to me
14:06:45 <flaper87> sigmavirus24: any reason why you want to block it until you see some code?
14:07:06 <sigmavirus24> Mostly I want to make sure it's at least started and will be delivered
14:07:25 <sigmavirus24> Steve and I are on the same team and the amount of time we have had of late to work on this stuff has been next to none
14:07:32 <flaper87> I think we should just move everything to backlog and then move implemented specs to the <release> folder
14:07:55 <sigmavirus24> Unless someone will collaborate with Steve, I want to make sure we can get some progress started that someone else can pick up on
14:07:59 <flaper87> I really don't like that we're blocking specs on review until there's some code
14:08:06 <sigmavirus24> flaper87: that's fair
14:08:12 <sigmavirus24> I know that's not how other projects work either
14:08:15 <flaper87> We need to prioritize on reviews more than specs
14:08:36 <flaper87> sigmavirus24: yeah, FWIW, we're moving to "Everythin in backlog" in Zaqar too
14:08:48 <jokke_> flaper87: ++ ... Personally I prefer not to do implementation before the design has been approved :P
14:08:50 <flaper87> it's less frustrating, it allows you to focus on the feature without focusing on "when"
14:08:51 <sigmavirus24> I know Swfit merges it and then allows updates as things change
14:09:01 <sigmavirus24> flaper87: jokke_ you've sold me =P
14:09:22 <flaper87> ah and yeah, updating the spec is more than fine if there were changes
14:09:31 <flaper87> it helps to keep track of why/when/feedback
14:09:41 <sigmavirus24> Yep
14:09:49 <flaper87> ok, with that in mind, I'll send an email to openstack-dev
14:10:05 <flaper87> so we can get feedback from rosmaita, nikhil_k-away  and others
14:10:07 <sigmavirus24> #action flaper87 to send (yet another) email to openstack-dev
14:10:08 <sigmavirus24> =P
14:10:12 <flaper87> sigmavirus24: LOL
14:10:25 <jokke_> flaper87: do you think the spec processes differ enough by now that it would be good time to write X-proj spec for that :P
14:10:28 <sigmavirus24> Sorry. Not sorry.
14:10:35 <sigmavirus24> jokke_: how meta
14:10:36 <sigmavirus24> =P
14:10:57 <flaper87> jokke_: we should have a spec explaining how to write spec and then discuss when it should be implemented
14:11:17 <sigmavirus24> Maybe we should discuss that in #openstack-dev after the meeting. We're 1/3 of the way through ou rtime
14:11:22 <jokke_> flaper87: and not approve that spec before everyone comply :D
14:11:30 <sigmavirus24> oh god
14:11:34 <flaper87> sigmavirus24: next
14:11:46 <sigmavirus24> #topic Configurable Checksum for Glance
14:11:47 <sigmavirus24> #link https://review.openstack.org/#/c/191542/
14:11:56 <sigmavirus24> It's still a wip and Brianna left some good feedback that I need to address
14:12:04 <sigmavirus24> I don't think we need to spend much time on it
14:12:43 <flaper87> I like the idea but I'm not an expert on the topic
14:13:01 <flaper87> I'll defer to experts and +2 the idea of having other ways to guarantee the integrity of images
14:14:01 <sigmavirus24> So this is mostly to help Brianna with her Image Signing and Encryption spec
14:14:22 <sigmavirus24> MD5 is a problem, but I'm not certain it's a problem that will immediately affect us today or is actively being exploited
14:14:36 <sigmavirus24> That said, it's a problem and having a way to fix it now is better than having to rush a fix later
14:14:43 <sigmavirus24> Everyone okay with moving on?
14:14:51 <sigmavirus24> oh
14:14:52 <sigmavirus24> Also
14:14:56 <flaper87> next++
14:15:11 <sigmavirus24> I'm talking with #openstack-security and #cryptography developers about perhaps better ways than cryptographic hashing algorithms
14:15:12 <sigmavirus24> so
14:15:14 <sigmavirus24> That's a thing
14:15:28 <sigmavirus24> This spec may change further or a secondary spec may enter the ring. That's not certain yet
14:15:35 <sigmavirus24> #topic Glance Store Refactor
14:15:37 <sigmavirus24> #link https://review.openstack.org/#/c/188050/
14:15:51 <flaper87> Another spec that would be great to have in a backlog dir while it's implemented
14:15:57 <sigmavirus24> I agree
14:16:00 <flaper87> sigmavirus24: has Cindy contacted you?
14:16:03 <sigmavirus24> Yes
14:16:10 <flaper87> I believe she's in need of reviews
14:16:11 <sigmavirus24> I've agreed to review the impl
14:16:15 <sigmavirus24> Oh is it ready for review?
14:16:22 <sigmavirus24> flaper87: she has 3 on the list
14:16:27 <flaper87> mmh, I *think* so
14:16:29 <flaper87> mmh
14:16:31 <flaper87> ok
14:16:33 <flaper87> I'll ping her
14:16:40 * flaper87 writes that down with a red-color pen
14:16:42 * sigmavirus24 just pinged her in #openstack-glance
14:16:49 <jokke_> I'm more and more thinking that perhaps we should implement glance_store API v2 rather than trying to refactor the current one :P
14:17:08 <flaper87> jokke_: which is pretty much what she's doing
14:17:22 <flaper87> It's call refactor because we can't maintain 2 APIs
14:17:29 <flaper87> so, eventually, we need to remove the old one
14:17:37 <sigmavirus24> Yep
14:17:39 <flaper87> especially because it's a library
14:17:48 <flaper87> next ++
14:18:01 <sigmavirus24> I just left a comment/question
14:18:08 <sigmavirus24> But I think it doesn't block it
14:18:10 <jokke_> flaper87: somewhere around the time we have glance_store API v3 that does not provide FS- & swift-stores ;)
14:18:34 <sigmavirus24> lol
14:18:42 <flaper87> mmh, not sure why it shouldn't provide FS/swift-stores
14:18:44 <sigmavirus24> jokke_: why don't you like the swift store?
14:18:45 <flaper87> ah fuck u
14:18:46 <sigmavirus24> =P
14:18:49 <flaper87> you're trolling me
14:18:51 <sigmavirus24> lol
14:18:54 * flaper87 fell on that one
14:18:57 * sigmavirus24 high fives jokke_
14:19:07 <sigmavirus24> #topic Automatic Task Triggering
14:19:08 <sigmavirus24> #link https://review.openstack.org/#/c/188388/
14:19:09 <jokke_> \\o \o/ o// o/7
14:19:29 <flaper87> I've addressed sabari's comments and I've had several discussions with him on IRC
14:19:32 <flaper87> I have few nits there
14:19:41 <flaper87> and I'm ready to start implementing it
14:19:50 <sigmavirus24> I haven't read it
14:19:54 <flaper87> but but but but, ^
14:19:55 <sigmavirus24> Since it was added during the meeting
14:19:56 <sigmavirus24> Give me a second
14:20:08 <sigmavirus24> =P
14:20:13 <flaper87> sigmavirus24: sure, you can also read it after the meeting but please, do it :D
14:20:19 <sigmavirus24> I will
14:20:25 <flaper87> I think that spec adds great value to our tasks' engine
14:20:33 <flaper87> or whatever we want to call that thing we have
14:21:13 <jokke_> flaper87: ok, why should glance guess whe certain tasks should be ran rather than someone (automated tool) calling glance and telling it to do so?
14:21:58 <flaper87> it's not really guessing, you're telling glance when to run it
14:22:07 <flaper87> it just happens when "certain things" happen in glance
14:22:21 <flaper87> A good example is that an automated tool won't detect when an image has been added to glance
14:22:41 <sigmavirus24> flaper87: it /could/ but we wont' go there =P
14:22:43 <flaper87> which will endup in the image being saved in the store in a format that you don't want (just following up on one of the examples I added there)
14:22:49 <jokke_> flaper87: it does if it's listening the notifications (in the case we happen to advertice such things)
14:22:52 <flaper87> sigmavirus24: ssshhhh
14:23:00 * sigmavirus24 agrees with jokke_
14:23:16 <sigmavirus24> but I'm not sure anyone actually uses ceilometer
14:23:24 <flaper87> jokke_: right and you're asking ops to go and listen on notifications and write their own tools to have images converted to raw because that's what they want
14:23:36 <flaper87> sigmavirus24: ceilo won't help, you'd have to write the whole thing
14:23:37 <sigmavirus24> flaper87: IFTT integration
14:23:47 <sigmavirus24> *IfTTT
14:23:49 <sigmavirus24> whatever it is
14:23:55 <sigmavirus24> "If This Than That"
14:24:20 <kragniz> then that
14:24:31 <flaper87> kragniz: look who's here
14:24:42 <flaper87> So, that said, I bet jokke_ is writing a long IRC message
14:24:52 * kragniz totally contributes here
14:25:12 <jokke_> flaper87: no ... trying to think how to put my thought process to words
14:25:36 <flaper87> anyway, I don't think it'll add huge costs of maintenance and it still brings something useful to OPs
14:25:56 <sigmavirus24> kragniz: thought I'd trick flaper87 and jokke_ with that
14:26:17 <sigmavirus24> Yeah. I know CERN and someone else was looking for this at the summit
14:26:29 <jokke_> mainly my point is that if we have tooling following what happens in glance (not in single glance node) we have way better resiliency to get things done, possibilities to offload such tasks to dedicated nodes etc.
14:26:57 <flaper87> but we already have that in glance
14:27:08 <flaper87> we even adopted taskflow that is suppose to take care of managing workers
14:27:53 <kragniz> adding explicit triggers is a lot more friendly than asking people to write extra tools for something that can be useful to quite a few ops
14:28:02 <flaper87> If this doesn't make sense in Glance, I'd even argue that having tasks makes sense
14:28:17 <flaper87> People could simply have their tasks implemented elsewhere
14:28:23 <flaper87> and then do whatever they want
14:28:51 <flaper87> Not to mention that, as of today, you need to write fucking json on the CLI to trigger a task
14:29:01 <sigmavirus24> lol
14:29:07 <flaper87> s/fucking//
14:29:11 <sigmavirus24> less than a minute left
14:29:14 <flaper87> sorry, I start cursing when I rant
14:29:17 <flaper87> :D
14:29:17 * kragniz agrees with flaper87 here
14:29:28 <jokke_> flaper87: ok, one clarification as tl;dr ... plan is to get general configurable ruleset rather than attacking the usecases hardcoded which are mentined in the spec?
14:29:36 <flaper87> ok, please, drop comments there, read it and at least say you agree so I can start coding
14:29:57 <flaper87> jokke_: configurable rule set is the goal
14:30:03 <flaper87> things in the spec are just examples
14:30:08 <flaper87> (sabari requested them)
14:30:08 <sigmavirus24> :+1:
14:30:12 <sigmavirus24> And we're out of time
14:30:15 <flaper87> sigmavirus24: GH hipster
14:30:18 <sigmavirus24> #endmeeting