16:03:22 <jgriffith> #startmeeting
16:03:23 <openstack> Meeting started Wed Jun 20 16:03:22 2012 UTC.  The chair is jgriffith. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:03:24 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:03:26 <DuncanT> A similar cinder-manage command would, I think, be useful - it is safer than manually grubbing in the database, and requiring root to use it (to get the db creds from the .conf file) makes it fairly safe
16:03:48 <dricco> here too
16:03:48 <jgriffith> DuncanT: Absolutely
16:03:52 <jgriffith> dricco: Hello
16:04:00 <jgriffith> #topic status update
16:04:29 <jgriffith> For those that saw my email regarding testing out cinder with my github you may have run into some problems :(
16:04:40 <jgriffith> It seems glance isn't gettting cloned properly
16:05:19 <jgriffith> For those that tried it (clayg) I apologize for the confusion
16:05:29 <jgriffith> Anyway...
16:05:51 <jgriffith> The good news is euca- commands are working now
16:06:15 <jgriffith> I've added tests to devstack to attach/detach, snapshot and deletes
16:06:31 <jgriffith> I think we're pretty close to being ready to go out into the wild
16:07:02 <jgriffith> I'm pushing to make Cinder default in devstack after f2 release
16:07:26 <jgriffith> After that we have a lot of work to do in terms of features
16:07:59 <jgriffith> Anybody else tried running anything, or have any updates on current status?
16:08:08 <jgriffith> I won't speak for sleepsonthefloor :)
16:08:36 <sleepsonthefloor> jgriffith: finishing up moving the snapshot api changes to an extension
16:08:43 * heckj wanders in to lurk
16:08:54 <sleepsonthefloor> may have to update cinder client as well
16:09:11 <sleepsonthefloor> after that, it would seem that nova tests are the main task to be done?
16:09:26 <jgriffith> sleepsonthefloor: Sounds good... I had some updates in the client, but moving those to an extension may mean another turn on them
16:09:36 <jgriffith> sleepsonthefloor: Yes, nova tests.... I'm still working on those
16:09:40 <DuncanT> We've got a working devstack instance going here, including working euca commands... not got our driver port working on it yet but working on it
16:09:52 <jgriffith> DuncanT: cool
16:10:35 <jgriffith> sleepsonthefloor: I almost have a completed fake cinder service finished
16:10:39 <sleepsonthefloor> In terms of the merge timeline, the relevant keystone changes are in
16:10:45 <DuncanT> I also put up https://github.com/Funcan/nova-volume-end2end though it needs some tweaks to work with a recent euca2ools library
16:11:27 <jgriffith> sleepsonthefloor: So what I was thinking was get everything merged, saving Nova for last
16:11:55 <jgriffith> For devstack merge everything with n-vol as default, then submit a patch to switch the default after that
16:12:40 <jgriffith> any thoughts from anybody else on the ordering there?
16:13:27 <sleepsonthefloor> jgriffith: makes sense to me.  once the snapshot api extension and related client work are done, cinder and cinder client should be ready.  hopefully those land today/tomorrow
16:13:55 <jgriffith> sleepsonthefloor: Great.. and I'm hoping close to the same timeline for nova
16:14:14 <jgriffith> The nova tests may still have some holes, but it should be sufficient by then
16:14:32 <jgriffith> DuncanT: Do you want to talk a bit about your end2end tests?
16:15:32 <jgriffith> clayg: Make up your mind :)
16:15:33 <DuncanT> Sure, they (currently) use the euca api to create instance and volumes, write data to them, move the volume to different instances and check the result
16:15:57 <clayg> jgriffith: i can't stay, i just wanted to pop in and say I'm playing with cinder (got all my issues worked out last night)
16:15:58 <DuncanT> There is a similar test for snapshots there
16:16:10 <clayg> I'll try to watch for messages that show up in yellow
16:16:47 <jgriffith> DuncanT: very cool... any thoughts about moving this into devstack, or working with ci team to do something ie gate test?
16:16:50 <clayg> I'm going to try and find bugs in cinder, track down all the patches so I can repoduce cinder in a multinode xen/nova setup - and maybe write some unittests!
16:17:03 <jgriffith> clayg: Awesome!!
16:17:30 <DuncanT> Certainly we can look at that, yeah. Devstack integration should be fairly easy
16:17:43 <jgriffith> DuncanT: A bit of it is already there in euca.sh, but I like the actual data transfer addition
16:17:58 <DuncanT> Also intend to have a version for the nova/cinder api soon
16:18:08 <jgriffith> I need to step away for one minute... sorry, keep talking I'll catch up
16:18:26 <jgriffith> Never mind... I'm good
16:18:55 <jgriffith> DuncanT: Sounds great, keep us updated
16:19:28 <jgriffith> DuncanT: Come across any issues thus far or are things looking pretty good?
16:19:57 <DuncanT> Not found anything yet
16:20:19 <jgriffith> Ok, sounds good...
16:20:30 <DuncanT> Will be hitting it a bit harder as I get our driver wired up to it
16:20:41 <jgriffith> One thing we need to think about is the whole specifying device mount point...
16:21:03 <jgriffith> AFICT this doesn't really work in n-vol or cinder
16:22:07 <jgriffith> Ok... anything else for status updates?  Questions?
16:22:25 <DuncanT> Under KVM the mount point specification is entirely bogus with nova today... and in a difficult-to-explain-to-customers way
16:22:48 <jgriffith> DuncanT: exactly, I'd like to figure out a way to address that
16:23:06 <jgriffith> DuncanT: not critical right now, but want it on our radar screen
16:24:03 <jgriffith> Ok... onward
16:24:10 <jgriffith> #topic hack day in SF
16:24:39 <jgriffith> So don't know if everybody here saw the email from Joshua M about a cinder hack day in SF?
16:25:03 <jgriffith> DuncanT: I know this probably isn't realistic for you :(
16:25:30 <jgriffith> I was wondering if any folks from inktank, redhat etc were lurking about today that were on that list/interested
16:25:51 <jgriffith> Also renuka you're out that way, would you be interested in such an event?
16:25:58 <renuka> yea im in
16:26:00 <heckj> jgriffith: when is the hackday?
16:26:40 <jgriffith> heckj: Well he sent out the email last week and was talking yesterday timeframe :)
16:27:05 <heckj> heh
16:27:07 <jgriffith> I thougth it would be good to bring it up here this week and gauge interest before I get on a plane :)
16:27:35 <heckj> I'm in Seattle, but am kind of interested. Wouldn't be able to contribute a whole lot right off that bat...
16:27:41 <jgriffith> Also one question I had was if it would be more productive to wait until the basic integration was done and out there
16:27:45 <DuncanT> I'd love to come along but I'm not sure it is practical :-(
16:28:09 <jgriffith> DuncanT: Yeah, I figured... although maybe we could all come to Ireland?  I've wanted to visit for a while now.
16:28:24 <renuka> integration might be a good idea
16:28:34 <DuncanT> That sounds like a great plan to me... it's a lovely place
16:28:52 <jgriffith> renuka: Yeah, I think folks could contribute a lot more once there's a stable foundation
16:29:39 <jgriffith> So let's put it this way... if we set such a thing up say in July after F2 would there be enough interest from folks here to show up and get together?
16:29:42 * heckj is happy to wait and learn
16:30:14 <renuka> i'm in whichever way
16:31:02 <jgriffith> Ok, I think I'll try and coordinate something.  There are enough people in the area there that it could be fun/useful
16:31:30 <jgriffith> I'll keep folks posted, but I don't think it makes as much sense to do it before we're integrated
16:31:55 <jgriffith> I'd like to get some things like Boot From Volume knocked out in that type of setting
16:31:59 <jgriffith> :)
16:32:33 <renuka> speaking of which... boot from volume now works for xen :)
16:32:40 <renuka> code in review
16:33:00 <jgriffith> renuka: Awesome!
16:33:14 <jgriffith> renuka: Which review was that?  I've looked at a couple from you lately
16:33:37 <renuka> https://review.openstack.org/#/c/8156/ It's mostly in the compute space
16:33:55 <renuka> but i guess it is of interest to volume people :)
16:34:06 <jgriffith> renuka: very much so :)
16:35:01 <jgriffith> Ok...
16:35:13 <jgriffith> We've diverged a bit
16:35:19 <jgriffith> #topic open discussion
16:35:33 <jgriffith> Fire away!
16:36:15 <jgriffith> ha!  I'm going to get off easy this week!
16:36:38 <DuncanT> One question related to a recent mailing list thread
16:36:45 <jgriffith> DuncanT: shoot
16:37:04 <DuncanT> Any thoughts on whether delete should work on states otehr than available & error?
16:37:53 <DuncanT> Deleting an 'attached' volume seems odd... you could be an implicit detattach but I suspect anybody trying to delete an attached disk has typo'd
16:38:01 <jgriffith> DuncanT: my thought is no, however there's the thing about being abl to force a state
16:38:36 <DuncanT> There are ways to get stuck in a state where nova database thinks the device is attached but nothing else does
16:38:57 <DuncanT> nova-manage force-attach can be useful there sometimes but an end-user can't run that
16:39:22 <DuncanT> Similarly it is possible to get stuck in creating or attaching with no way out as the end user
16:39:32 <jgriffith> DuncanT: Yes, so we need a way to "fix" that, but in terms of what you can delete I think it should be dependent on the states you mention
16:39:50 <jgriffith> Also that brings up the question of deleting volumes with snapshots again :)
16:40:36 <DuncanT> Yeah... currently our code lets you but then the snapshots become useless because snapshots have no provider location :-(
16:40:40 <jgriffith> DuncanT: We'll need to put together a blueprint for adding something like "correct-state" to the api/client
16:41:17 <DuncanT> We've certainly got usecases where users would rather not pay for both the volume and the snapshot when all they need is the snapshot
16:41:33 <jgriffith> DuncanT: Yeah, I think we may change definition and implementation of snapshots in the future based on past conversations
16:41:57 <DuncanT> Sounds good to me
16:42:03 <renuka> DuncanT: that seems to be a matter of billing policy
16:42:49 <renuka> i think some hypervisors might be storing snapshots as deltas... that is the reason they dont allow deleting the original volume
16:43:05 <DuncanT> renuka: It's also a matter of conceptual model... if the user only wants to keep a golden snapshot, why shouldn't they be able to delete the volume that the snapshot was created from?
16:43:27 <DuncanT> I realise what semantics a given implementation supports can vary
16:43:30 <jgriffith> DuncanT: renuka I seem to recall we started this topic at length a while back :)
16:43:43 <jgriffith> I proposed the "golden copy" model be a clone
16:44:28 <DuncanT> jgriffith: I don't remember coming up with a model mapped onto an api... I do remember the discussion
16:44:57 <DuncanT> Should I send out an email about it and give people time to think?
16:45:11 <jgriffith> DuncanT: no we didn't resolve it
16:45:47 <jgriffith> DuncanT: Probably a good idea, but it's not something I intend to try and solve in the next week or two
16:45:56 <jgriffith> DuncanT: Just to be fair/honest
16:46:16 <jgriffith> Although if others have time/bandwidth to tackle it and are ready that's another story
16:46:48 <DuncanT> Oh, I don't expect a speedy resolution, just want to keep the discussion alive so we don't get stuck with not changing anything do to lack of agreement on where we'd like to go
16:47:02 <jgriffith> My concern would be if there are changes in how we do things like BFV etc they may impact it so I won't to be able to really focus everyone with a big picture in place
16:47:19 <jgriffith> DuncanT: I agree with that completely
16:48:38 <jgriffith> DuncanT: sending out something to ML is probably a good idea, it you want to run with it sounds good to me
16:49:11 <DuncanT> Ok, I'll put something together
16:50:05 <jgriffith> Anybody else have anything?
16:51:10 <jgriffith> Next week should be exciting, we'll hit our first big milestone :)
16:51:28 <jgriffith> Alrighty, thanks everyone!
16:51:33 <jgriffith> #endmeeting