15:01:09 <ihrachys> #startmeeting neutron_upgrades
15:01:09 <openstack> Meeting started Mon Feb 15 15:01:09 2016 UTC and is due to finish in 60 minutes.  The chair is ihrachys. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:11 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:01:13 <openstack> The meeting name has been set to 'neutron_upgrades'
15:01:25 <ajo> o/
15:01:26 <korzen> hello
15:01:28 <ihrachys> o/ all. giving some small time for everyone to fasten their seatbelts
15:03:02 <rossella_s> hi
15:03:08 <electrocucaracha> GM
15:03:15 <ihrachys> ok let's get working.
15:03:32 <mhickey> Hi; sorry I am late
15:03:37 <ihrachys> no announcements my side, so just going straight into topics
15:03:41 <ihrachys> #topic Partial Multinode Grenade
15:04:09 <ihrachys> not much progress, still the same 3 failures.
15:04:21 <ihrachys> we have some patches to add more diagnostics to grenade runs: https://review.openstack.org/#/q/branch:master+topic:neutron-worlddump
15:04:37 <ihrachys> and honestly I lean to setting devstack-gate like setup locally and debug it from there
15:04:59 <ihrachys> since throwing patches into gate and hope for the best does not seem very effective
15:05:08 <ihrachys> that's about it I guess
15:05:21 <ihrachys> now juicy stuff
15:05:21 <ihrachys> #topic Object implementation
15:05:27 <ihrachys> rossella_s: wanna update?
15:05:39 <rossella_s> ihrachys, yes!
15:06:00 <ajo> About the previous topic, I agree, we need a mechanism like that to jump in to CI VMs and eventually debug
15:06:25 <slunkad> hello
15:06:25 <ihrachys> ajo: sadly requires infra involvement, so probably not this time. but yes.
15:06:30 <ihrachys> slunkad: hi
15:06:39 <ihrachys> rossella_s: shoot :)
15:06:41 <ajo> ihrachys, yes, we just need to raise the concern with infra
15:07:12 <rossella_s> ihrachys, so I was lazy and didn't do anything...korzen is still working on the subnet ovo, mhickey is working on the extra dhcp extension
15:07:37 <rossella_s> dguitarbite has not started on the port sec extension
15:07:40 <mhickey> rossella_s: some work also done on address pairs
15:08:02 <rossella_s> I've seen saisriki patch for the sqlachemy types
15:08:07 <ihrachys> yeah, I started going thru all the bits you all folks work on today. will continue reviews.
15:08:39 <ihrachys> that types patch is of concern to me, yes.
15:08:39 <ihrachys> https://review.openstack.org/#/c/277558/
15:08:40 <rossella_s> slunkad, is working on the sec group extension...still WIP and I didn't have the time to review
15:08:49 <korzen> I will continue the effort to introduce the composite key case
15:09:14 <rossella_s> korzen, great
15:09:16 <ihrachys> korzen: I think we have a solid base for the composite keys to merge it this week.
15:09:26 <electrocucaracha> I uploaded some changes of subnetpool but it requires the change of composite keys that korzen has proposed.  Subnetpoolprefix requires it
15:09:29 <slunkad> rossella_s: yes I wasn't able to update my patch this week will try to do that this week
15:09:43 <slunkad> *last
15:09:55 <ihrachys> electrocucaracha: we could rebase patches on top of korzen's work if that's what you depend on
15:10:37 <rossella_s> actually I have another concern...maybe we should talk about it during the discussion slot
15:11:04 <ihrachys> rossella_s: as long as it's objects, we can do it now.
15:12:07 <rossella_s> ihrachys, it's not so specific to object, let's leave it for later
15:12:08 <ihrachys> rossella_s: are you done on objects?
15:12:10 <ihrachys> ok
15:12:13 <rossella_s> ihrachys, yes
15:12:30 <ihrachys> so I was wrapping my head around the types patch:
15:12:31 <ihrachys> #link https://review.openstack.org/#/c/277558/
15:12:58 <ihrachys> so my understanding is that without the patch like that (and db models migrated to using new types), we can't use OVO fields for IP address
15:13:46 <ihrachys> concern here is that unless netaddr representation is identical to our current string based one, we could need migration (I hope representation is the same though)
15:13:57 <ihrachys> and even if the representation is the same, types may need migration.
15:13:59 <ihrachys> is that correct?
15:14:03 <ihrachys> slunkad: rossella_s ?
15:14:04 <rossella_s> ihrachys, yes
15:14:24 <ihrachys> that's a bit of sad honestly. :)
15:14:29 <rossella_s> ihrachys, I know
15:14:34 <rossella_s> ihrachys, I can't see any workaround
15:14:38 <ihrachys> first since we may be blocked until we get the types in (and people push them into oslo.db)
15:14:57 <ihrachys> and second since migrations mean higher downtime to do the migration.
15:15:12 <rossella_s> ihrachys, well we could use string for now...and then migrate but it's going to be more complex probably
15:15:23 <mhickey> rossella_s: why was IPAddress field added to the o.vo?
15:15:30 <rossella_s> ihrachys, or we stick with strings :/
15:15:52 <rossella_s> mhickey, what do you mean?
15:15:59 <ihrachys> mhickey: apparently because nova already used it
15:16:06 <ihrachys> mhickey: the whole library came from nova world
15:16:09 <mhickey> https://github.com/openstack/oslo.versionedobjects/blob/master/oslo_versionedobjects/fields.py#L402
15:16:47 <mhickey> so how does nova persist the field?
15:17:07 <korzen> I've heard that IP address in o.vo was added recetly
15:17:21 <mhickey> korzen: ok
15:17:35 <ihrachys> mhickey: the netaddr.IPAddress (not string) is passed into SQLAlchemy
15:17:36 <mhickey> sorry thinking out loud; haven't had time to investigate
15:17:48 <ihrachys> mhickey: and then the type makes the proper convertion for you.
15:17:48 <rossella_s> mhickey, https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/types.py
15:18:36 <rossella_s> ihrachys, anyway this discussion should be probably brought to the neutron meeting?
15:18:36 <mhickey> ok, so everything started in nova and is now started to be centralized?
15:18:41 <korzen> the new sql type is sql alchemy decorator to handle inserts
15:18:45 <ihrachys> rossella_s: is there a strict need to map the type to INET for psql?
15:19:02 <rossella_s> ihrachys, TBH didn't check
15:19:17 <ihrachys> because if we can unconditionally map to String, then at least we don't need to modify schema
15:19:36 <rossella_s> ihrachys, I think that works, it's what we are doing now
15:19:53 <rossella_s> ihrachys, the down side is that there's no type check ...
15:20:10 <ihrachys> right. I actually don't think that the proposal to have it in oslo.db may work. since different projects have different requirements to map alias types to backend types.
15:20:54 <ihrachys> rossella_s: I hear you. but at least we can then postpone/avoid alembic migrations and figure out the way out later.
15:20:55 <rossella_s> ihrachys, I agree
15:20:57 <korzen> the workaround for us would be to implement conversion to string in code, without schema and sqlalchemy decorator
15:20:58 <mhickey> ihrachys: but I thought that is a reason to centralize to standardize?
15:21:36 <rossella_s> ihrachys, we can postpone the pain :)
15:21:52 <ihrachys> mhickey: we standardize on using OVO field type, yes. we diverge on how we handle the result on sqlalchemy level.
15:22:09 <mhickey> ihrachys: ok
15:22:17 <ihrachys> ok, I think I have some mind map now. thanks folks for clarification.
15:22:37 <korzen> the question is should we use string? instead of IP address?
15:22:38 <ihrachys> I will try to convince reviewers it may be ok to have it in neutron tree :)
15:22:48 <rossella_s> ihrachys, :D
15:22:51 <ihrachys> korzen: I think on object level it should still be IPAddress
15:23:13 <korzen> ihrachys then we need to convert on insert
15:23:23 <korzen> otherwise the code blows
15:23:32 <rossella_s> ihrachys, anyway you raised a good point. This conversion might slow us down quite a bit due to migrations
15:23:36 <ihrachys> korzen: yes, we need a SQLA type for that. just not an exact copy of what nova has.
15:23:46 <ihrachys> to avoid alembic
15:24:17 <rossella_s> ihrachys, maybe we should investigate how painful would be to move to these new types
15:24:43 <ihrachys> yeah, definitely something that should be tackled quick to avoid the blocker for other patches.
15:24:45 <rossella_s> ihrachys, to get the whole picture
15:24:55 <korzen> I would suggest to have local type with string(64) not string(39) like nova
15:25:02 <rossella_s> ihrachys, right, like high priority task
15:25:19 <ihrachys> korzen: yeah, that's another thing. it should be of size we already use in models for ip-address fields
15:26:16 <rossella_s> so my proposal would be to have a kind of POC with the introduction of a new type and migration needed
15:26:38 <korzen> rossella_s and until then?
15:26:40 <rossella_s> we can move forward from there...we need to get knowledge regarding what's exactly needed
15:27:08 <electrocucaracha> for stringfield sizes I found this bp https://blueprints.launchpad.net/nova/+spec/string-field-max-length
15:27:11 <rossella_s> korzen, until then we have to use strings...
15:27:17 <mhickey> rossella_s: ++; better work out issue head on
15:27:46 <ihrachys> I guess we need someone seasoned to spin on it real quick
15:27:46 <rossella_s> korzen, mhickey the type investigation should be higher priority since we are kind of blocked
15:27:50 <ihrachys> I may take it.
15:27:59 <ihrachys> unless someone wants to
15:28:07 <mhickey> I am open
15:28:09 <rossella_s> ihrachys, that would be great!!!
15:28:35 <rossella_s> this week I am bug deputy so I won't have time
15:28:45 <mhickey> rossella_s: ack! :)
15:28:51 <ihrachys> mhickey: ok let's discuss it off the meeting, I bet we can share :)
15:29:04 <ihrachys> rossella_s: np, we'll tackle it. :)
15:29:06 <mhickey> ihrachys: beers I hope! :)
15:29:33 <korzen> we can share beer in March
15:29:46 <korzen> BTW, travel already booked from my site
15:29:46 <mhickey> korzen: if I can make it...
15:29:47 <ihrachys> definitely. it's Czech Republic, ya know
15:29:52 <ihrachys> mhickey: you MUST!
15:29:57 <slunkad> ihrachys: I would like to help if I can in anyway
15:30:01 <mhickey> ihrachys: trying my best
15:30:14 <mhickey> I do like the Czech Beer
15:30:24 <ihrachys> korzen: nice. we'll do the personal round call later on who's in
15:30:34 <ihrachys> ok, one more thing related to objects
15:30:47 <ihrachys> ajo landed 1st piece of rpc callbacks rolling upgrade: https://review.openstack.org/265347
15:31:00 <rossella_s> noticed that
15:31:10 <ihrachys> and another piece was due to be avail for review today: https://review.openstack.org/268040 (probably not there yet?)
15:31:14 <mhickey> ajo: ++
15:31:16 <ihrachys> so keep an eye :)
15:31:24 <rossella_s> ihrachys, got it!
15:31:51 <ihrachys> ok let's switch to other non-object-y stuff
15:31:54 <ihrachys> #topic Other patches
15:32:14 <ihrachys> korzen shared a link with me for a nice patch lately
15:32:16 <ihrachys> #link https://review.openstack.org/#/c/124946/
15:32:46 <ihrachys> that one adds a framework that allows to add test resources to new alembic scripts and run migrations with them as part of test suite
15:33:06 <rossella_s> sweet
15:33:21 <ihrachys> so you could merge a script that e.g. adds a field to an object, and make sure the script is working with some objects of that type pre-inserted
15:33:22 <ajo> :)
15:33:25 <mhickey> nice
15:33:39 <ihrachys> still in deep review, but seems to be in right hands
15:33:52 <rossella_s> ihrachys, we can use that to test the type stuff ;)
15:34:16 <ihrachys> rossella_s: right you are, I guess
15:34:19 <ihrachys> also mhickey has some doc patch for db upgrade procedures for ops
15:34:21 <ihrachys> #link https://review.openstack.org/#/c/276335/
15:34:30 <ihrachys> have a look, complain, review ;)
15:34:35 <rossella_s> oh yes
15:34:46 <ihrachys> note it's ops visible, not dev docs
15:34:52 <rossella_s> we are such a productive team :)
15:35:08 <korzen> I would like to start the 'create an hook to modify the object field before writing in the DB' if anyone else is not interested
15:35:12 <mhickey> rossella_s: not bug deputy week! :)
15:35:35 <korzen> as it is in the backlock
15:35:40 <korzen> log*
15:36:19 <ihrachys> korzen: not sure what's that. but I guess I just need to get my hands dirty with subnet patch to understand the context.
15:36:39 <rossella_s> ihrachys, it's commented on the review
15:36:52 <ihrachys> right, I won't bother you explaining that to me now
15:37:01 <ihrachys> fyi the review mentioned is https://review.openstack.org/#/c/264273
15:37:16 <ihrachys> any other patches we should care about?
15:37:31 <korzen> ihrachys, yes, the case is that we do not have place to modify the params before pushing in to DB
15:38:01 <rossella_s> ihrachys, I had to refresh my memory about it..anyway korzen explained that
15:38:04 <korzen> it is needed when we have different OVO and DB params naming
15:38:26 <korzen> the thing is to get common method in base class
15:38:34 <ihrachys> oh ok I think I follow now
15:38:51 <korzen> and overdie it in child classes
15:39:10 <korzen> this would reduce the need to override the create,update,delete methods
15:39:58 <ihrachys> electrocucaracha: btw is the patch for sphinx integration for db models in review?
15:40:16 <ihrachys> ok I see it now
15:40:17 <ihrachys> #link https://review.openstack.org/#/c/274874/
15:40:28 <ihrachys> need to add to the wiki
15:40:33 * ihrachys will do it later
15:40:54 <ihrachys> ok I guess no more patches
15:40:56 <ihrachys> let's move on
15:40:59 <ihrachys> #topic Open discussion
15:41:03 <electrocucaracha> yes, but I have concerns about the pydot dependency that doesn't support py34
15:41:26 <ihrachys> electrocucaracha: but docs job is py27 only, right?
15:41:35 <electrocucaracha> yes
15:41:47 <ihrachys> electrocucaracha: then what's the concern, for now?
15:42:19 <ihrachys> oh I think I know what it is... is it because the dep is part of test-requirements?
15:42:24 <electrocucaracha> it's just that it's displayed in gerrit
15:43:18 <ihrachys> what do you mean?
15:44:38 <electrocucaracha> I was thinking that maybe that could be a roadblock for merging that change... but if it's ok for me it's better
15:45:01 <electrocucaracha> in the other hand I'm looking for alternatives like pydot3
15:45:14 <electrocucaracha> but I saw that they have some issues
15:45:29 <electrocucaracha> https://github.com/log0/pydot3/issues/1
15:45:34 <ihrachys> electrocucaracha: I think yes, since we have it in test-requirements.txt, it may be an issue
15:45:46 <ihrachys> but you can try to add a python27 marker to your dep
15:46:40 <ihrachys> can't find an example just now, but will give you one later.
15:47:08 <electrocucaracha> perfect, I didn't know that exists something to difference them
15:47:10 <rossella_s> I wanted to raise another point...maybe we should create a topic branch to speed things up
15:47:18 <ihrachys> let's get it solved off the meeting.
15:47:31 <ihrachys> rossella_s: all for it. suggestions?
15:47:46 <ihrachys> "ovo"? :) nice and short and cryptic!
15:47:54 <rossella_s> ihrachys yeah!
15:48:17 <rossella_s> ihrachys, actually I wanted to ask you if you see downsides...you did that for qos right?
15:48:23 <ihrachys> thoughts folks?
15:48:31 <ihrachys> rossella_s: yeah we had a topic for qos. worked smooth.
15:48:38 <rossella_s> ihrachys, cool
15:48:43 <rossella_s> so we can get things moving
15:48:44 <ihrachys> especially since we needed to track stuff in other repos too
15:49:01 <ihrachys> and then we can deprecate the list of patches on wiki
15:49:18 <rossella_s> yeah
15:49:22 <ihrachys> I guess everyone is for it? :)
15:49:28 <korzen> agree
15:50:01 <mhickey> agree
15:50:03 <ihrachys> ok, then let's use 'ovo' everywhere.
15:50:52 <ihrachys> the topic is not currently used by anyone in the whole openstack world, so should be fine
15:51:07 <rossella_s> haha we are original
15:51:16 <ihrachys> ok, one more thing I wanted to cover is... the code sprint.
15:51:40 <rossella_s> ihrachys, a lazy question regarding the sprint...any idea regarding the daily rate of the hotels more or less? otherwise I will check myself
15:51:54 <ihrachys> it's due in a month, so would be cool if we start thinking of it, booking hotels and stuff. please notify me when you are 100% going, or if you have questions.
15:52:22 <ihrachys> rossella_s: I suspect it to be quite low, like 50-60 USD, but I will check for you later.
15:52:39 <rossella_s> ihrachys, no worries, I can check myself
15:52:57 <korzen> like I said, flight and hotel booked ;)
15:53:07 <ihrachys> korzen: that's cool.
15:53:16 <mhickey> tentative
15:53:27 <rossella_s> I will write today to get the authorization
15:53:32 <rossella_s> it should be fine
15:53:49 <ihrachys> korzen: what's the price for you?
15:54:06 <korzen> 90 Euros per day in Continental?
15:54:21 <ihrachys> oh ok. haven't actually booked anything here myself :)
15:54:27 <rossella_s> hahah
15:54:38 <ihrachys> probably was more optimistic than it's the case :)
15:54:42 <rossella_s> ihrachys, one thing you could check it's the special rate for RH
15:54:45 <korzen> the Barcelo was 130Euro
15:54:47 <mhickey> ihrachys: are you not near the officve; I thought we would have house party! :)
15:55:16 <ihrachys> right. I may check on the price. let me talk to a local girl later in the week and get back to you if there are better offers.
15:55:57 <rossella_s> ihrachys, thanks
15:56:18 <korzen> I hope the public transport is ok in Brno
15:56:21 <ihrachys> ok I guess that's about it. keep up the good job. we still bootstrap things for objects, but definitely progress is here, and we'll be in good shape in a month.
15:56:29 <ihrachys> korzen: it is, absolutely
15:56:49 <ihrachys> kudos!
15:56:50 <rossella_s> yay
15:56:54 <ihrachys> #endmeeting