09:00:07 <anteaya> #startmeeting nova-net-to-neutron-migration
09:00:08 <openstack> Meeting started Tue Feb 10 09:00:07 2015 UTC and is due to finish in 60 minutes.  The chair is anteaya. Information about MeetBot at http://wiki.debian.org/MeetBot.
09:00:10 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
09:00:12 <openstack> The meeting name has been set to 'nova_net_to_neutron_migration'
09:00:16 <anteaya> hello hello
09:00:17 <jlibosva> hello
09:00:18 <obondarev> hi
09:00:21 <spandhe> hello!
09:00:32 <anteaya> yay spandhe is here!
09:00:33 <belmoreira> hi
09:00:37 <anteaya> great
09:00:47 <spandhe> :)
09:00:49 <anteaya> jlibosva obondarev belmoreira nice to see you
09:00:55 <anteaya> gus: are you about?
09:01:19 <anteaya> let's get going
09:01:28 <anteaya> #topic the state of the Neutron spec (obondarev)
09:01:41 <obondarev> so the first part of the spec was merged which is great
09:01:41 <anteaya> obondarev: you're up
09:01:46 <anteaya> yay team!
09:01:48 <obondarev> and means we're moving forward
09:02:01 <spandhe> great!
09:02:03 <anteaya> nice work everyone
09:02:04 <obondarev> the second part lacks reviews a bit
09:02:17 <obondarev> but also lacks some details of the process
09:02:19 <gus> (hi)
09:02:41 <obondarev> we can wait for comments to iterate further
09:02:48 <anteaya> nice summary
09:02:49 <obondarev> or just start adding details
09:02:59 <obondarev> anteaya: thanks
09:03:17 <anteaya> so yes, congratulations to everyone for their hard work in getting the first patch to the spec merged
09:03:22 <anteaya> well done!
09:03:49 <anteaya> we can address next steps for the merged portion in the documentation topic
09:03:53 <gus> I think a lot of people weren't looking at the second until the first was submitted, so I think (hope) a bunch of those will jump over to the second patch now.
09:03:59 <anteaya> now for the second part
09:04:11 <anteaya> gus: that hopefully is the case
09:04:32 <obondarev> I hope so too
09:04:39 <anteaya> at the nova mid-cycle some folks stated that they need to see proof of concept code to review the second part of the spec
09:05:19 <anteaya> so my feeling right now is that the two patches we have up, obondarev's and jlibosva's, will be reviewed first before the second part of the spec is reviewed
09:05:36 <anteaya> does that make sense to folks?
09:05:58 <obondarev> anteaya: it does
09:06:05 <anteaya> great
09:06:07 * jlibosva nods
09:06:11 <gus> yep.
09:06:15 <obondarev> also I think one major thing that is not well detailed in the spec is about working in mixed nova-net/neutron env
09:06:27 <anteaya> so shall we move topics to implementation and see where we are on those patches?
09:06:37 <anteaya> obondarev: ah okay go ahead with your point
09:06:41 <obondarev> where internal and external connectivity should be kept for both types of hypervisors
09:06:57 <obondarev> currently we have "Bring up neutron network (l3) nodes" as the final step
09:07:10 <obondarev> but I'm not sure how to provide external connectivity for "has_transitioned" hypervisors then
09:07:17 <obondarev> while we're in the middle of the big migration loop
09:07:36 <obondarev> I'll add comments on this to the review
09:07:54 <obondarev> and I can start to look into this
09:08:06 <anteaya> does anyone else have any feedback on obondarev's point?
09:08:13 <gus> The routers (or gateways I think nova calls them) should have the same mac+ip address, and the wire format is compatible (because we chose it that way).
09:08:54 <gus> so the idea was the nova gateways would stay up until all the hosts are transitioned across, and then in that latter (poorly defined atm) step we would switch them out for neutron routers.
09:09:09 <gus> I don't have good suggestions for how we actually do that switch :/
09:09:23 <obondarev> gus: I see
09:09:52 <obondarev> this needs some experiments/prototyping as well I guess
09:10:02 <gus> yes, definitely.
09:10:42 <gus> obondarev: does that seem reasonable?
09:11:05 <obondarev> gus: yes, so let's get initial steps done then
09:11:28 <gus> yep.
09:11:32 <anteaya> obondarev: thanks for raising the point
09:11:49 <anteaya> obondarev: you are able to capture these thoughts in a comment on the patch?
09:12:03 <obondarev> anteaya: I will
09:12:10 <anteaya> obondarev: thank you
09:12:25 <anteaya> any objection to moving to a discussion of implementation?
09:12:41 <anteaya> #topic the state of implementation (obondarev)
09:12:45 <anteaya> here we are
09:12:49 <anteaya> obondarev: to you
09:12:57 <obondarev> sure
09:13:10 <obondarev> jlibosva: can you please update on db migration?
09:13:44 <anteaya> we seem to lose him
09:13:45 <jlibosva> well, I had some unplanned events last week so I didn't do much work on it (my apologies)
09:13:50 <anteaya> ah here we are
09:14:01 <anteaya> jlibosva: what are your current thoughts?
09:14:13 <jlibosva> I did some work yesterday mostly about creating networks out of floating ips pools
09:14:41 <obondarev> jlibosva: do you mean external networks?
09:14:50 <jlibosva> yes
09:14:57 <obondarev> jlibosva: ok, cool
09:15:10 <anteaya> jlibosva: is that work reflected in the current patch?
09:15:11 <jlibosva> I'm now trying to find out where to get gateway for external subnets
09:15:23 <jlibosva> anteaya: no, it's only local. not sent to review yet
09:15:42 <anteaya> jlibosva: can we try to get in the habit of submitting all the work
09:15:46 <anteaya> even if it doesn't work yet
09:16:01 <anteaya> it helps folks see what you mean when you bring it up in discussion
09:16:10 <jlibosva> anteaya: yes, sorry
09:16:25 <anteaya> jlibosva: it is okay, it has to become a habit, which takes time
09:16:28 <anteaya> and thank you
09:16:33 <anteaya> you were saying
09:17:25 <anteaya> you are working on getting gateways for external subnets
09:17:33 <obondarev> jlibosva: you can raise you concerns on this in the ML
09:17:59 <jlibosva> obondarev: good idea, I will try to look more on the code and then I'll send an email
09:18:02 <anteaya> the gateways would come from where?
09:18:03 <anteaya> nova?
09:19:04 <jlibosva> well, the only source of information I have is nova database. And I believe the gateway of exeternal network should be the gateway on the physical network
09:19:37 <anteaya> hmmm, so that would be different for each operator potentially?
09:20:19 <anteaya> or you would have to access the nova database?
09:20:33 <jlibosva> I think so. We could provide it as a parameter. But now it's getting more and more parameters so maybe some config file would make better sense
09:20:46 <jlibosva> that was for previous answer
09:20:49 <anteaya> I do like config files
09:20:56 <jlibosva> during migration nova db is already accessed
09:21:10 <anteaya> if you went to the mailing list, who would you be hoping would respond?
09:21:13 <belmoreira> jlibosva: in the case of using a "flat network" configuration with linux bridge, the gateway configuration in nova-network is not needed since it will use the compute node configuration.
09:21:59 <jlibosva> belmoreira: right, but the migration should cover all network managers
09:22:19 <anteaya> I thought our use case was the flat network
09:22:24 <anteaya> am I wrong?
09:22:47 <jlibosva> really? I'm doing all of them
09:22:58 <anteaya> or the use case we are addressing our work to
09:23:02 <anteaya> then this is a good question
09:23:12 <anteaya> gus obondarev can you share your thoughts?
09:23:12 <obondarev> anteaya: I think we should support all of them
09:23:20 <jlibosva> in the meantime - I sent my current work to gerrit
09:23:24 <anteaya> gus: ?
09:23:26 <gus> the hope was anything that has a (direct) equivalent in neutron.  Obviously if there's a simpler combination we should do that first.
09:23:29 <anteaya> jlibosva: thank you
09:23:47 <anteaya> is the flat network the simplier choice?
09:24:10 <jlibosva> yes
09:24:23 <anteaya> then my vote is for the flat network first
09:24:34 * gus agrees.
09:24:47 <obondarev> "This process supports any nova-network and neutron deployment options which are wire-compatible"
09:24:58 <jlibosva> My opinion is that from db perspective the change between flat and vlan wouldn't be large. I might be wrong though
09:24:59 <obondarev> this is from the merged part of the spec
09:25:04 <gus> At this point it's important to get an end-to-end proof-of-concept working, even if we have to leave a few TODOs around.
09:25:27 <obondarev> gus: agree
09:25:33 <belmoreira> gus: +1
09:25:44 <anteaya> let's get somethign we can test
09:26:04 <jlibosva> ok
09:26:20 <anteaya> and yes, thanks obondarev for finding that from the spec, we have to acknowledge we have a todo here to expand once we get something working
09:26:34 <anteaya> fair enough?
09:26:39 <jlibosva> belmoreira: so when using flat network, I assume every fip pool will be tied to a bridge and thus I don't need to care about gateways
09:26:42 <jlibosva> is that correct?
09:26:55 <obondarev> anteaya: agree
09:27:24 <belmoreira> jlibosva: I believe so
09:27:26 <jlibosva> thanks
09:27:57 <anteaya> jlibosva: do you have what you need to continue working for now?
09:28:18 <jlibosva> anteaya: I'm short on time :) Otherwise, yes
09:28:30 <anteaya> jlibosva: understood and thank you
09:28:40 <anteaya> obondarev: do you want to take your turn now?
09:28:43 <obondarev> moving on then
09:28:48 <anteaya> great thanks
09:28:48 <obondarev> anteaya: yep
09:28:53 <obondarev> speaking of nova-net proxy
09:29:03 <obondarev> I've put a couple on new revisions on gerrit
09:29:21 <obondarev> # link https://review.openstack.org/#/c/150490/
09:29:37 <obondarev> got some general concerns for dansmith
09:29:45 <anteaya> I think you need to remove the whitespace for #link to work
09:29:57 <obondarev> anteaya: yep, sorry
09:30:03 <anteaya> np, try again?
09:30:06 <obondarev> #link https://review.openstack.org/#/c/150490/
09:30:09 <anteaya> thanks
09:30:35 <obondarev> so I'd really like to sort them before moving forward
09:30:42 <obondarev> I mean Dan's comments
09:30:42 <anteaya> let's take a look at the latest patchset shall we?
09:30:50 <obondarev> so I replied to comments and waiting for dansmith to review
09:30:54 <anteaya> obondarev: I feel the same
09:31:02 <anteaya> let's see if we can discuss the comments
09:31:11 <anteaya> and understand them a bit better
09:31:14 <obondarev> sure
09:31:20 <anteaya> to the patch!
09:32:52 <obondarev> so I guess the main concern is what is nova-network proxy and whether we need it at all
09:33:10 <anteaya> well if that is what the concerns boil down to
09:33:20 <anteaya> that is a concern that needs to be addressed
09:33:27 <anteaya> any input from others
09:33:42 <anteaya> do you feel obondarev's assessment of the comments is accurate?
09:34:32 <gus> right, I don't think Dan understands the need for a step between "store state in neutron DB" and "all compute nodes have migrated"
09:34:50 <gus> .. so we should perhaps explain that in the review.
09:35:02 <anteaya> that might be a good beginning
09:35:07 <obondarev> or move the discussion to the spec maybe?
09:35:30 <anteaya> let's keep the discussion in the proxy patch for now
09:35:42 <anteaya> it helps retain the history of the conversation
09:35:55 <anteaya> the spec can have a comment that references the proxy patch
09:36:02 <gus> yeah, replying on the review keeps the response easily correlated.
09:36:17 <obondarev> ok
09:36:19 <anteaya> actually the spec should have a comment that references both the proxy patch and the db migration patch
09:36:28 <gus> obondarev: Do you want me to respond to Dan, or shall we see what he comes back with after you existing response?
09:36:43 <anteaya> I would like to see some movement here
09:36:48 <anteaya> other than just wait
09:36:53 <obondarev> gus: let's wait I think
09:36:57 <gus> ;)
09:37:16 <anteaya> we have a status update at the cross project meeting today
09:37:28 <anteaya> many folks will be reading these patches for the first time as a result
09:37:37 <anteaya> let's have the latest information up for them please?
09:37:49 <anteaya> so if we think there is a hole in an explaination
09:37:53 <anteaya> can we fill it?
09:37:56 <belmoreira> but the proxy is a mandatory step in the migration? If a deployment just migrate the DB and is fine with a small downtime my understanding is that it can avoid the proxy.
09:38:04 <anteaya> otherwise I will just need to keep explaining it
09:38:21 <obondarev> anteaya: well the rationale is in the spec
09:38:43 <gus> anteaya: ack.  I'll add a short comment on the change so we're sure that aspect isn't left hanging.
09:38:49 <gus> belmoreira: yes, exactly.
09:39:04 <anteaya> obondarev: true, but folks can trip over any percieved gap
09:39:08 <anteaya> gus: thank you
09:39:41 <obondarev> anteaya: undestood
09:39:54 <anteaya> obondarev: thanks :)
09:40:01 <anteaya> any more comments here?
09:40:29 <anteaya> let's move on
09:40:47 <anteaya> #topic documentation (obondarev)
09:41:01 <anteaya> obondarev: :)
09:41:15 <obondarev> yep
09:41:30 <obondarev> no updates from me on this one I'm afraid
09:41:35 <anteaya> okay
09:41:45 <anteaya> we need to make some contact with documentation folks
09:41:57 <anteaya> edgar said he would do the work, but he can't attend the meetings
09:42:07 <anteaya> which is tough as we lose that communication
09:42:29 <anteaya> I'll keep working on finding someone who can attend meetings who can work with edgar on documentation
09:42:47 <obondarev> last update from him was that he can start work on docs based on the specs
09:43:02 <anteaya> okay that is great
09:43:15 <anteaya> do we know where to look yet to track his progress?
09:43:37 <obondarev> he said he'll send the link once he has patch
09:43:39 <anteaya> edgar if you are reading the logs, can you tell me where to look to see what you are doing?
09:43:52 <anteaya> okay great, I'm looking forward to that
09:44:06 <anteaya> we are working towards having something up for the operators mid-cycle
09:44:10 <anteaya> #link https://wiki.openstack.org/wiki/Operations/Meetups
09:44:30 <anteaya> it takes place in the second week of march
09:44:35 <anteaya> which doesn't leave much time
09:44:46 <anteaya> okay so let's move on?
09:45:03 <anteaya> since we can't do much more without someone from docs to help us here
09:45:18 <anteaya> #topic testing (belmoreira)
09:45:35 <anteaya> belmoreira: how far away are we from having something to test?
09:45:44 <belmoreira> I don't have any update.
09:45:53 <anteaya> belmoreira: you said you think we can test the db migration in isolation of the proxy code?
09:45:59 <anteaya> belmoreira: fair enough
09:46:03 <belmoreira> we should wait for the proxy and db migration developments
09:46:18 <anteaya> but if we can test the db migration code in isolation
09:46:30 <anteaya> can we consider that as somethign testable?
09:47:44 <jlibosva> I haven't tried to run Neutron after db migration, just observed the db. But I believe networks, subnets, ports and security groups should be migrated successfully
09:48:12 <anteaya> can we consider that as something to try?
09:48:41 <belmoreira> we will test the DB migration tool against our DB. However before we need to perform changes in order to work with our setup. So I will wait for a more stable version
09:48:59 <jlibosva> not with current patchset as I sent what I had without trying to run it
09:50:01 <anteaya> okay great
09:50:14 <anteaya> so what do we need for it to be considered more stable?
09:50:59 <jlibosva> I'll run the script and once it passes my "small environemnt" migration, I'll send new PS
09:51:04 <belmoreira> anteaya: I mean with more review iterations
09:51:39 <anteaya> belmoreira: if it passes jlibosva's small environment, will you consider it sufficient to test?
09:51:58 <anteaya> belmoreira: I'm trying to find out what you need to see, so I can help you get it
09:53:26 <belmoreira> anteaya: I know. Because we need to work on top of it that's why I'm waiting for more iterations. Fair?
09:53:48 <anteaya> well if I knew what you were looking for from more iterations taht would help
09:54:02 <anteaya> since number of patchsets doesn't really convey stability to me
09:54:27 <anteaya> I understand you want something better, I just don't know what better looks like to your eyes
09:55:15 <anteaya> so let's see if we can get jlibosva's patchset into a shape that belmoreira feels comfortable testing
09:55:23 <anteaya> is that a reasonable goal?
09:55:58 <anteaya> jlibosva: can you work with belmoreira to make that happen?
09:56:00 <jlibosva> I still don't understand, when to say "now it's in the shape for testing"
09:56:03 <belmoreira> anteaya: I will look into the last patchset
09:56:15 <gus> belmoreira: fwiw, the tool only needs read-only access to your current tables (or could work off a snapshot)
09:56:19 <anteaya> belmoreira: thank you, that would be a big help
09:56:30 <anteaya> jlibosva: me either
09:56:44 <jlibosva> belmoreira: you can also do db dump, restore db somewhere in testing environment and test db migration
09:57:07 <anteaya> belmoreira: or is there more work for you to test this than we currently understand?
09:57:09 <gus> but it seems premature to test it out on a large dataset before jlibosva has declared it "plausible"
09:57:26 <anteaya> gus: yes to be sure
09:57:50 <jlibosva> I can confirm now that current PS crashes
09:58:03 <belmoreira> my point is that the test is very interesting for us... for this exercise maybe less because our DB maybe will not fit in what is expected by the script...
09:58:14 <anteaya> jlibosva: so a non-crashing patchset, that is a worthy metric
09:58:31 <anteaya> ah so you don't want to test this, belmoreira?
09:58:33 <belmoreira> or in a standard nova-network configuration
09:58:38 <anteaya> then I made a poor assumption
09:58:56 <anteaya> spandhe: do you have any ability to help us test the db migration patchset?
09:59:06 <gus> belmoreira: ah I see, so you're expecting to use some customised version of this migration tool?
09:59:31 <anteaya> so time to wrap up
09:59:33 <spandhe> anteaya: I think I can..
09:59:40 <anteaya> spandhe: oh that would be great
09:59:47 <anteaya> spandhe: can you work with jlibosva this week?
09:59:55 <anteaya> see what you can accomplish?
10:00:03 <jlibosva> that would be good
10:00:12 <anteaya> and we need to wrap up
10:00:16 <spandhe> sure.. jlibosva.. shall we talk again later? what timezone ar u in?
10:00:19 <belmoreira> anteaya: I really want to test it... and work on top of it... I can give feedback however our DB as some particularities that will not fit the generic test that is expected.
10:00:21 <anteaya> another great meeting folks, thank you
10:00:25 <obondarev> thanks, bye everyone
10:00:30 <jlibosva> thanks, bye
10:00:36 <gus> belmoreira: ack, understood.
10:00:37 <anteaya> belmoreira: I made assumptions I shouldn't have, sorry about that
10:00:39 <gus> thanks all.
10:00:43 <belmoreira> thanks
10:00:55 <anteaya> and show up to the cross-project meeting today if you can
10:01:07 <anteaya> I'll be giving a status update and welcome your input
10:01:15 <anteaya> thanks for your hard work everyone
10:01:19 <anteaya> #endmeeting