Monday, 2015-06-15

*** amotoki has joined #openstack-neutron-release03:56
*** amotoki has quit IRC04:04
*** amotoki has joined #openstack-neutron-release04:08
*** amotoki has quit IRC04:31
*** yamamoto has joined #openstack-neutron-release13:08
*** yamamoto has quit IRC13:20
*** yamamoto has joined #openstack-neutron-release13:37
*** amuller has joined #openstack-neutron-release13:54
*** yamamoto has quit IRC14:18
*** carl_baldwin has joined #openstack-neutron-release14:53
* mestery stumbles in 6 minutes before the first official "Neutron PTL / Lieutenant Office Hours" begins14:54
*** carl_baldwin has quit IRC14:57
mesteryWow, I scared carl_baldwin away!14:57
amullerheh14:59
mesterySo, office hours begin!15:00
mesteryThere is really no agenda here per se, just a sync point.15:00
mesteryamuller: How goes the fullstack testing and merges?15:00
*** carl_baldwin has joined #openstack-neutron-release15:00
mesterycarl_baldwin: You're back! :)15:01
amullermestery: I have a local patch that I've yet to push that generalizes the multinode stuff15:01
*** amotoki has joined #openstack-neutron-release15:01
mesteryamuller: That's pretty huge!15:01
amullerI got some comments from Cedric and Kuba15:01
mesterynice15:01
mesteryamuller: Will it be enabled as part of some job upstream, or is that the next step?15:01
* carl_baldwin wonders what he missed.15:01
amullerthe fullstack tests are being skipped right now due to DB mysql tests are being skipped15:02
mesterycarl_baldwin: Nothing yet :)15:02
amullerthe DB tests in the functional job15:02
mesteryamuller: Ack15:02
amullerCedric has a patch for that15:02
amullerwe can't merge any fullstack stuff before that15:02
mesterySpeaking of CI jobs ...15:02
amulleror we might merge stuff that when it gets unskipped it will fail15:02
mesteryI had a conversation with boris-42 this morning, the multiple API workers patch armax had broke the rally jobs due to their high concurrency rates15:02
mesteryOnce we stabilize that, we shoudl look at making rally jobs voting15:02
mesterythoughts?15:02
amullerI'm not familiar enough with Rally to know what how do they decide to fail a test15:03
amullerbut I think that performance testing in general should never fail. Rather, it should output a historical graph (Say, 3 months) of data15:03
amullerso you can see trends and blame specific patches15:03
mesteryamuller: Those semantics are hazy to me as well, all I know is they've been first to find some issues lately (pymsql as well as the API Works issues)15:03
mesteryamuller: Makes sense to me, +115:04
amullerBoris is already on that btw I talked to him, about the historical graphs15:04
mesterycool15:05
mesterythanks amuller15:05
amullerOnce the full stack tests are ready on my side I'll probably ask HenryG to take a look15:05
amullerand we can start merging them15:05
amullermestery: btw all of the full stack tests are voting as part of the functional job15:05
mesteryamuller: Excellent!15:05
amullerExcellent but also dangerous :)15:05
mesteryheh :)15:06
*** carl_baldwin has quit IRC15:07
HenryGamuller: I will be happy to review15:08
amullerHenryG: Cool, I'll let you know when15:08
*** carl_baldwin has joined #openstack-neutron-release15:08
amullerHenryG: still iterating on them15:08
HenryGamuller: np, I have enough to keep me busy :)15:08
amullerHenryG: Oh you're not actively looking for more work? ;-)15:09
HenryGamuller: on the contrary :)15:09
HenryGamuller: In the not-too-distant future I would like to see the API tests using fullstack15:10
amullerHenryG: ahh, that's an interesting one15:10
HenryGGet away from depending on a running devstack15:10
amullerso right now we tear down and start up all of the processes for *every* test15:10
amullerso on a VM of mine each test takes a minimum of 15 seconds15:11
amullerI think that we'll want, for example, all tests that belong to the same class to reuse the same environment/processes15:11
mesteryamuller: ++ to that15:11
HenryGSure, I understand there will be some challenges.15:12
amullerso it'll take a while before we can consider the full stack framework to be ready for something like the API tests15:12
HenryGthat's fine15:12
HenryGBut let's keep it in our sights15:12
amullerright now my goals are to unskip them, get the clean ups right (Both of these items are being handled by Cedric thankfully), get them ready to test L3 HA + l2pop, DVR, DVR + L3 HA15:12
amullerHenryG: ack15:13
HenryGamuller: I think Cedric's unskip patches are ready to go15:16
mesteryamuller HenryG: This spec is re-proposed from Kilo, could you eyeball it for me? https://review.openstack.org/#/c/185765/15:17
mesteryIt's just a re-propose.15:17
* HenryG looks15:17
mesteryAlso, would be great to have folks look at the "unaddressed interfaces" spec from Isaku: https://review.openstack.org/#/c/183125/15:17
*** yamamoto has joined #openstack-neutron-release15:18
amullermestery: I'll ask ajo to look at the service group thingie15:19
amullerlooks like it might have to do with QoS and filtering rules they wanted to do there15:19
mesterySpreading the spec review load ... carl_baldwin, any interest in looking at a "make octavia the default open source lbaas backend" spec? https://review.openstack.org/#/c/174616/15:19
mesteryThis was discussed at the sumit.15:19
mesteryamuller: It's somewhat orthogonal, but yes, that woudl be great!15:19
carl_baldwin mestery:  I missed the discussion at summit unfortunately.  But, I can fit it in if you need it.15:20
mesterycarl_baldwin: Well, no worries.15:20
mesteryWe have a giant backlog of specs I've been slowly peeling through.15:20
amullermestery: There were talks about defining types of traffic as a first class citizen that could be reused by QoS and tap as a service15:20
mesteryWe have no deadlines per se, but I'd like to settle on these before Liberty-1 if possible, especially not filed as RFEs.15:20
mesteryamuller: Yes, I see that!15:20
mesteryamuller: I mean, I recall that discussion15:20
carl_baldwinmestery: Don’t be hesitent to throw some my way.  I’ll have a look at this one.15:22
mesterycarl_baldwin: thanks!15:22
mesteryRegarding "VLAN aware VMs" (https://review.openstack.org/#/c/94612/), a meeting has been scheduled for tomorrow at 1700UTC in #openstack-meeting-4.15:23
*** yamamoto has quit IRC15:23
HenryGmestery: For the service group spec, can we add APIImpact to the commit msg? I don't know if the api-wg looked at it during kilo, but they should now.15:24
mesteryHenryG: Makes sense to me.15:24
HenryGI'll just do it right in gerrit15:24
mesteryHenryG: cool!15:24
HenryGdone15:27
carl_baldwinmestery: dougwig: I’m looking at https://review.openstack.org/#/c/174616/1/specs/liberty/lbaas-ref-octavia.rst16:07
carl_baldwinI have two questions.16:07
mesterycarl_baldwin: please, go ahead16:08
carl_baldwinFirst, I don’t have a sense for how ready Octavia is to fill the role of default reference implementation.  I see a few +1s from some lbaas guys.  Is it pretty generally agree that it is ready?16:08
dougwigcarl_baldwin: it's not ready now, it should be by the end of liberty.  if it's not ready, we will not flip the switch to make it the ref.16:09
carl_baldwinSecond, is there a deprecation plan for the existing implementation?  The spec says it will stay in and deployers will choose to change.  Will there be a deadline to move?16:09
mesterycarl_baldwin: Yes, deprecated in liberty removed in Mxxx16:09
mesteryassuming octavia is ready16:09
mesterydougwig: ^^^^16:09
dougwigcarl_baldwin: that's kinda hand-waivey because we don't know of anyone actually using the ref in the production, because it's that awful. so this isn't a question of moving from something working to something better.  :)16:10
carl_baldwinAh, so hence the “Dependencies:  A working Octavia implementation"16:10
dougwigso we'll deprecate, continue to support, but i'm guessing no one will notice.16:10
carl_baldwindougwig: Okay.16:10
carl_baldwindougwig: I gave -1 but I don’t think it will take much change to get my +2 on it.  Take a look.16:17
dougwigcarl_baldwin: ok16:17
dougwigthanks16:17
carl_baldwinMostly, I think we should say explicitly that the current impl will be removed if that’s what we intend to do.  Just so there aren’t any surprised.16:18
mesterycarl_baldwin: you convinced me to go -1 as well :)16:30
mesteryBut like you, I could easily sway back to +2/+A16:30
*** amuller is now known as amuller_afk17:11
*** amuller_afk is now known as amuller17:51
*** amuller is now known as amuller_afk18:36
*** amotoki has quit IRC19:33
*** amuller_afk is now known as amuller19:57
* mestery is here for office hours if anyone wants to talk20:00
amullerHenryG: Rossella's OVS functional tests patch: https://review.openstack.org/#/c/140042/20:29
amullergave it a +2 after I worked with her a bit20:30
HenryGamuller: I "put it in a tab" (tm), but I need to step out for a bit20:31
amullerHenryG: awesome, thanks :)20:32
amullerHenryG: it'll unblock some of her work20:32
*** yamamoto has joined #openstack-neutron-release20:45
*** yamamoto has quit IRC20:49
*** amuller has quit IRC22:27
*** yamamoto has joined #openstack-neutron-release23:23
*** yamamoto has quit IRC23:27
*** yamamoto has joined #openstack-neutron-release23:54

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!