Tuesday, 2016-02-02

*** manand has quit IRC00:03
*** salv-orl_ has quit IRC00:08
*** gangil has quit IRC00:08
*** dlundquist1 has joined #openstack-neutron-ovn00:08
*** dlundquist has quit IRC00:11
*** roeyc has joined #openstack-neutron-ovn00:14
*** gangil has joined #openstack-neutron-ovn00:23
*** gangil has joined #openstack-neutron-ovn00:23
*** dlundquist1 has quit IRC00:26
openstackgerritMatthew Kassawara proposed openstack/networking-ovn: Docs: Implement doc8 linter  https://review.openstack.org/27494800:27
*** azbiswas has joined #openstack-neutron-ovn00:56
*** gangil has quit IRC01:15
*** gangil has joined #openstack-neutron-ovn01:15
*** gangil has joined #openstack-neutron-ovn01:15
*** chandrav has quit IRC01:15
*** azbiswas has quit IRC01:54
*** fzdarsky_ has joined #openstack-neutron-ovn02:03
*** azbiswas has joined #openstack-neutron-ovn02:07
*** fzdarsky|afk has quit IRC02:07
*** roeyc has quit IRC02:09
*** chandrav has joined #openstack-neutron-ovn02:14
openstackgerritMerged openstack/networking-ovn: Add msec resolution to ovn-northd console logs.  https://review.openstack.org/27487602:22
*** s3wong has quit IRC02:27
*** gangil has quit IRC02:28
*** chandrav has quit IRC02:39
*** azbiswas has quit IRC02:40
*** chandrav has joined #openstack-neutron-ovn02:49
*** thumpba has quit IRC02:51
*** chandrav has quit IRC03:00
*** yamamoto has joined #openstack-neutron-ovn03:09
*** jckasper has joined #openstack-neutron-ovn03:14
*** allan_h has quit IRC03:15
*** gangil has joined #openstack-neutron-ovn03:16
*** gangil has quit IRC03:16
*** gangil has joined #openstack-neutron-ovn03:16
*** azbiswas has joined #openstack-neutron-ovn03:18
*** yamamoto has quit IRC03:20
*** manand has joined #openstack-neutron-ovn03:32
*** azbiswas_ has joined #openstack-neutron-ovn03:48
*** azbiswas has quit IRC03:52
*** armax has quit IRC04:02
*** armax has joined #openstack-neutron-ovn04:06
*** yamamoto_ has joined #openstack-neutron-ovn04:08
*** azbiswas_ has quit IRC04:13
*** manand has quit IRC04:47
*** chandrav has joined #openstack-neutron-ovn04:51
*** azbiswas has joined #openstack-neutron-ovn04:53
*** azbiswas has quit IRC04:54
*** gongysh has joined #openstack-neutron-ovn05:00
*** allan_h has joined #openstack-neutron-ovn05:00
*** allan_h has quit IRC05:32
*** gongysh has quit IRC06:04
*** chandrav has quit IRC06:22
*** armax has quit IRC07:03
*** palexster has quit IRC08:22
*** palexster has joined #openstack-neutron-ovn08:26
openstackgerritNuman Siddique proposed openstack/networking-ovn: Create periodic status task to check dhcp agents in Ovn Worker  https://review.openstack.org/27486408:56
*** numans has joined #openstack-neutron-ovn08:57
*** gangil has quit IRC09:00
*** roeyc has joined #openstack-neutron-ovn09:29
*** gongysh has joined #openstack-neutron-ovn09:56
*** openstackgerrit has quit IRC10:17
*** openstackgerrit has joined #openstack-neutron-ovn10:17
openstackgerritBabu Shanmugam proposed openstack/networking-ovn: DPDK support for OVN  https://review.openstack.org/27510310:20
*** gongysh has quit IRC10:25
*** yamamoto_ has quit IRC10:37
openstackgerritBabu Shanmugam proposed openstack/networking-ovn: Removed unnecessary code from the plugin  https://review.openstack.org/27512511:07
*** roeyc has quit IRC11:13
*** fzdarsky_ has quit IRC11:47
*** fzdarsky has joined #openstack-neutron-ovn11:47
*** roeyc has joined #openstack-neutron-ovn12:08
*** yamamoto has joined #openstack-neutron-ovn12:15
*** yamamoto has quit IRC12:22
*** yamamoto has joined #openstack-neutron-ovn12:23
*** yamamoto has quit IRC12:27
*** numan_ has joined #openstack-neutron-ovn12:35
*** numans has quit IRC12:36
*** roeyc has quit IRC12:39
*** yamamoto has joined #openstack-neutron-ovn12:46
*** roeyc has joined #openstack-neutron-ovn12:49
*** yamamoto has quit IRC12:56
*** gongysh has joined #openstack-neutron-ovn13:25
*** yamamoto has joined #openstack-neutron-ovn13:27
*** yamamoto has quit IRC13:27
*** yamamoto has joined #openstack-neutron-ovn13:28
*** thumpba has joined #openstack-neutron-ovn13:42
*** regXboi has joined #openstack-neutron-ovn13:45
*** salv-orlando has joined #openstack-neutron-ovn14:01
*** thumpba has quit IRC14:06
*** jckasper has quit IRC14:18
*** jckasper has joined #openstack-neutron-ovn14:18
*** jckasper has quit IRC14:20
*** jckasper has joined #openstack-neutron-ovn14:21
openstackgerritMerged openstack/networking-ovn: Vagrant: Add docs for instance external net access  https://review.openstack.org/27485614:23
openstackgerritMatthew Kassawara proposed openstack/networking-ovn: Vagrant: Modify OpenStack services  https://review.openstack.org/27479514:31
Sam-I-Amrussellb: moo.14:43
Sam-I-Amrussellb: can you look at https://review.openstack.org/#/c/274795/ ?14:44
russellbI can.14:44
russellbbut will I?!14:44
Sam-I-Amif you're nice.14:44
Sam-I-Amonce that merges... and possibly this next patch i'm working on, i'm trying to make the mtu handed out by dhcp vary based on the mtu chosen for the vagrant vm net interfaces14:46
russellbSam-I-Am: does metadata work?14:50
Sam-I-Amrussellb: yes14:50
Sam-I-Amon both network types14:50
russellbthat's the first successful test report of metadata with OVN btw14:51
russellbdoes our base devstack setup need to get fixed to make that work too?14:51
russellbmaybe for your TODO :)14:52
Sam-I-Amrussellb: my guess is the dhcp option made it work, not something specific with md itself.14:53
* russellb nods14:53
Sam-I-Amif you dont do enable_isolated_metadata, it relies on the router for doing some iptables magic to proxy requests14:53
Sam-I-Amso, interesting note about enable_isolated_metadata14:53
Sam-I-Amdhcp is enabled (or not) at subnet creation. routers always come later. so that option always assumes your network as no router and enables the dhcp option.14:54
Sam-I-Amthe router md magic only comes into play if you disable that option and always have routers14:55
russellbSam-I-Am: commented on your patch14:58
Sam-I-Amrussellb: hmmmmmmmmmm15:00
Sam-I-Amrussellb: i guess that depends on the chances of the proxy magic ending up in the native l3 implementation15:00
russellbnot actively worked on15:00
russellband likely low priority15:00
Sam-I-Amand/or if the native dhcp agent (if that is such a thing) would handle enable_isolated_metadata15:01
russellbyeah, the case of no l3 agent and no dhcp agent is a bit up in the air15:01
* russellb punts that down the field15:02
Sam-I-Amok, so moving my dhcp thinger to the devstacks is probably better than having it in the vagrants15:02
Sam-I-Ami had it in the vagrants to test this specific case, but it does apply to a larger audience15:02
russellbk :)15:03
Sam-I-Amlet me fixerate, test, then beg for +2s15:03
*** thumpba has joined #openstack-neutron-ovn15:13
regXboirussellb: ping15:18
regXboirussellb: FYI, I'm trying to test out http://openvswitch.org/pipermail/dev/2016-January/064854.html and get some solid numbers on how it improves things in my tests - it looks like I'm going to have to add some instrumentation to measure loop times and loop counts as a simple view of cpu usage isn't enough to draw a conclusion :(15:20
regXboirussellb: thus I'm wondering - does a restack recompile ovn-controllerd even if I'm not recloning?15:21
*** chandrav has joined #openstack-neutron-ovn15:23
*** azbiswas_ has joined #openstack-neutron-ovn15:25
*** azbiswas_ has quit IRC15:28
russellbregXboi: it should recompile/reinstall yes15:29
regXboirussellb: thx15:29
russellbregXboi: my patch is only going to make a different in particular scenarios15:29
regXboirussellb: ack, I want to see if the one I'm testing is one of those scenarios15:30
russellbspecifically a large # of networks, where a given hypervisor only has ports on a smaller subset of those networks15:30
regXboiwell, my test should be in that mode15:30
russellbit cuts out some processing in that case, otherwise it won't make a difference15:30
regXboier, wait ... no I'm not15:30
regXboino wait, I'm confusing myself - I am :)15:30
regXboiI think :)15:30
regXboiwell, I guess the easiest way is to just put in the instrumentation and run the darn thing15:31
regXboiand then I'll know15:32
russellbOR maybe you'll come out with more questions15:34
*** chandrav has quit IRC15:39
*** numan_ has quit IRC15:43
*** roeyc has quit IRC16:03
*** yamamoto has quit IRC16:05
*** yamamoto has joined #openstack-neutron-ovn16:07
*** yamamoto has quit IRC16:08
openstackgerritRussell Bryant proposed openstack/networking-ovn: Add more thoughts to the HA section of the FAQ.  https://review.openstack.org/27484416:10
*** gongysh has quit IRC16:18
*** roeyc has joined #openstack-neutron-ovn16:19
*** armax has joined #openstack-neutron-ovn16:52
openstackgerritRussell Bryant proposed openstack/networking-ovn: Update the HA section of the FAQ.  https://review.openstack.org/27484417:06
*** Guest75006 has joined #openstack-neutron-ovn17:07
*** Guest75006 has quit IRC17:07
*** Guest75006 has joined #openstack-neutron-ovn17:07
*** Guest75006 is now known as arosen17:07
openstackgerritMatthew Kassawara proposed openstack/networking-ovn: Docs: Implement doc8 linter  https://review.openstack.org/27494817:07
Sam-I-Amrussellb: i'll be pondering your patch17:08
regXboiand I'm restacking on the other patch :)17:08
Sam-I-Ami'm working on elevendyseven patches at once17:09
russellbjoin the clubu17:10
openstackgerritMatthew Kassawara proposed openstack/networking-ovn: Vagrant: Share local directories  https://review.openstack.org/27533117:14
Sam-I-Am^ should be an easy one17:14
*** salv-orlando has quit IRC17:14
* russellb usually lets mestery review the vagrant changes first :)17:15
Sam-I-Amits slowly becoming more robust17:15
russellbreally great work17:16
Sam-I-Ammore or less simulating what we should test for prod environments... just on vms.17:16
russellbi really like having people that understand production environments helping :)17:16
Sam-I-Amonce those patches merge, and i fix that patch you commented on, i'm trying to add automatic mtu adjustment17:17
russellbif you need more for your todo list, figuring out a multi-node env we could run in the gate would be cool too :-D17:17
Sam-I-Ami just had too many dependent patches locally and knew i was reaching git-destruction level17:17
Sam-I-Amrussellb: yeah that and gating with native l317:17
Sam-I-Amperhaps non-voting for a while17:18
russellbi'm all for making it voting ASAP17:18
russellbjust need to have a new devstackgaterc file that disables tests as needed17:18
russellbit'll be a lot at first17:18
russellbbut have to start somewhere17:19
arosenrussellb: did you figure what what was causing the test failure?17:19
russellbarosen: no :(17:19
Sam-I-Amtheres some other potential problems with devstack itself that might need fixin17:19
arosenit seemed like the failures were always in tempest.api.networking :(17:19
russellbarosen: it's still happening occasionally, and my last lead turned out to not work out17:20
aroseni ran it like 100 times locally and reproduced the failure twice but wasn't able to track down what is causing it yet :(17:20
russellbwell, nice that you can reproduce17:20
russellbperhaps realted, we're getting some errors in our log even on successful runs17:20
arosenyea but not very often17:20
russellbso maybe that's a good place to start17:20
arosenhard to find what's giong on in the logs.17:20
arosenI tried putting some sleeps in the code to see if i could make it happen more easily but nothing yet.17:21
Sam-I-Amthe good news, devstack doesn't run the 'bad things' when you service_disable q-l3 ... so if you dont mind fixing that stuff in ovn's plugin, we can do that. ideally devstack needs a lot of help so plugins dont have to bandaid it.17:21
arosenyea i definitely wanna understand whats causing this seems really weird.17:21
russellbthis is from a successful test run: http://logs.openstack.org/05/269705/6/check/gate-tempest-dsvm-networking-ovn/98526a3/logs/screen-q-svc.txt.gz?level=TRACE17:21
russellbsome weird stuff17:22
russellbfirst, lots of 2016-02-01 20:23:29.031 27514 WARNING neutron.notifiers.nova [-] Nova returned NotFound for event: [{'tag': u'17bbce2d-8b02-470f-b7a6-e193c68cdfa0', 'name': 'network-vif-deleted', 'server_uuid': u'e897b7b1-dfef-4f77-ac39-60a8b4d1b9cb'}]17:22
russellbwhich seems quite odd.17:22
russellbthere's a NetworkInUse failure in there, though it didn't cause the test to fail17:22
arosenI can explain that warning :)17:22
arosenWhen you delete a port17:22
russellbooh, please do17:22
arosenwe tell nova that the port has been deleted so it can update it's cache to remove it from the instance.17:23
arosenso in this case the port is usually deleted by nova.17:23
arosennova delete vm117:23
arosenso it's a race that nova doesn't have the instance running anymore.17:23
russellbah, maybe shouldn't be a WARNING then17:23
russellbsounds more like DEBUG17:23
arosenit's kinda silly.17:23
arosenyea i agree.17:24
* russellb patches that17:25
arosencool :)17:25
russellbwe also get a db deadlock traceback on port insert sometimes17:25
*** salv-orlando has joined #openstack-neutron-ovn17:26
arosenYea, the dhcp-agent's rpc interface uses transactions over the plugins :(17:27
aroseni think that's the cause of that.17:27
arosenhas there been anymore progress on the ovn dhcp service?17:28
russellbno, it's a bit stalled17:29
russellbthere was discussion about another way to do it, which depends on a new feature ben was working on, that is under review now17:29
russellbso progress in a sense, but we're not close17:29
*** thumpba has quit IRC17:29
russellbarosen: https://review.openstack.org/27533617:31
*** openstackgerrit has quit IRC17:32
russellbbbl lunch17:32
*** openstackgerrit has joined #openstack-neutron-ovn17:32
*** allan_h has joined #openstack-neutron-ovn17:36
arosenalright, going to dig back into random test failure.17:43
*** salv-orl_ has joined #openstack-neutron-ovn17:50
*** salv-orlando has quit IRC17:54
Sam-I-Amrussellb: pssst - https://review.openstack.org/#/c/275331/18:14
*** roeyc has quit IRC18:17
*** fzdarsky is now known as fzdarsky|afk18:19
*** thumpba has joined #openstack-neutron-ovn18:21
*** thumpba has quit IRC18:27
*** salv-orl_ has quit IRC18:28
*** salv-orlando has joined #openstack-neutron-ovn18:28
*** salv-orlando has quit IRC18:31
*** salv-orlando has joined #openstack-neutron-ovn18:32
*** dlundquist has joined #openstack-neutron-ovn18:33
Sam-I-Amwell this is frustrating... not vagrant isnt picking up on my plugin.sh changes18:37
*** azbiswas has joined #openstack-neutron-ovn18:39
*** thumpba has joined #openstack-neutron-ovn18:42
arosendo you guys all use vagrant because you do development on a mac?18:43
*** salv-orlando has quit IRC18:43
Sam-I-Amarosen: i'm running vagrant on a linux box18:43
*** salv-orlando has joined #openstack-neutron-ovn18:45
*** numan_ has joined #openstack-neutron-ovn18:47
arosenDoes that use virtualbox on linux or i guess you can configure it to use w/e?18:48
mesteryI use Vagrant because I can quickly spin-up repeatable environments for testing arosen18:52
mesterySam-I-Am: I'd really like https://review.openstack.org/#/c/274795/ to land18:52
mesteryarosen: ^^^^18:52
mesterySam-I-Am: Do you plan to respin the above to address russellb comment soon?18:53
Sam-I-Ammestery: yeah, working on the underlying patch now18:56
Sam-I-Amfor some reason its not picking up my changes to plugin.sh18:56
Sam-I-Amwell, vagrant isnt... and i have it looking at my local copy of networking-ovn18:57
Sam-I-Ammestery: its basically one thing that needs to move to plugin.sh - enable_isolated_metadata18:58
Sam-I-Amthe other setting in neutron.conf only applies for deployments with > 1 compute node... so not stock devstack. that will stay in the vagrant patch.18:58
mesterySam-I-Am: Ack, thanks18:59
Sam-I-Amif this doesnt work, there must be something else preventing that iniset from being called18:59
Sam-I-Ami ran across this problem yesterday when i try trying to make mtu-via-dhcp configurable18:59
Sam-I-Ambut i thought that was just because vagrant kept recloning networking-ovn from master19:00
arosenmestery: ^ -> https://review.openstack.org/#/c/274795/ to land19:01
mesterySam-I-Am: Is that ok that arosen just boomed that one in?19:01
arosenI can unboom19:02
Sam-I-Ammestery: yep19:02
mesteryarosen: Boom away my friend19:02
Sam-I-Amthat one makes my other bits easier19:02
mesterySam-I-Am: I like to make things easier19:02
Sam-I-Amall the dependent-but-not-really patches yesterday almost got me into trouble with git19:02
Sam-I-Ami still dont know why my stuff didnt work, but fewer git hacks makes it easier to troubleshoot19:03
*** numan_ has quit IRC19:03
openstackgerritMerged openstack/networking-ovn: Vagrant: Modify OpenStack services  https://review.openstack.org/27479519:07
*** jamespd_ has joined #openstack-neutron-ovn19:07
*** dlundquist has quit IRC19:10
*** jamespd has quit IRC19:12
*** dlundquist has joined #openstack-neutron-ovn19:12
azbiswasrussellb: would love your thoughts on https://review.openstack.org/#/c/274274/ when you get some time.19:14
*** gangil has joined #openstack-neutron-ovn19:16
*** gangil has joined #openstack-neutron-ovn19:16
Sam-I-Amsigh... i dont know why this isnt working19:17
Sam-I-Ammy changes to plugin.sh are being ignored19:18
* mestery hands Sam-I-Am a tissue19:18
Sam-I-Ami even commented out the thing that i thought was working, and its still doing that thing19:18
Sam-I-Amy u no work19:19
arosenazbiswas: I just did a read through of your doc.19:21
arosenin this model there is no ovn gateway?19:21
arosennat is done directly on the host?19:21
azbiswasYes nat is done directly on the host19:21
azbiswasnat can be done either using conntrack NAT support or 1-1 translation19:22
arosenit also assumes that the hosts are all connected to the external network?19:22
arosenin this case how does it work for a 1 to many nat?19:22
azbiswasFloating IP is 1-119:23
azbiswasWe are not solving 1-many (SNAT) yet19:23
arosenhrm, I just wonder if not solving that problem at this stage makes things more awkard.19:24
arosenyou need one - many nat to access the internet.19:24
Sam-I-Amarosen: can you review https://review.openstack.org/#/c/274795/ ?19:25
arosenI wonder if it makes more sense to take that into consideration of the design since this will only be something in the interim  until 1 to many is there right?19:25
Sam-I-Amnot that one19:25
*** salv-orlando has quit IRC19:25
azbiswasTrue, in this case we are assuming that the VM needs a Floating IP to access the internet and vice-versa. We will get to SNAT (1-many) once the OVN community figures out a strategy for that.19:26
arosenSam-I-Am:  +219:26
Sam-I-Amone less thing i need to mangle locally19:27
arosenazbiswas:  yea i understand. Though I wonder if implementing this in the interim makes this more complex long term? Do you think we'll be able to easily make it one to many after we implement this/19:27
arosenazbiswas: are the lrouter flow additions something that is going to be down out side of the current northbound interfaec?19:28
arosenor are you saying in ovn those flows will be added?19:29
azbiswasWe've had some thoughts on how SNAT might affect this design. If you add your SNAT concerns in the comments - I can try and address them in this design as well i.e. make sure we are at least not 2 far off from SNAT changes needed.19:29
azbiswasYes - to flows will be added.19:29
arosenazbiswas: sounds good i'll just follow up on the review it self.19:30
azbiswasarosen: The new flows need to be OKed by the OVN community as well.19:31
russellbarmax: azbiswas based on the last OVN IRC meeting, we're expecting a proposed OVN gateway design that includes NAT in the next few weeks19:33
russellbi imagine we'll need to merge these proposals somehow19:33
armaxrussellb: I supposed you wanted to tag arosen instead, didn’t you?19:34
azbiswasThat's good to know - will the gateway reside on the host?19:35
mesteryarmax: We just want you to know we're always thinking of you over here19:35
russellbi think it's a proposal for a central gateway (or set of gateays) i don't know yet, not my proposal19:35
armaxmestery: oh19:36
russellbwas hoping we could make the 2 work together19:36
mesteryI bet it follows this: https://github.com/openvswitch/ovs/blob/master/ovn/OVN-GW-HA.md19:36
russellbarmax: it's true19:36
armaxrussellb: that you thinking of me constantly?19:36
armaxrussellb: or that you wanted to ping arosen?19:36
russellblet's not get too personal19:36
armaxmestery: that’s eerily scary19:37
* armax goes and touches wood19:37
mesteryarmax: Don't read too much into it19:37
mesteryarmax: That's just wrong19:37
* azbiswas reading https://github.com/openvswitch/ovs/blob/master/ovn/OVN-GW-HA.md19:40
russellbit won't necessraily follow that19:40
russellbthat was kind of a brain dump from someone before they left vmware, heh19:40
mesteryheh :)19:41
azbiswasto where?19:41
arosenwhoops sorry i mean't russellb19:41
armaxI am busy here you know?19:42
russellbgrad school19:42
mesteryarmax: You can't fool us19:42
* armax put OVN in the his blacklist19:42
russellbarmax: we try so hard to play nice though!19:43
armaxI am muting notifications from this channel…just so you know!19:44
armaxrussellb, mestery, arosen ^^^^19:44
russellbalright guys, time to pack up this room and move to #openstack-neutron19:44
armaxoh boy19:44
mesteryrussellb armax: We can just bug him on google hangout chat :)19:45
armaxI long for simpler times19:45
Sam-I-Amthis looks transient? https://review.openstack.org/#/c/274948/19:48
Sam-I-Amrussellb: ^19:50
russellbi'm going to offer a bounty of a large basket of cookies for fixing that soon19:53
russellbi'm trying not to think about it this minute ... i don't want to lose the peace I found during my lunchtime walk on the beach19:56
mesteryNice russellb :)19:57
Sam-I-Amwait, what is this beach thing?19:59
russellbyou heard me!19:59
*** thumpba has quit IRC20:00
Sam-I-Amcrap, well i may just let the gate check this devstack patch20:01
russellbthat's not what everyone does for every patch?20:01
Sam-I-Ami like to test things locally when i can20:02
Sam-I-Ami suspect something is weird with vagrant though20:02
Sam-I-Amjust cant put my finger on it20:02
regXboirussellb: is there anybody working on incremental processing within lflow_run/20:04
regXboier ?20:04
russellbben said he was going to work on it20:04
regXboiok, I may ask if there is anything to test on Thursday20:05
russellbok, i don't think that work has started yet20:05
openstackgerritMatthew Kassawara proposed openstack/networking-ovn: Devstack: Provide metadata route via DHCP  https://review.openstack.org/27539220:06
openstackgerritMerged openstack/networking-ovn: Removed unnecessary code from the plugin  https://review.openstack.org/27512520:07
openstackgerritMerged openstack/networking-ovn: Vagrant: Share local directories  https://review.openstack.org/27533120:07
russellbregXboi: there's some similar talk happening around ovn-northd20:09
russellbbut no code yet20:09
regXboirussellb: for a test of stamping out 800 copies of n1 -- r1 -- x1 (each in separate projects), your patches in the ovn-controller-perf branch result in a 2% improvement in the number of flows processed, a 10% improvement in the number of matches processed, and a 16% improvement in the time spent in lflow_run after the test completes20:09
regXboiI'm going to find your patch in ovs-dev and reply with that information20:09
mesterySam-I-Am: With 275331 merged, I'm seeing htis error on a fresh vagrant: http://paste.openstack.org/show/485769/20:09
regXboiand say this is a good first step, but things are still supra linear and so incremental processing is a must20:10
russellbyes, it's just a short term low hanging fruit optimization20:10
Sam-I-Ammestery: yeah, you gotta make those. i thought i put that in the docs.20:10
russellbincremental processing is absolutely the answer for real performance improvements here20:10
* regXboi goes to look for the patch20:10
mesterySam-I-Am: Who reads docs?20:10
* mestery ducks20:10
Sam-I-Amrather, i think the docs say to clone those repos prior to the vagrants20:10
Sam-I-Amvagrant lacks a way to handle missing shared dirs, so there's a hack i put into setup-base to check if they're at least a git repo20:11
Sam-I-Ambut still requires that the dirs exist in some shape or form20:11
Sam-I-Ammestery: sigh @ docs ... :/20:11
mesterySam-I-Am: How does it work sharing hte same ~/devstack across all the VMs in that setup if htey all need different local.conf files?20:12
mesteryDoes that work?20:12
Sam-I-Amok, i'm going to be reaaaaally happy when my new keyboard and mouse arrive20:12
Sam-I-Ammestery: best i could tell, it builds vms in series, so it overwrites the config file before deploying the next one20:14
Sam-I-Amor it should... i could have missed something20:14
mesterySam-I-Am: I'm slightly nervous about that, I'd almost prefer to *not* share devstack and only share networking-ovn for this reason20:14
Sam-I-Amok. my primary concern was sharing networking-ovn to test local changes. devstack was just another attempt at reducing build time.20:15
mesterySam-I-Am: Patch inbound20:15
Sam-I-Ammestery: maybe just comment it out by default?20:15
mesterySam-I-Am: Exactly20:15
Sam-I-Ammight need some docs around it20:15
Sam-I-Amyou know, those things.20:16
openstackgerritKyle Mestery proposed openstack/networking-ovn: Vagrant: Stop sharing the ~/devstack directory  https://review.openstack.org/27539720:18
mesterySam-I-Am: ^^^20:18
Sam-I-Ammestery: lgtm20:21
* Sam-I-Am finds +2 thing20:21
*** salv-orlando has joined #openstack-neutron-ovn20:28
Sam-I-Amrussellb: instances attached to provider nets shouldnt hit any tunnels, right?20:51
*** thumpba has joined #openstack-neutron-ovn20:51
russellbunless you've set up the network the provider net is mapped to with tunnels20:53
Sam-I-Amso this mtu stuff will be more of a hack20:53
russellbovn doesn't know what's on the other side of the bridge20:53
Sam-I-Amreality is, we dont need to pass option 26 to provider nets, just self-service/private nets20:53
* regXboi ponders how to cheaply implement some sort of incremental processing in lflow_run20:54
Sam-I-Ambut theres no clean way to do that20:54
Sam-I-Amits all or nothing with neutron right now20:54
Sam-I-Amits broken now and i think my patch makes it less broken, but we're not going to solve this now20:54
*** fzdarsky|afk has quit IRC20:55
*** thumpba has quit IRC20:56
*** thumpba has joined #openstack-neutron-ovn20:59
*** s3wong has joined #openstack-neutron-ovn21:02
Sam-I-Amrussellb: should i use the somewhat-broken path_mtu option or the broken-in-different ways dnsmasq.conf option?21:03
Sam-I-Ampath mtu might provide the correct mtu for provider nets21:04
Sam-I-Amthats the only benefit there21:04
*** salv-orl_ has joined #openstack-neutron-ovn21:04
*** salv-orlando has quit IRC21:04
russellbSam-I-Am: you tell me21:05
Sam-I-Amrussellb: sigh, nevermind. path_mtu is ML221:05
Sam-I-Amyet somehow, advertise_mtu is neutron itself :/21:06
* Sam-I-Am facepalm21:06
Sam-I-Ambeing that we're not ml2, we probably need our own bits for handing mtu values to instances after subtracting any overlay overhead21:07
*** salv-orl_ has quit IRC21:08
*** salv-orlando has joined #openstack-neutron-ovn21:08
Sam-I-Amrussellb: if your head hasnt exploded yet, give it time.21:10
russellbi think i have to be able to absorb any of it to get to the exploding part21:10
Sam-I-Ambut long term, just using dnsmasq.conf to set option 26 is not going to work in production.21:10
openstackgerritRichard Theis proposed openstack/networking-ovn: WIP: Add network availability zone support  https://review.openstack.org/27541521:30
openstackgerritMatthew Kassawara proposed openstack/networking-ovn: Devstack: Add native MTU option  https://review.openstack.org/27541821:33
Sam-I-Amrussellb: ^ its hacky, but the best we can do right now... i think.21:35
Sam-I-Amthis is one of those problems where i really dont know where to start21:35
Sam-I-Amthe discussion for fixing mtu is all around the ml2 plug-in21:35
Sam-I-Amso we'll probably need to implement something different for ovn that makes use of the mtu extension (which is semi-useless)21:36
Sam-I-Ameven if ovs doesnt have mtu problems, instances will21:36
*** flaviof has joined #openstack-neutron-ovn21:40
*** azbiswas has quit IRC21:45
*** flaviof has quit IRC21:51
*** flaviof has joined #openstack-neutron-ovn21:51
*** azbiswas has joined #openstack-neutron-ovn21:53
*** thumpba has quit IRC21:54
Sam-I-Ambbiab - venturing away from screen22:00
openstackgerritRichard Theis proposed openstack/networking-ovn: WIP: Add network availability zone support  https://review.openstack.org/27541522:12
*** markmcclain has quit IRC22:25
*** markmcclain has joined #openstack-neutron-ovn22:26
regXboirussellb: two part question for you if you are around: (1) do you know where networking-ovn defines its OVS Transaction Timeout from and (2) can that be changed via a configuration setting?22:31
russellb1) off hand no22:31
russellband so 2) don't know22:32
regXboiok I'll go spend some time tomorrow and look22:32
regXboibecause I think in the short term we are going to want to change (2)22:33
russellbyeah i'd have to dig myself22:33
regXboino worries - I asked, I'll pick up the shovel22:33
russellbheh, ok22:34
arosenregXboi:  it's in the config file for ovn22:34
* arosen *the plugin not ovn*22:34
regXboiarosen: thx - that was simple enough :)22:34
* russellb out for the night22:37
*** mamulsow has joined #openstack-neutron-ovn22:38
*** rtheis has quit IRC22:38
*** armax_ has joined #openstack-neutron-ovn22:44
*** armax has quit IRC22:45
*** armax_ is now known as armax22:45
*** azbiswas has quit IRC23:00
*** azbiswas has joined #openstack-neutron-ovn23:01
*** azbiswas_ has joined #openstack-neutron-ovn23:02
*** gangil has quit IRC23:04
*** azbiswas has quit IRC23:05
*** gangil has joined #openstack-neutron-ovn23:06
*** gangil has joined #openstack-neutron-ovn23:06
*** salv-orlando has quit IRC23:13
*** salv-orlando has joined #openstack-neutron-ovn23:13
openstackgerritAaron Rosen proposed openstack/networking-ovn: test  https://review.openstack.org/27543923:18
*** regXboi has quit IRC23:21
*** thumpba has joined #openstack-neutron-ovn23:34
openstackgerritMatthew Kassawara proposed openstack/networking-ovn: Devstack: Provide metadata route via DHCP  https://review.openstack.org/27539223:40
*** gangil has quit IRC23:48
*** gangil has joined #openstack-neutron-ovn23:54
*** gangil has joined #openstack-neutron-ovn23:54
openstackgerritMatthew Kassawara proposed openstack/networking-ovn: Floating IP support design (ready for review).  https://review.openstack.org/27427423:58
*** gangil has quit IRC23:58

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