| *** yamamoto_ has quit IRC | 01:31 | |
| *** gongysh has joined #openstack-dragonflow | 01:35 | |
| *** yamamoto has joined #openstack-dragonflow | 01:35 | |
| *** gongysh has quit IRC | 01:47 | |
| *** gongysh has joined #openstack-dragonflow | 04:31 | |
| *** gongysh has quit IRC | 05:05 | |
| *** gongysh has joined #openstack-dragonflow | 05:10 | |
| *** gongysh has quit IRC | 05:13 | |
| *** NatanBro has joined #openstack-dragonflow | 05:15 | |
| openstackgerrit | Omer Anson proposed openstack/dragonflow master: Trunk: Update subport status https://review.openstack.org/487305 | 05:26 |
|---|---|---|
| oanson | Morning | 05:34 |
| openstackgerrit | Omer Anson proposed openstack/dragonflow master: Add fullstack test for snat https://review.openstack.org/465524 | 05:34 |
| *** kkxue has joined #openstack-dragonflow | 05:44 | |
| dimak | Good morning | 05:45 |
| dimak | oanson, saw your trunk port patch, +1 | 05:45 |
| oanson | Thanks. | 05:45 |
| dimak | Maybe we should do it for fips as well | 05:45 |
| oanson | Why only +1? :) | 05:45 |
| oanson | Probably | 05:46 |
| oanson | dimak, I think the snat patch should be good to go (finally) | 05:46 |
| dimak | It gates ok? | 05:46 |
| oanson | It should | 05:48 |
| oanson | It had some unit tests issues, but the fullstack finally worked. | 05:49 |
| *** kkxue has quit IRC | 05:54 | |
| *** kkxue has joined #openstack-dragonflow | 05:55 | |
| dimak | oanson, kudos | 05:57 |
| oanson | Yeah. It only took 20 tries :) | 05:57 |
| dimak | Gate debugging is the best debugging ;) | 05:57 |
| dimak | Maybe except for debugging with prod | 05:57 |
| oanson | Embedded gate production environment debugging | 05:58 |
| oanson | dimak, dnat should work out-of-the-box, now? Or are there still issues? | 06:59 |
| oanson | dimak, lihi, https://review.openstack.org/#/c/487305/ and https://review.openstack.org/#/c/465524/ finally pass gate! :) | 07:47 |
| openstackgerrit | Merged openstack/dragonflow master: Answer to "Parameter Request List" option in DHCP app https://review.openstack.org/475718 | 08:03 |
| dimak | oanson, afaik dnat should work | 08:06 |
| oanson | I think I ran into a race-condition: FIP is create the same time as the floating ip port. FIP reaches controller first, so the code that goes through _associate_fip crashes | 08:06 |
| openstackgerrit | Eyal Leshem proposed openstack/dragonflow master: Add support for dnsmsq and dhcpd dhcp-option-names https://review.openstack.org/480231 | 08:07 |
| dimak | oanson, hmm, need to look into, is there some logs in gate? | 08:08 |
| oanson | Local environment | 08:09 |
| irenab | oanson: fip race sounds like something I had when bringing up kuryr overcloud heat stack | 08:17 |
| oanson | I guess that's something https://review.openstack.org/#/c/480196/ will solve, but I doubt I'll be able to solve it today | 08:21 |
| *** kkxue has quit IRC | 08:21 | |
| *** kkxue has joined #openstack-dragonflow | 08:22 | |
| dimak | oanson, I had some ideas in this direction | 08:23 |
| oanson | Shoot | 08:23 |
| dimak | That might let us remove topology | 08:23 |
| dimak | each object will have a set of 'dependency' objects | 08:24 |
| dimak | e.g. a fip depends on both ports | 08:24 |
| dimak | root all dependencies from the OvsPorts | 08:24 |
| dimak | and you have a tree of all objects needed | 08:25 |
| oanson | It won't be a tree, per se. FloatingIP and OVSPort are not reachable from each other in this dependency graph. | 08:25 |
| dimak | Well, at least all topics needed | 08:25 |
| *** kkxue has quit IRC | 08:25 | |
| dimak | we get fips topic from ovsport | 08:26 |
| dimak | we get external network topic from fip | 08:26 |
| *** kkxue has joined #openstack-dragonflow | 08:26 | |
| dimak | once you depend on a topic, you'll have to sync all of it anyway | 08:26 |
| dimak | to stay consistent | 08:26 |
| oanson | Not sure how this would work. The dependency graph is: OvsPort -> lport. FIP -> lport, FIP-> floating lport. lport -> network | 08:27 |
| dimak | ovsport -> lport (add topic) | 08:27 |
| dimak | sync topic | 08:27 |
| dimak | get to a fip, depends on both lports, add relevant topics and sync | 08:28 |
| dimak | although there's plenty room for circular refs | 08:28 |
| oanson | Actually, circular refs should not happen. | 08:28 |
| dimak | among topics, not objects | 08:29 |
| openstackgerrit | Rajat Sharma proposed openstack/dragonflow master: Replace uuid.uuid4() with uuidutils.generate_uuid() https://review.openstack.org/487750 | 08:29 |
| oanson | ovsport -> lport->lswitch => get external network only if lport is on extenal network. But the FIP will belong to the tenant, not the external network, so we should be fine | 08:29 |
| oanson | It should work | 08:29 |
| oanson | And we have a set of topics we listen to. And we can skip objects and subobjects if we already cached/handled them | 08:29 |
| oanson | dimak, I'm guessing the DF controller will have to verify that all dependents are already cached, and pull new topics if needed? | 08:34 |
| dimak | yes | 08:36 |
| *** gongysh has joined #openstack-dragonflow | 08:36 | |
| dimak | oanson, on a totally unrelated note, I though of moving the chassis registration to a chassis health app | 08:37 |
| dimak | proper name pending | 08:37 |
| dimak | and make it non-optional | 08:38 |
| dimak | (for local controller) | 08:38 |
| oanson | We could. Although chassis registration is a core functionality. | 08:38 |
| oanson | If it's not optional, why put it in an app? | 08:38 |
| oanson | You also mentioned moving the event handling out of the controller, so it can be reused (e.g. bgp service) | 08:39 |
| oanson | I think that's a better direction | 08:39 |
| dimak | So we can eventually re-use controller code | 08:39 |
| dimak | yes | 08:39 |
| oanson | I think we just disagree on which part should be moved out :) | 08:39 |
| dimak | Yeah | 08:40 |
| dimak | Some partitioning should happen | 08:40 |
| oanson | Not so unrelated then - how will this play with the conversation above? That will be in the local controller, or general event handling? | 08:40 |
| dimak | General event handling | 08:40 |
| dimak | BGP can take advantage of it as well | 08:41 |
| dimak | It helps dealing with out-of-order updates and notifications | 08:41 |
| dimak | missing* | 08:41 |
| *** kkxue has quit IRC | 09:49 | |
| *** kkxue has joined #openstack-dragonflow | 09:50 | |
| *** yamamoto has quit IRC | 10:00 | |
| openstackgerrit | Omer Anson proposed openstack/dragonflow master: Trunk: Update subport status https://review.openstack.org/487305 | 10:15 |
| *** yamamoto has joined #openstack-dragonflow | 10:31 | |
| dimak | oanson, just a small question on snat patch | 10:45 |
| oanson | Shoot | 10:45 |
| dimak | It's there | 10:45 |
| dimak | And I took the trunk status one | 10:46 |
| oanson | dimak, replied. | 10:48 |
| oanson | And I saw. Thanks! | 10:48 |
| dimak | oanson, voted | 10:49 |
| dimak | I'll split test_apps on top of snat patch | 10:50 |
| oanson | Great. Thanks! | 10:50 |
| oanson | Cool. | 10:50 |
| dimak | I think it's time we did that | 10:50 |
| oanson | Definitely. | 10:50 |
| oanson | I think you'll find there's a lot of repeated code in there that can be moved out | 10:50 |
| oanson | You might want to do that first | 10:50 |
| dimak | Any specifics? | 10:51 |
| dimak | (I haven't looked yet) | 10:51 |
| oanson | Too many to mention :) But nothing of the top of my head | 10:52 |
| openstackgerrit | Rajat Sharma proposed openstack/dragonflow master: Switch from oslosphinx to openstackdocstheme https://review.openstack.org/487810 | 11:14 |
| openstackgerrit | Merged openstack/dragonflow master: Add support for dnsmsq and dhcpd dhcp-option-names https://review.openstack.org/480231 | 11:47 |
| *** yamamoto has quit IRC | 12:14 | |
| *** gongysh has quit IRC | 12:24 | |
| openstackgerrit | Merged openstack/dragonflow master: Move tunnel creation code to tunneling app https://review.openstack.org/487576 | 12:27 |
| oanson | dimak is there a quick tutorial on how to test SFC on Dragonflow? | 12:38 |
| kkxue | +1 | 12:39 |
| *** lewo has quit IRC | 12:39 | |
| dimak | oanson, working on it :) | 12:39 |
| oanson | Great. Thanks. Please cc kkxue when you are done :) | 12:39 |
| dimak | I will :) | 12:40 |
| dimak | kkxue, you're on gerrit? | 12:40 |
| dimak | I'll upload a heat template to our docs, i'll put to on it | 12:40 |
| kkxue | no, i'm not | 12:41 |
| dimak | I'll ping you on irc then :) | 12:41 |
| kkxue | ok，thx! | 12:42 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Support qos in network https://review.openstack.org/403577 | 12:44 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Remove OvsdbQos https://review.openstack.org/487830 | 12:44 |
| dimak | oanson, I resurrected Xiao's network QoS patch | 12:45 |
| oanson | Cool. Thanks. | 12:45 |
| dimak | It's not a big change, I'll see if tests need to be revised but otherwise it was pretty good to go then, moreso now :) | 12:47 |
| oanson | We did a lot of renaming and moving things about. I guess that was most of the work | 12:47 |
| dimak | yes | 12:47 |
| dimak | btw | 12:47 |
| dimak | qos won't work with trunk-ports now :( | 12:47 |
| dimak | and they return true from lport.is_vm_port() | 12:48 |
| oanson | I think they are a is_vm_port | 12:49 |
| oanson | And qos relies on tc, so it would only work with 'real' ports | 12:49 |
| oanson | i.e. those that the linux kernel sees too | 12:50 |
| openstackgerrit | Omer Anson proposed openstack/dragonflow master: Implement utc_timeout for PGSQL and MSSQL https://review.openstack.org/487833 | 12:50 |
| dimak | oanson, yes | 13:00 |
| dimak | We can implement it for non real ports by funneling their traffic through real ports :P | 13:01 |
| oanson | Yes. But I don't want to do that. | 13:01 |
| oanson | The *best* solution is openflow metering. But OVS doesn't support that | 13:01 |
| oanson | It's somewhere on my todo list | 13:01 |
| dimak | it just adds implicit max-kbps ;) | 13:02 |
| oanson | I think metering supports burst as well | 13:02 |
| irenab | oanson: dimak : need to check this with DF: https://bugs.launchpad.net/neutron/+bug/1627987 | 13:10 |
| openstack | Launchpad bug 1627987 in neutron "[RFE] SR-IOV accelerated OVS integration" [Wishlist,In progress] - Assigned to Moshe Levi (moshele) | 13:10 |
| *** yamamoto has joined #openstack-dragonflow | 13:14 | |
| *** yamamoto has quit IRC | 13:21 | |
| openstackgerrit | Merged openstack/dragonflow master: Trunk: Update subport status https://review.openstack.org/487305 | 13:31 |
| *** mlavalle has joined #openstack-dragonflow | 13:59 | |
| *** NatanBro has quit IRC | 14:15 | |
| openstackgerrit | Merged openstack/dragonflow master: Add fullstack test for snat https://review.openstack.org/465524 | 14:49 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Support qos in network https://review.openstack.org/403577 | 15:28 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Remove OvsdbQos https://review.openstack.org/487830 | 15:28 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Remove ovsdb.objects https://review.openstack.org/487908 | 15:28 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Remove ovsdb.objects https://review.openstack.org/487908 | 15:36 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Support qos in network https://review.openstack.org/403577 | 15:36 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Remove OvsdbQos https://review.openstack.org/487830 | 15:36 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Remove ovsdb.objects https://review.openstack.org/487908 | 18:27 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Support qos in network https://review.openstack.org/403577 | 18:27 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Remove ovsdb.objects https://review.openstack.org/487908 | 18:29 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow master: Support qos in network https://review.openstack.org/403577 | 18:29 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/dragonflow master: Updated from global requirements https://review.openstack.org/487991 | 19:44 |
| *** yamamoto_ has joined #openstack-dragonflow | 21:13 | |
| *** yamamoto_ has quit IRC | 21:20 | |
| *** gongysh has joined #openstack-dragonflow | 21:36 | |
| *** gongysh has quit IRC | 21:36 | |
| *** yamamoto has joined #openstack-dragonflow | 22:04 | |
| *** openstack has joined #openstack-dragonflow | 22:46 | |
| *** leyal has joined #openstack-dragonflow | 22:47 | |
| *** mlavalle has quit IRC | 23:40 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!