| *** zenoway has joined #openstack-dragonflow | 00:08 | |
| *** zenoway has quit IRC | 00:12 | |
| *** vikasc_ has quit IRC | 00:54 | |
| *** vikasc has quit IRC | 00:54 | |
| *** Duan has joined #openstack-dragonflow | 01:32 | |
| *** DuanKebo1 has joined #openstack-dragonflow | 01:36 | |
| *** yamamoto has quit IRC | 02:45 | |
| *** yamamoto has joined #openstack-dragonflow | 02:52 | |
| *** zenoway has joined #openstack-dragonflow | 02:52 | |
| *** zenoway has quit IRC | 02:56 | |
| *** vikasc_ has joined #openstack-dragonflow | 02:57 | |
| openstackgerrit | duankebo proposed openstack/dragonflow: Adjust Default configurations of Dragonflow https://review.openstack.org/308284 | 03:30 |
|---|---|---|
| *** vikasc_ has quit IRC | 03:42 | |
| *** yamamoto has quit IRC | 03:58 | |
| openstackgerrit | hujie proposed openstack/dragonflow: uniform all the interface for all objects https://review.openstack.org/315433 | 04:29 |
| *** yamamoto has joined #openstack-dragonflow | 04:39 | |
| *** vikasc has joined #openstack-dragonflow | 06:00 | |
| vikasc | ping gsagie | 06:03 |
| gsagie | vikasc: hi | 06:03 |
| vikasc | Can i look into this bug https://bugs.launchpad.net/dragonflow/+bug/1557412 ? | 06:03 |
| openstack | Launchpad bug 1557412 in DragonFlow "VM connected to two different networks" [High,New] - Assigned to Gal Sagie (gal-sagie) | 06:03 |
| vikasc | Its assigned to you | 06:03 |
| vikasc | Earlier one was invalid. | 06:04 |
| gsagie | vikasc: sure, feel free | 06:04 |
| gsagie | thanks :) | 06:04 |
| vikasc | gsagie, thanks :) | 06:04 |
| gsagie | DuanKebo: here? | 06:10 |
| gsagie | Duan, DuanKebo1 :) | 06:10 |
| *** zenoway has joined #openstack-dragonflow | 06:24 | |
| DuanKebo1 | Hi gal | 06:25 |
| gsagie | DuanKebo: can you send me an example local.conf for running with Redis, want to add it as default to the gate | 06:33 |
| *** todin has joined #openstack-dragonflow | 06:34 | |
| todin | morning | 06:34 |
| DuanKebo1 | OK, second | 06:35 |
| DuanKebo1 | Q_ENABLE_DRAGONFLOW_LOCAL_CONTROLLER=True | 06:38 |
| DuanKebo1 | #DF_SELECTIVE_TOPO_DIST=True | 06:38 |
| DuanKebo1 | DF_REDIS_PUBSUB=True | 06:38 |
| DuanKebo1 | enable_plugin dragonflow http://git.openstack.org/openstack/dragonflow | 06:38 |
| DuanKebo1 | #enable_service df-etcd | 06:38 |
| DuanKebo1 | #enable_service df-etcd-server | 06:38 |
| DuanKebo1 | enable_service df-redis | 06:38 |
| DuanKebo1 | enable_service df-redis-server | 06:38 |
| DuanKebo1 | #enable_service df-redis-publisher-service | 06:38 |
| DuanKebo1 | #enable_service df-publisher-service | 06:38 |
| DuanKebo1 | enable_service df-controller | 06:38 |
| DuanKebo1 | enable_service df-ext-services | 06:38 |
| gsagie | morning todin | 07:08 |
| gsagie | thanks DuanKebo1 | 07:09 |
| hujie | gampel: ping | 07:17 |
| openstackgerrit | Li Ma proposed openstack/dragonflow: Fix ubuntu 16.04 support for zookeeper installation https://review.openstack.org/315336 | 07:19 |
| gampel | hi | 07:21 |
| gampel | hujie: hi | 07:21 |
| hujie | Hi :) | 07:21 |
| hujie | I want to discuss the review you leave in my db consistency patch | 07:22 |
| gampel | hujie: sure when we we have the version in the object | 07:23 |
| gampel | we should update only newer versions of the cache | 07:23 |
| hujie | you mean the new messages maybe arrived earlier than the older ones? | 07:24 |
| gampel | yes | 07:24 |
| hujie | for a specific object, if it is processed by the same Neutron server, I think it's processed one by one | 07:25 |
| hujie | and if it is processed by multi-server, I think we have distributed lock to make sure the sequence | 07:25 |
| hujie | what's the senario you mean by arrived earlier? | 07:26 |
| gampel | let say you send multiple update on the same object from the API the load balance will not send it always to the same neutron server | 07:26 |
| hujie | yes | 07:26 |
| gampel | and in the sql DB they got a lock and update x on Neutron server 1is before update y on neutron server 2 | 07:27 |
| gampel | then both send the update via the pub sub and y arrive to compute node C1 before update x | 07:28 |
| gampel | Know the SQL db the value of the object is y and in the local cache of C1 it is y | 07:29 |
| gampel | That is way we do not send the value in the pubsub now we send just notification and do get | 07:30 |
| gampel | We want to change that and avoid the extra latency | 07:30 |
| *** zhuweiqiong has quit IRC | 07:30 | |
| gampel | I have to go lets continue later | 07:30 |
| hujie | Neutron server 1 update x and send the update via pub sub , I think it's a atomic operation in the control of distributed lock | 07:31 |
| hujie | so if x is updated into sql first , it will arrive local controller first | 07:32 |
| *** vikasc has quit IRC | 07:32 | |
| hujie | In redis pub/sub implement, we fill the message with the whole object :) | 07:35 |
| *** vikasc has joined #openstack-dragonflow | 07:48 | |
| openstackgerrit | zhuweiqiong proposed openstack/dragonflow: Add redis HA implement https://review.openstack.org/309642 | 08:05 |
| *** saggi has joined #openstack-dragonflow | 08:12 | |
| gampel | hujie: it is not atomic it is sent via pub sub | 08:16 |
| gampel | when you send to events via pub sub from two different servers you can not guarantee order | 08:17 |
| gampel | hujie: ping | 08:18 |
| *** irenab has joined #openstack-dragonflow | 09:22 | |
| openstackgerrit | hujie proposed openstack/dragonflow: add data sync mechanism for keep db consistency https://review.openstack.org/300877 | 09:31 |
| vikasc | hi nick-ma | 10:17 |
| nick-ma | vikasc: hi | 10:18 |
| vikasc | nick-ma, if you have couple of minutes, I want to ask a very basic question | 10:18 |
| nick-ma | ok, sure. | 10:18 |
| vikasc | nick-ma, functionalityt wise OVN seems to be very similar to DF | 10:19 |
| vikasc | nick-ma, I am wondering if ovn is there(which is enabling sgs etc ovs in sdn way), why DF is needed | 10:20 |
| vikasc | nick-ma, which is enabling sgs etc using ovs in sdn way | 10:21 |
| nick-ma | sgs? | 10:21 |
| vikasc | security groups | 10:21 |
| vikasc | nick-ma, I was going through this, http://openvswitch.org/support/slides/OVN-Vancouver.pdf | 10:22 |
| vikasc | nick-ma, It says ovn enables logical router, logical switch, vteps, security groups in ovs using ovn-controller | 10:23 |
| nick-ma | i don't know if ovn and df can be co-existed in the same environment. df is a little bit similar to ovn. the difference is that df is pythonic and governed by the openstack community, but ovn is totally not. | 10:23 |
| vikasc | nick-ma, okay.. might be DF will support a bigger set of functionalities than OVN | 10:24 |
| gsagie | vikasc: there are 3 main difference that led us to start DF as oppose to using OVN (and i am active contributor to OVN) | 10:25 |
| vikasc | gsagie, delighted that you are answering this :) | 10:26 |
| gsagie | 1) the pluggable DB and pub-sub (which i tried to push to OVN) - very important for control plane scaling / HA / Redundancy and more which are none existent in OVSDB currently (only DB supported by OVN) | 10:26 |
| gsagie | 2) We distribute policy level abstraction data to the local controller, this let us implement network services really fast in an easy way and implement very nice features like "selective proactive" - distribute only relevant information to each compute node | 10:27 |
| gsagie | It also allow us to have nice integrations with various different NICs and hardwares (i can elaborate more if you want) | 10:27 |
| gsagie | OVN as you know distribute logical flows already compiled in a centralized DB location | 10:27 |
| vikasc | gsagie, please elaborate on this hardware and NIC part. | 10:28 |
| vikasc | gsagie, or link to any existing doc | 10:28 |
| gsagie | 3) OVN uses 2 DB's in addition to Neutron DB which to us brings many consistency issues in production which we are tring to minimize (Dragonflow has only another DB and nick-ma is working on our consistency model which i suspect is going to be a bit harder for OVN given their DB model) | 10:29 |
| gsagie | and another important part is our plans to support reactive OpenFlow pipelines where OVN (at least from talking with Ben and Justin) tries to keep everything proactive - which means you need to insert all possible features to OVS (for example DHCP if you follow that up) | 10:29 |
| gsagie | thats the main differences that led us continue with the model in DF, of course we borrowed many good ideas from OVN implementation that we already fully use in DF | 10:30 |
| vikasc | many thanks gsagie, It will take sometime for me to digest your points :) | 10:31 |
| gsagie | vikasc: the idea is that its hard to fully offload OVS to smart NICs (we are working with some NIC vendors right now) , there are many new features like connection tracking and ARP responding and so on and this keep progress which makes it hard for HW to follow, so the idea is that Dragonflow locally on the compute node can detect the HW capabilities and re-compile the pipeline to use the HW offload capabilities | 10:31 |
| gsagie | but still use the SW OVS for everything else, but this require a smart controller which gets policy level data in the local compute node and know how to translate it | 10:32 |
| gsagie | vikasc: np :) let me know if you need any more information | 10:32 |
| gsagie | and of course the points that vikasc mentioned about python and the full code being under OpenStack | 10:33 |
| vikasc | gsagie, enough for now. will take hour for each :) . Will get back for sure with doubts. thanks once again | 10:33 |
| gsagie | that nick-ma mentioned, sorry | 10:33 |
| vikasc | got it | 10:34 |
| nick-ma | gsagie: thanks, i will save your conclusion on df to my notes, to make sure i'll 'answer' them quickly enough if some one want me to answer similar questions. :-) | 10:36 |
| *** yamamoto has quit IRC | 10:36 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/dragonflow: Updated from global requirements https://review.openstack.org/317374 | 10:36 |
| vikasc | thanks nick-ma and gsagie . Will read more on the pointers. | 10:36 |
| gsagie | nick-ma: :) | 10:48 |
| *** yuanwei has quit IRC | 11:35 | |
| *** yuanwei has joined #openstack-dragonflow | 11:35 | |
| *** yuanwei has quit IRC | 11:39 | |
| *** yuanwei has joined #openstack-dragonflow | 11:39 | |
| *** yuanwei has quit IRC | 11:41 | |
| *** yamamoto has joined #openstack-dragonflow | 11:51 | |
| *** yamamoto has quit IRC | 11:54 | |
| *** irenab has quit IRC | 12:12 | |
| *** vikasc has quit IRC | 12:15 | |
| *** yamamoto has joined #openstack-dragonflow | 12:19 | |
| openstackgerrit | duankebo proposed openstack/dragonflow: Enable router scheduler for dragonflow plugin Closes-Bug: #1574631 https://review.openstack.org/317430 | 12:24 |
| openstack | bug 1574631 in DragonFlow "after running stack router interface is deleted by df-l3-agent" [High,New] https://launchpad.net/bugs/1574631 - Assigned to Duan Kebo (duankebo) | 12:24 |
| openstackgerrit | duankebo proposed openstack/dragonflow: Enable router scheduler for dragonflow plugin https://review.openstack.org/317430 | 12:25 |
| *** irenab has joined #openstack-dragonflow | 12:42 | |
| *** irenab has quit IRC | 12:43 | |
| *** irenab has joined #openstack-dragonflow | 13:01 | |
| *** openstackgerrit has quit IRC | 13:32 | |
| *** openstackgerrit has joined #openstack-dragonflow | 13:33 | |
| *** yamamoto has quit IRC | 13:36 | |
| *** yamamoto has joined #openstack-dragonflow | 14:50 | |
| *** yamamoto has quit IRC | 15:05 | |
| *** yamamoto has joined #openstack-dragonflow | 15:30 | |
| *** yamamoto has quit IRC | 15:59 | |
| *** yamamoto has joined #openstack-dragonflow | 16:01 | |
| *** saggi has quit IRC | 16:04 | |
| *** yamamoto has quit IRC | 16:23 | |
| *** yamamoto has joined #openstack-dragonflow | 16:24 | |
| *** yamamoto has quit IRC | 16:29 | |
| *** yamamoto has joined #openstack-dragonflow | 16:30 | |
| *** yamamoto has quit IRC | 16:32 | |
| *** zenoway has quit IRC | 16:36 | |
| *** zenoway has joined #openstack-dragonflow | 16:36 | |
| *** zenoway has quit IRC | 16:41 | |
| *** yamamoto has joined #openstack-dragonflow | 16:46 | |
| *** zenoway has joined #openstack-dragonflow | 17:05 | |
| *** zenoway has quit IRC | 17:09 | |
| openstackgerrit | Merged openstack/dragonflow: allow packet data to be passed as bytearray to applicative tests https://review.openstack.org/305131 | 18:00 |
| *** irenab has quit IRC | 18:00 | |
| *** irenab has joined #openstack-dragonflow | 18:02 | |
| *** irenab has quit IRC | 18:14 | |
| *** irenab has joined #openstack-dragonflow | 18:24 | |
| *** zenoway has joined #openstack-dragonflow | 18:48 | |
| *** mastermind has joined #openstack-dragonflow | 18:54 | |
| *** zenoway has quit IRC | 18:57 | |
| *** zenoway has joined #openstack-dragonflow | 18:57 | |
| *** omnipresent has joined #openstack-dragonflow | 19:00 | |
| *** mastermind has quit IRC | 19:02 | |
| *** mastermind has joined #openstack-dragonflow | 19:03 | |
| *** zenoway has quit IRC | 19:04 | |
| *** zenoway has joined #openstack-dragonflow | 19:06 | |
| *** omnipresent has quit IRC | 19:07 | |
| *** mastermind has quit IRC | 19:09 | |
| *** mastermind has joined #openstack-dragonflow | 19:10 | |
| *** zenoway has quit IRC | 19:13 | |
| *** zenoway has joined #openstack-dragonflow | 19:14 | |
| *** omnipresent has joined #openstack-dragonflow | 19:17 | |
| *** mastermind has quit IRC | 19:17 | |
| *** zenoway has quit IRC | 19:20 | |
| *** zenoway has joined #openstack-dragonflow | 19:21 | |
| *** mastermind has joined #openstack-dragonflow | 19:23 | |
| *** omnipresent has quit IRC | 19:24 | |
| *** zenoway has quit IRC | 19:26 | |
| *** yuanwei has joined #openstack-dragonflow | 19:33 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!