13:01:29 #startmeeting tricircle 13:01:30 Meeting started Wed Feb 3 13:01:29 2016 UTC and is due to finish in 60 minutes. The chair is joehuang. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:01:31 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:01:33 The meeting name has been set to 'tricircle' 13:01:53 hi, Bean, could you introduce yourself> 13:02:18 ok 13:02:36 hi Pengfei 13:02:49 hi, Joe 13:02:54 i'm an undergraduate student from Shanghai Jiao Tong University. 13:03:11 Could also introduce youself, Pengfei? 13:03:51 Yep! 13:04:11 Zhiyuan was graduated from SJTU? 13:04:35 I'm also from SJTU, and also be the student of Pro. Jin 13:04:58 i'm new developer of openstack, deployed my first openstack in june, 2015. 13:04:58 great, like a family of SJTU 13:05:13 Yes, graduated in 2014 13:05:29 Yipei from HUST 13:05:45 #rollcall 13:05:53 #topic rollcall 13:05:58 #info joehuang 13:06:02 #info zhiyuan 13:06:07 #info Yipei 13:06:14 #info Pengfei 13:06:22 #info BeanZhang 13:06:40 #topic progress review 13:06:56 hi, Yipei, how about you 13:07:06 i have already installed devstack in two nodes. i think find a solution to fix the bug, but i am not sure whether it is correct 13:07:19 interesting 13:07:32 i set the OS_REGION_NAME statically as "RegionOne" 13:07:47 in Node2? 13:07:56 yes 13:08:17 when starting cinder, i set it as Pod2 13:08:29 it returns no error 13:09:06 it can run, but will it use the keystone in the node1? 13:09:31 yes 13:09:55 Now Bean and I want to deploy the tricircle by some Devstack nodes in SJTU's OpenStack, and try to learn more practicely about the project by this way:-) 13:09:55 it shares the keystone with node1 13:10:00 and tricircle may not see the node2 services, you have to manully configure the services in the node1 keystone 13:10:02 so the region of services in node2 are registered as RegionOne or Pod2? 13:10:22 yes, make your hand dirty is the fast way 13:11:04 other services are registered as Pod2 13:11:08 bingo! 13:11:11 except keystone 13:11:26 to Pengfei and Bean, if any help need, just send mail in the openstack dev maillist 13:11:33 yeah 13:11:44 thanks a lot! 13:11:47 i receive the maillist all the way 13:12:20 not sure if this is a good fix, since OS_REGION_NAME is a configurable option in DevStack 13:12:24 to Zhiyuan, I am afraid it will overlap one Region configuration 13:12:36 if we occur some problem or any questions, we will send mail to [openstack-dev][tricircle] maillist 13:12:53 ok, got it 13:13:03 BeanZhang that is right 13:13:06 that's the correct way to get help 13:13:21 to BeanZhang, are you going to install OpenStack on bare metal machine, or virtual machine? 13:13:23 we alwasys watch the mail-list 13:13:45 virtual machine 13:14:13 any suggestions? 13:14:24 but Chinese new year is approaching, maybe sometime a later response 13:14:36 i can try to verify whether the solution is correct, i just solve it today 13:15:05 the l3 networking functionality requires vlan network support 13:15:06 virtual machine ok, but need to be aware of the VLAN configuration, some guide in the github 13:15:26 Go on please, Yipei 13:15:35 ok, got it 13:15:59 What I use is vmware vcenter 13:16:12 espeically make sure the service catalog and endpoint in node1 keystone are correct 13:16:36 to Zhiyuan, your progress briefly 13:16:39 ok, got it 13:16:40 pls 13:17:26 yep! we made devstack with neutron before, and vlan by vlan may cause some problems 13:17:34 ok, l3 north-south networking functionality is done, but as Joe commented in the patch, I need to improve the patch 13:18:01 this is a great step 13:18:18 If user associate an unbound port to a floating ip, I need to create the port on bottom pod 13:18:24 basic cross pod networking will be ready after the patch merged 13:18:34 zhiyuan has this patch be merged? 13:19:01 i may have someone want to review the networking patch 13:19:05 not yet for the N-S part, E-W part is merged 13:19:30 helpful more people to do the review 13:19:41 yeah! 13:19:53 okay please add me for the N-S part so that I could add more people to review it 13:20:18 and also zhiyuan can share the test for the distributed query 13:20:34 send it to the maillist? 13:20:57 in the meeting now, a short summary 13:21:07 and also good in the ml 13:21:25 yes that one is good for ml 13:22:52 hi, zhipeng added 13:23:16 Bean and Pengfei, your account to the reviewer list? 13:23:42 In my test, I starts two bottom pods. In each pod, N ports are created, so when we query ports from top pod, we get 2N ports. I increase N and count the response time of query operation 13:23:45 or you can add comment in the patch, and then will be listed in the ptach 13:24:11 #link https://review.openstack.org/#/c/274521/ 13:24:11 three observations 13:24:24 (1) Neutron client costs much more time than curl, the reason may be neutron client needs to apply for a new token in each run. 13:24:34 (2) Eventlet doesn't bring much improvement, the reason may be we only have two bottom pods. I will add some logs to do a further investigation. 13:24:46 (3) Query 1000 ports in top pod costs about 1.5s when using curl, which is acceptable. 13:25:08 I will send the detailed data to the mail list later 13:25:18 good, please 13:25:22 great findings 13:25:34 #action share the distributed query findings 13:26:26 for quota management, the algorithm is updated in the document, please review in the document to see if it applicable 13:26:53 and the patch for quota engin and quota set operation is submitted for review 13:27:23 this is a quite big patch https://review.openstack.org/#/c/274986/ 13:28:03 for almost 2/3 of the code ported from the Cinder 13:28:07 thank joe for the notification! now I add myself in the reviewers ,name: Pengfei.Shi 13:28:30 Great 13:28:59 and for the qutoa management, please also listed yourself to the reviewer, please 13:29:57 The spring festival is coming soon, I hope all of us have a pleasant new year 13:30:31 do you have started your vacation? 13:30:43 it's already getting warmer at the chat room :) 13:30:45 to Pengfei, Bin, Yipei? 13:30:50 yeah 13:30:58 we have gone home-:) 13:30:59 me too 13:31:10 i arrived home yesterday 13:31:10 yeah 13:31:15 I have asked for one day leave and will go home on Feb 6th 13:31:19 so you are working at home , SOHO 13:31:49 now I lying in the bed,hia~hia~ 13:31:50 I'll start the vacation tomorrow, forgive me the review will be slower 13:31:57 haha 13:32:04 Pengfei and me have a question to ask 13:32:10 pls 13:32:19 Pengfei and me deployed nested openstack before. The bare metal network and nested network are VLAN, but the nested network does not work, maybe we will meet the same problem. Do you ever occur similar problem? 13:33:01 you may need to configure the Switch hardware 13:33:13 to let all VLAN network bypass 13:33:34 Zhiyuan, do you think so? 13:34:09 would Pengfei add some? 13:34:57 i remember the hardware switches have be configured as bypass 13:35:18 trunk mode 13:35:20 our enviroment has some other services so it cannot be set of bypass ....TT 13:35:22 If the two OpenStacks are in different bare metal machines, yes, hardware configuration is needed. But I start the two virtual machines running OpenStack in the same machine, so I only need to configure vcenter swtich 13:35:46 s/swtich/switch 13:36:20 the two virtual machines are connected by a software switch 13:36:39 I think we can discuss this in M-L or tricircle channel, it's a little bit late now 13:37:18 deploy all the virtual machines of nested openstack in one compute node is a smart method 13:37:30 so Pengfei and Bin can try one bare matel to see if it works 13:37:36 ok 13:37:44 nice! 13:37:48 good 13:38:06 So happy new year to all of you 13:38:08 just to mention that the test result has been sent, please check :) 13:38:24 Happy New Year!~ 13:38:26 happy new year :) 13:38:31 thanks for the attending, see you next year 13:38:35 Happy new Monkey year! 13:38:36 :) 13:38:52 great work zhiyuan 13:38:58 #endmeeting