13:02:41 #startmeeting tricircle 13:02:42 Meeting started Wed Mar 9 13:02:41 2016 UTC and is due to finish in 60 minutes. The chair is joehuang. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:02:43 Hi joe 13:02:44 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:02:46 The meeting name has been set to 'tricircle' 13:02:51 hi Pengfei 13:02:59 #topic rollcall 13:03:06 #info joehuang 13:03:11 because funtion will automatically return a response incase of invalid input 13:03:11 #info Yipei 13:03:15 #help 13:03:17 #info Bean 13:03:23 #info Pengfei 13:03:50 Hi, joe 13:03:54 Hi, Pengfei 13:03:57 hi Haixie 13:04:03 sorry Haixia 13:04:12 to Khayam, yes 13:05:48 as to exception test, just use a fake function to raise exception, and check to see if the caller can catch and print correct logging including stack trace 13:07:08 ok 13:07:19 I have a research on VTN, it can bridge multiple different vlans to a vBridge. Thus can realize vlan l2 networking accross openstacks. 13:07:33 great news, haixia 13:07:41 But it seems to require hardware switch to support. 13:07:53 :( 13:08:05 no software solution? 13:08:58 software solution can be supported, but I really don't kown if it can support vxlan. 13:09:03 to Khayam, what I am afraid is that some parameter in Log.exception has not been initialized when using in the log.Exception 13:09:43 I asked somebody, still cann't get futher information. 13:10:32 to Khayam, if some variable is used without initilization, a new excetion will be throw and exit 13:11:01 You mean VTN can connect two VLAN to one L2 network 13:11:08 I have another concern, even vtn support L2 bridging with software. We need to replace bottom agent with ODL. May some features will be lost. 13:11:14 yes, 13:11:22 but can't make sure these two VLAN will be bridged with VxLAN 13:11:40 yes, joe. it's my concern. 13:12:00 this is a little bit too heavy 13:12:41 Pengfei and Bean, do you have some new ideas about L2 networking across pods 13:13:35 thanks. let me know if anyone is interesterd in linuxcon tokyo from july 13 to 15 13:14:10 I have no idea for the moment 13:14:32 ok. we may discuss it tomorrow in #openstack-tricircle 13:14:58 L2GW has uploaded code for review 13:15:10 we may need to familiar with neutron first. 13:15:22 but not sure it support software L2GW or not currently 13:15:25 I have write a blog last week, and add it in the FAQ of wiki. 13:15:31 https://wiki.openstack.org/wiki/Tricircle 13:15:36 Great! 13:15:48 last week, i have installed the devstack and tricircle with virtualbox 13:16:03 and review the code commited by Pengfei 13:16:21 I have encountered a problem when i reboot the host with devstack. I run ./rejoin_stack.sh, but it didn’t work. 13:16:30 Run ./unstack.sh and ./stack.sh, it echos Error: Service g-api is not running, Error: Service g-reg is not running, [ERROR] /home/stack/git/devstack/functions-common:1646 More details about the above errors can be found with screen, with ./rejoin-stack.sh 13:16:38 Maybe it’s the devstack’s issue or tricircle’s issue. Then i run ./unstack.sh, ./clean, rm -rf /opt/stack/*, reboot the host and ./stack.sh again. It works. 13:16:47 I have a idea.Maybe we can bridge vlan and vxlan with some softeware instance, like L2GW. But we need to have a little change to the bottom. 13:16:49 Do you have any good method to deal with the problem when the devstack doesn’t work ? 13:17:01 https://bugs.launchpad.net/magnum/+bug/1479475 13:17:01 Launchpad bug 1479475 in Magnum "Document recovery steps if devstack host restart" [Undecided,New] - Assigned to Vilobh Meshram (vilobhmm) 13:17:16 pengfei: i found you add a link to paste.openstack.org, but that link is not permanent 13:17:58 so what I need to do? 13:17:59 it will expire in 30 days 13:18:58 to Haixia, what's change? it would be better to use un-touched bottom openstack, or some package could be installed from git repository 13:19:04 pengfei: maybe put the content directly in your blog? 13:19:39 ok~ thanks for your suggestion. 13:20:34 hi, joe, I know what you mean. 13:20:49 So we have to keep research on it. 13:20:49 Any one has good suggestions? 13:21:14 Hi, pengfei, do you have any ideas on L2 networking accross openstack? 13:21:41 I think we can start design the framework for different L2 networking cross pods 13:22:28 and first use shared VLAN for demonstration of the framework 13:22:29 no ideas...囧 13:22:30 hi, joe, you mean overlay and vlan, we can design differently? 13:22:44 beanzhang: rejoin-stack.sh works well for me sometimes after host reboots, if problem occurs, you need to check the log to find out why... 13:22:47 I know. 13:22:58 a frame work with different driver for different L2 netwoking solution 13:23:36 yeah, for vlan, we can use this first. 13:24:09 ok 13:24:31 in https://github.com/openstack/tricircle/blob/master/devstack/local.conf.node_2.sample OVS_BRIDGE_MAPPINGS=bridge:br-bridge,externn:br-ext 13:24:35 but the first driver is shared VLAN, than means, to create a network with same VLAN segment id in bottom pods 13:24:36 The externn is misspelled? 13:25:20 hi Bean, please file a bug and fix it 13:25:30 OK 13:27:36 joe: so the hosts in the bottom pods need to be in the same network? I mean, connected to the same physical switch? 13:27:51 for shared VLAN mode 13:27:59 for the second driver is different vlan. We can keep searching for existing solutions for bottom. But if there's doesn't yet, what should we do? 13:28:14 yes if the switch doen't support VxLAN gateway function 13:28:48 How can we begin formally writing codes? Are there some tutorials about such as how to name the function and what python library to import for using? 13:28:56 if hardware switches support VxLAN gateway function, like TOR as VTEP 13:29:25 then we have to configure the TOR with 1:1 VLAN to VxLAN mapping 13:29:48 then no need for different pods must connect to same switch 13:30:38 one pod one switch, but switches are interconnected with VxLAN for VLAN segment 13:31:08 If not ,then pods must connect to same switch for the shared VLAN 13:31:20 is it clear? 13:31:28 pengfei: you mean coding style? 13:31:35 yes 13:32:18 pengfei: http://docs.openstack.org/developer/hacking/ 13:32:45 the VLAN to VxLAN configuration in TOR or core switch should be configured mannually if we want deploy different pod connecting to different switch 13:33:04 I know what you mean. But I have some other concerns. It's usually used with port hierachy binding. Vlan is 'local vlan', and maybe have some conflict to existing vxlan networks. 13:33:25 thanks a lot! 13:33:25 as what I did in agile controller. 13:33:50 is there any open source agile controller version 13:33:52 But maybe there's other methods can implement what you said. 13:34:05 It's not open. 13:34:13 Enterprise line's product. 13:34:16 yes, shared VLAN can help us to do this 13:34:27 no controller needed 13:34:58 joe, I will have a deep research on it. 13:35:00 it should be shared VLAN segment id exactly 13:35:19 ok, please 13:35:50 if anybody interest on it , please contact me. 13:36:14 to pengfei, for coding style, you can use pep8 or flake8 to check your style 13:36:36 simply run "tox -epep8" 13:36:38 joe, xiaojie have some researches on L3 netwoking. shall we have a talk on it today? 13:37:23 sudo tox -epep8 13:37:57 to Haixia, please 13:38:32 to Yipei, how about you 13:38:36 thanks to joe and zhiyuan, I will read about it 13:39:18 if you use pycharm, wrong coding style could be promopted instantly 13:39:21 I have some finding when fixing the bug of devstack of installation on two nodes. 13:39:36 Simply setting OS_REGION_NAME as RegionOne on node2 does not work on my vm 13:39:41 ok, can it work now 13:39:46 yes 13:39:49 joe, xiaojie have some problems with wireless. I will doi. 13:40:04 to Haixia, ok 13:40:32 Thirst one, there's a bug fix for router, to suport vxlan to other routers. 13:40:35 the two nodes share one keystone 13:40:41 #link http://paste.openstack.org/show/489778/ 13:41:07 by add flooding vtep list or by multicast to suport this. 13:41:09 to Haixia, I noticed it 13:41:35 yeah, but i think some limits. 13:41:47 When create_XXXX_accounts, it is required to go for keystone in RegionOne to verify whether the orperation is legal. Because node1 and node2 share the same keystone service. 13:41:51 is it correct? 13:41:52 but it's freezed in Mitaka, and can't support DVR currently 13:42:09 if pods num is large, flooding list will be increaing. 13:42:16 yeah, I konw. 13:42:42 even it supports DVR, i really don't think flooding is a good method. 13:42:49 They assume only one 13:43:12 yeah~ 13:43:37 to Yipei, I am afraid you installed two independent regions with shared KeyStone, but no Tricircle insatlled 13:43:45 yes 13:44:12 i install devstack on two nodes with tricircle 13:44:26 but when booting vm, i encounter some problems 13:44:32 to Yipei, could follow Pengfei's blog? He has installed two nodes and L3 networking successfully 13:44:59 yes, i read it 13:46:04 Hi, all I think we have a lot to discuss, as more and more items are doing 13:46:05 but simply setting OS_REGION_NAME as RegionOne on node2 does not work on my vm 13:46:06 Second one, with port hierachy binding, vlan and vxlan mapping is 1:1, if router use shared vlan and it's mapped to vxlan. So tenants cann't be isolated for the same vxlan. 13:46:24 but if router or tenants used diffent vlan, 4096 is the limit. 13:46:41 I need to switch the OS_REGION_NAME to RegionTwo when necessary, during installing devstack on node2. 13:46:51 when uploading image to glance, we need to set the OS_REGION_NAME as RegionTwo, since node2 has its own glance service. 13:47:10 to Yipei, use shared Glance now 13:48:01 no need 1:1 for hierarchy port binding 13:48:01 but i need to it will have some problems when executing upload_image 13:48:18 i think glance services are disabled in node2 from the config sample 13:48:19 ok, let's move the talk to #openstack-tricircle tomorrow 13:48:26 ok 13:48:33 the time is limited 13:49:04 to Haixia, we can discuss your concerns tomorrow in #openstack-tricircle 13:49:21 ok. 13:49:45 it also on 1:00 pm UTC? 13:49:50 We need to keep online in #openstack-tricircle, so that more frequently communication 13:50:03 no time limitation if we are online 13:50:06 Yipei, github tricircle Cross-pod L3 networking with DevStack, from my perspective, there is no RegionTwo 13:50:08 oh, I know. 13:50:21 Just RegionOne, Pod1, Pod2 13:50:26 i know, pod2 i mean 13:50:27 to Bean, agree +1 13:51:03 RegionTwo->pod2 13:51:06 to Bean, agree +1 13:51:18 ok, let's conclude the meeting and continue discssion in #openstack-tricircle 13:51:25 ok 13:51:38 see you 13:51:46 #endmeeting