| *** joehuang has joined #openstack-tricircle | 00:47 | |
| *** dongfeng has joined #openstack-tricircle | 00:59 | |
| joehuang | we start the meeting at #openstack-meeting now | 01:01 |
|---|---|---|
| *** zhiyuan has joined #openstack-tricircle | 02:13 | |
| joehuang | hello, dongfeng | 02:13 |
| dongfeng | hello | 02:17 |
| dongfeng | @joehuang | 02:17 |
| joehuang | currently we have to cache the neutron url from keyston | 02:19 |
| joehuang | but in keystone federation scenario, through pod's region name information, we may not able to get the neutron endpoint | 02:20 |
| joehuang | so how to deal with the neutron endpoint url in federation scenario | 02:20 |
| dongfeng | you mean we need access distributed keystone and get info from them? | 02:24 |
| zhiyuan | joe's idea is to provide an API for admin to configure the endpoint | 02:28 |
| dongfeng | change the path "127.0.0.1/tricircle" to another value? | 02:32 |
| zhiyuan | no | 02:33 |
| zhiyuan | as you know, endpoints for services(like neutron) are retrieved from keystone and stored in our database | 02:33 |
| zhiyuan | in the keystone federation scenario, we may not directly retrieve endpoints from keystone, so we need to manually configure them | 02:35 |
| dongfeng | in federation scenario, how do we know the endpoint? | 02:37 |
| dongfeng | get it from each keystone? | 02:37 |
| zhiyuan | we expose an API and admin manually add it | 02:42 |
| dongfeng | read the endpoint from db and show it? | 02:43 |
| zhiyuan | no, admin knows the endpoint, so add it to Tricircle | 02:44 |
| dongfeng | CachedEndpoint saves all the values? | 02:45 |
| dongfeng | or add it to cached_endpoints? | 02:46 |
| dongfeng | another round of crud operations for endpoints? | 02:50 |
| zhiyuan | hmm, this needs discussion. whether add the endpoint to cached_endpoints table, or pods table | 02:50 |
| zhiyuan | no need for a complete CRUD support i think | 02:51 |
| dongfeng | you mean 1) we can config directly 2) add it to db | 02:51 |
| zhiyuan | 1) we get it from keystone 2) admin set it via API, both needs to write to the db | 02:52 |
| dongfeng | we get endpoints from different keystone? | 02:53 |
| dongfeng | the workflow is clear, but I don't know how to implement | 02:55 |
| joehuang | Admin can know the endpoint url through other ways, then fill it into tricircle | 02:55 |
| dongfeng | yes | 02:56 |
| joehuang | how to implement could be discussed | 02:56 |
| dongfeng | ok. how to fill? zhiuan said it may not through write to pods/endpoints table | 02:57 |
| zhiyuan | through write to db, whether write to pods table or endpoins table needs discussion | 02:58 |
| dongfeng | if we got the endpoint info, why not write it to endpoints table? | 02:59 |
| zhiyuan | (1) add a new field to pods table to store the manually configured endpoint (2) add a field to endpoints table to mark it as static | 02:59 |
| dongfeng | an endpoint is an pod info／ | 03:00 |
| dongfeng | ？ | 03:00 |
| zhiyuan | oh, after reading the client.py code, I think no need for a "static" mark, since we only update endpoints that can be retrieved from keystone | 03:01 |
| zhiyuan | if one endpoint is manually configured by admin, meaning that it cannot be retrieved from keystone, it will not be changed | 03:02 |
| dongfeng | manually config won't be troublesome? what if it is changed later／ | 03:03 |
| zhiyuan | yeah, the hard work is left to admin | 03:05 |
| zhiyuan | A better solution is to retrieve endpoints from other keystones. I haven't played with a federated keystone environment, so not quite know the details | 03:07 |
| dongfeng | can you show me what's the endpoints info looks like? I mean the data model | 03:09 |
| dongfeng | it seemed it not looks like item in pod/cached_endpoint table | 03:09 |
| joehuang | the priority for this enhancement is not very high | 03:12 |
| joehuang | we can delay when people ask for it | 03:13 |
| zhiyuan | First we better figure out how keystone federation works with a federation environment. | 03:14 |
| dongfeng | if federated keystone has shared db then this work may become easier? | 03:15 |
| zhiyuan | as I konw, federated keystone doesn't use shared db | 03:16 |
| dongfeng | oh | 03:16 |
| zhiyuan | https://docs.openstack.org/security-guide/identity/federated-keystone.html | 03:16 |
| dongfeng | my initial idea is to create another table in tricircle to collect the endpoint info | 03:17 |
| zhiyuan | the challenge is how to collect the endpoints, I think the current cached_endpoints table is enought to store endpoint info | 03:18 |
| zhiyuan | s/enought/enough | 03:19 |
| dongfeng | if collection process of endpoint can be dynamic, then we won't worry whether the collected item is changed later. just update if needed. | 03:22 |
| zhiyuan | yeah, if we can retrieve endpoints from each keystone under a federated keystone setup | 03:33 |
| *** Kevin_Zheng has quit IRC | 03:59 | |
| *** joehuang has quit IRC | 04:11 | |
| *** dongfeng has quit IRC | 06:36 | |
| *** dongfeng has joined #openstack-tricircle | 06:48 | |
| *** Yipei has joined #openstack-tricircle | 07:15 | |
| *** Yipei has quit IRC | 13:15 | |
| *** Meher has joined #openstack-tricircle | 13:20 | |
| Meher | Hello | 13:22 |
| dongfeng | hi | 13:22 |
| Meher | I introduce myself; Meher Hihi; I am doing my internship at Orange Labs Networks Lannion-France for the diploma of computer network and telecommunications engineer. I am working on innovative distribution solutions for the virtualization infrastructure of the network functions and more specifically on the Openstack Tricircle solution, which is why I join your community to participate in your discussions and learn from your | 13:23 |
| Meher | Indeed, I try to install Tricircle on a single node by following this documentation “https://docs.openstack.org/developer/tricircle/installation-guide.html#single-pod-installation-with-devstack”. I managed to install Devstack without any problems, but when I modify the local.conf file by adding the Tricircle plugin integration and the HOST_IP, the script does not want to work and stops on an error of Start of the Keyston | 13:23 |
| dongfeng | I am a student but also a developer in tricircle. | 13:25 |
| dongfeng | you mean do you installed a single node of tricircle successfully? | 13:27 |
| Meher | Have you tried installing Tricircle on a single node with Devstack? | 13:28 |
| dongfeng | yes. | 13:28 |
| dongfeng | I can install a single node by Devstack. | 13:28 |
| Meher | No I just managed to install Devstack without Tricircle | 13:28 |
| Meher | What is weird, I added only the local.conf file lines HOST_IP and enable_plugin tricircle https://github.com/openstack/tricircle/, and the script no longer works while it was running in the installation Of Devstack without Tricircle, it stops on a startup error of Keystone. I wanted to know if someone has encountered this problem. | 13:31 |
| dongfeng | a node with out tricircle is in multi-node installation scenario. under this circumstance, only HOST_IP is need to changed to the central host ip | 13:32 |
| dongfeng | I install one node recently. | 13:33 |
| dongfeng | please wait a minute I will show you my local.conf file | 13:34 |
| Meher | Okay thank you! | 13:35 |
| dongfeng | hello, this is my local.conf file for one node: http://paste.openstack.org/show/613927/ | 13:36 |
| dongfeng | what's the differences in the files between us? | 13:38 |
| Meher | The address you put in HOST_IP is the address of the machine? | 13:38 |
| dongfeng | yes. | 13:38 |
| Meher | The configuration I have is minimal since I followed https://docs.openstack.org/developer/tricircle/installation-guide.html#single-pod-installation-with-devstack | 13:39 |
| dongfeng | this is my ifconfig info: http://paste.openstack.org/show/613928/ | 13:39 |
| dongfeng | when I install a node, I observe the multi-pods installation guide but at last I only install one node. | 13:40 |
| dongfeng | because I only need one node environment for test use. | 13:41 |
| Meher | I will try to install with the documentation of multi-pods | 13:42 |
| dongfeng | you can run ./unstack.sh and ./clean.sh and then use my local.conf | 13:42 |
| Meher | Thank you very much for the help | 13:42 |
| dongfeng | that's all right. | 13:42 |
| *** Meher has quit IRC | 17:02 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!