| *** wanghao has joined #openstack-mogan | 00:29 | |
| *** little has joined #openstack-mogan | 00:52 | |
| little | Maybe the gerrit is down | 01:13 |
|---|---|---|
| *** kevinz has joined #openstack-mogan | 01:14 | |
| *** 7F1AAKIKM is now known as luyao | 01:33 | |
| openstackgerrit | liusheng proposed openstack/python-moganclient: Add 'Power State' column in server list --long comand https://review.openstack.org/422331 | 01:44 |
| zhenguo | morning mogan! | 01:49 |
| little | morning | 01:51 |
| liusheng | morning | 01:54 |
| *** zhangjl has joined #openstack-mogan | 01:59 | |
| zhenguo | #startmeeting mogan | 02:01 |
| openstack | Meeting started Thu Jan 19 02:01:00 2017 UTC and is due to finish in 60 minutes. The chair is zhenguo. Information about MeetBot at http://wiki.debian.org/MeetBot. | 02:01 |
| openstack | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 02:01 |
| openstack | The meeting name has been set to 'mogan' | 02:01 |
| zhenguo | o/ | 02:01 |
| zhangjl | o/ | 02:01 |
| openstackgerrit | Merged openstack/python-moganclient: Add 'Power State' column in server list --long comand https://review.openstack.org/422331 | 02:01 |
| zhenguo | ok, as always, the agenda: | 02:02 |
| zhenguo | #link https://wiki.openstack.org/wiki/Meetings/Mogan#Agenda_for_next_meeting | 02:02 |
| zhenguo | #topic announcements and reminders | 02:02 |
| zhenguo | Call for presentations for OpenStack Summit Boston | 02:03 |
| liusheng | o/ | 02:03 |
| zhenguo | The deadline for submission is Feb 6 | 02:03 |
| zhenguo | I would suggest you all register one | 02:03 |
| zhenguo | it's a good opportunity to promote Mogan! | 02:04 |
| little | right | 02:04 |
| shaohe_feng | o/ | 02:05 |
| zhenguo | shaohe_feng: o/ | 02:05 |
| zhenguo | shaohe_feng: you are alive, hah | 02:05 |
| shaohe_feng | Good. | 02:05 |
| shaohe_feng | we can submit a topic for mogan | 02:05 |
| zhenguo | yes, | 02:05 |
| little | great | 02:05 |
| zhenguo | but not one, we can t ry to submit more, lol | 02:06 |
| shaohe_feng | yes. | 02:06 |
| zhenguo | in different track | 02:06 |
| zhenguo | seems every presentation can have 3 speakers, right? | 02:07 |
| liusheng | zhenguo: yes | 02:07 |
| zhenguo | hah, try to make it, then we can fly to Boston | 02:08 |
| little | you guys have this experience? | 02:08 |
| zhenguo | wanghao and shaohe_feng has | 02:08 |
| zhenguo | they both have topic selected in previous summit | 02:09 |
| zhenguo | ok, not much announcement, | 02:10 |
| zhenguo | anything else here? | 02:10 |
| shaohe_feng | zhenguo: I have found the topic should not be very technicalization | 02:10 |
| zhenguo | shaohe_feng: yes, because we have a PTG for techs | 02:10 |
| zhenguo | I will attend the coming PTG if there's no problems with my visa | 02:11 |
| liusheng | zhenguo: good luch :) | 02:11 |
| liusheng | s/luch/luck | 02:11 |
| zhenguo | hah | 02:11 |
| shaohe_feng | zhenguo: It is better a cooperation topic accross company | 02:11 |
| shaohe_feng | zhenguo: Good luck. You are rich. | 02:12 |
| zhenguo | shaohe_feng: yes, so I suggest we can prepare it together | 02:12 |
| zhenguo | lol | 02:12 |
| zhenguo | maybe china mobile should have a presentation | 02:13 |
| liusheng | hope you can attract more contibutors and customers | 02:13 |
| shaohe_feng | zhenguo: I submmit a topic about live migration first time, just technical introduce. seldom select this one. | 02:13 |
| zhenguo | liusheng: I just want them to accept us | 02:13 |
| little | shaohe_feng: live migration in nova? | 02:14 |
| zhenguo | shaohe_feng: hah, seems because the summit has splitted | 02:14 |
| shaohe_feng | zhenguo: And the next one, I submmit topic to introduce how to optimize live migration, it is select as option. | 02:14 |
| shaohe_feng | little: yes. | 02:14 |
| zhenguo | shaohe_feng: hah, that's a experience | 02:15 |
| shaohe_feng | zhenguo: Then the GPU, I just tell what audiences care, and cooperation with Huawei. | 02:15 |
| shaohe_feng | zhenguo: it is selected. | 02:15 |
| zhenguo | we may submit one about the limitation of nova to manage baremetals | 02:16 |
| shaohe_feng | zhenguo: good point. | 02:16 |
| zhenguo | shaohe_feng: yes, but it's a pity that they don't send you there :( | 02:16 |
| wanghao | zhenguo: good idea for summit topic. | 02:17 |
| zhenguo | the other option is that baremetal cloud without nova | 02:17 |
| shaohe_feng | yes. Intel is stingy | 02:17 |
| zhenguo | hah | 02:17 |
| shaohe_feng | most compony want their employee to summit topic in these meeting | 02:18 |
| liusheng | we are also stingy | 02:18 |
| zhenguo | wanghao: maybe you can register one for customers view, like why you choose mogan | 02:18 |
| shaohe_feng | for this is a good chance to promote the compony itself. | 02:18 |
| shaohe_feng | wanghao: This is a good idea. | 02:19 |
| openstackgerrit | Zhong Luyao proposed openstack/mogan: Add fault_info field for object Instance https://review.openstack.org/421192 | 02:19 |
| wanghao | zhenguo: Good idea, I'd like try it. | 02:19 |
| zhenguo | yes, it's really weird that you topic is selected, but they don't want you go | 02:19 |
| zhenguo | wanghao: thanks | 02:19 |
| wanghao | zhenguo: But not sure the boss will let me go, tooooo | 02:19 |
| zhenguo | wanghao: hah, just try | 02:19 |
| wanghao | stingy too.... | 02:19 |
| shaohe_feng | Barcelona is a good tourist city | 02:20 |
| wanghao | zhenguo: np | 02:20 |
| shaohe_feng | every want to visit this city. | 02:20 |
| little | I just want to travel there | 02:20 |
| wanghao | shaohe_feng: yes, I was been there few years ago. | 02:21 |
| zhenguo | hah, | 02:21 |
| little | wanghao: u r a rich man | 02:21 |
| shaohe_feng | so most of our employee are asiggned to visit Barcelona even without any task for Intel. | 02:21 |
| shaohe_feng | wanghao: rich man | 02:22 |
| wanghao | little: huawei send me there .... | 02:22 |
| wanghao | shaohe_feng: .... for business, you know | 02:22 |
| liusheng | wanghao: 巴展 | 02:22 |
| shaohe_feng | billionaire | 02:23 |
| wanghao | liusheng: bingo | 02:23 |
| zhenguo | ok, hah we can continue to discuss this in open discusstion topic | 02:23 |
| zhenguo | let's move on? | 02:23 |
| wanghao | zhenguo: sure | 02:24 |
| zhenguo | ok | 02:24 |
| shaohe_feng | go ahead. | 02:24 |
| zhenguo | #topic task tracking | 02:24 |
| zhenguo | #link https://etherpad.openstack.org/p/nimble-task | 02:24 |
| * zhenguo gives folks a few minutes to review | 02:24 | |
| zhenguo | looks like things are moving along nicely | 02:25 |
| zhenguo | we can align the task status one by one | 02:25 |
| * liusheng dont | 02:26 | |
| little | OK | 02:26 |
| zhenguo | shaohe_feng: seems no owner for rootwrap task | 02:26 |
| liusheng | zhenguo: why we still use nimble-task, lol | 02:27 |
| shaohe_feng | liusheng: we need to call shell cmd | 02:27 |
| zhenguo | liusheng: oh, sorry, my mistake | 02:28 |
| little | hah | 02:28 |
| zhenguo | #link https://etherpad.openstack.org/p/mogan-task | 02:28 |
| zhenguo | let's move to the mogan task | 02:28 |
| shaohe_feng | liusheng: we should limit the privilege scope | 02:29 |
| zhenguo | I have moved the rootwrap task to TODO | 02:29 |
| liusheng | shaohe_feng: got it | 02:29 |
| shaohe_feng | liusheng: we should not use "sudo" to do every thing. | 02:29 |
| liusheng | shaohe_feng: sure | 02:29 |
| shaohe_feng | liusheng: now in configdrive, I just use "sudo" | 02:29 |
| zhenguo | shaohe_feng: yes, but no owner takes taht now, so move it to TOTO | 02:29 |
| liusheng | shaohe_feng: I talked about the etherpad link, hah | 02:29 |
| zhenguo | s/TOTO/TODO | 02:30 |
| shaohe_feng | zhenguo: yes, move it TODO. we can use "sudo" first. | 02:30 |
| zhenguo | shaohe_feng: ok | 02:30 |
| zhenguo | so the first now is config drive | 02:30 |
| shaohe_feng | many example for rootwrap in very project. such as nova, cinder, ironic. | 02:31 |
| liusheng | shaohe_feng: just stole it | 02:31 |
| shaohe_feng | liusheng: yes. | 02:31 |
| little | yes | 02:31 |
| shaohe_feng | zhenguo: it is almost done. I'm testing it. | 02:31 |
| zhenguo | if so, please feel free to do that | 02:31 |
| little | many projects use it | 02:31 |
| zhenguo | shaohe_feng: awsome! | 02:31 |
| zhenguo | you test it with vm or bm | 02:32 |
| shaohe_feng | zhenguo: VM. | 02:32 |
| shaohe_feng | zhenguo: still not setup bm | 02:32 |
| zhenguo | shaohe_feng: ok, | 02:32 |
| shaohe_feng | zhenguo: when you come to here to setup bm? | 02:32 |
| zhenguo | shaohe_feng: hah, | 02:32 |
| zhenguo | shaohe_feng: sorry for that, as now my office is really far from downtown | 02:33 |
| zhenguo | shaohe_feng: But I will try later, or maybe remote? | 02:33 |
| little | HuanBaoYuan? | 02:33 |
| zhenguo | little: exactly | 02:33 |
| shaohe_feng | zhenguo: Maybe I can take you here by car. | 02:33 |
| little | shaohe_feng: cool | 02:34 |
| zhenguo | shaohe_feng: hah | 02:34 |
| shaohe_feng | zhenguo: let me think how to setup remote? | 02:34 |
| shaohe_feng | so sorry, these days do not feel well to delay the task. | 02:34 |
| zhenguo | shaohe_feng: no problems, really appreciate it | 02:35 |
| zhenguo | shaohe_feng: ok, if the config drive task is nealy done, it can be landed before Spring Festival maybe | 02:35 |
| liusheng | shaohe_feng: health is most important | 02:35 |
| wanghao | shaohe_feng: health! health! health! | 02:36 |
| shaohe_feng | zhenguo: we can find a public place for these bm and Switch | 02:36 |
| wanghao | shaohe_feng: three times for important thing. | 02:36 |
| little | shaohe_feng: you are not in Huawei! | 02:36 |
| shaohe_feng | wanghao: yes. | 02:36 |
| zhenguo | hah | 02:36 |
| wanghao | liusheng: .... | 02:36 |
| wanghao | little: ... | 02:36 |
| zhenguo | only liusheng and me are with huawei | 02:36 |
| wanghao | zhenguo: .... | 02:36 |
| shaohe_feng | health! health! health! and then $! $! $! | 02:37 |
| zhenguo | lol | 02:37 |
| * liusheng envy you | 02:37 | |
| wanghao | haha | 02:37 |
| liusheng | ￥￥￥ is enough | 02:37 |
| little | liusheng: hah | 02:38 |
| zhenguo | I'm thinking whether we need to hold a meetup or mid-cycle to gather everyone together | 02:39 |
| liusheng | zhenguo: in Beijing ? | 02:39 |
| zhenguo | maybe, hah | 02:39 |
| zhenguo | but not time left before Spring Festival | 02:39 |
| wanghao | zhenguo: that also needs boss to support.... | 02:39 |
| zhenguo | and it's really hard to buy a ticket now | 02:40 |
| zhenguo | wanghao: yes | 02:40 |
| liusheng | zhenguo, wanghao yes | 02:40 |
| wanghao | zhenguo: we just can take train to beijing... | 02:40 |
| zhenguo | wanghao: hah | 02:40 |
| zhenguo | not sure when it is the next bug smash | 02:41 |
| zhenguo | oh, just realized this is task tracking topic! | 02:41 |
| liusheng | zhenguo: haha | 02:41 |
| zhenguo | we should discuss these in next one, hah | 02:41 |
| zhenguo | let's move one to next task | 02:41 |
| zhenguo | notification | 02:41 |
| zhenguo | it's yours liusheng | 02:42 |
| liusheng | zhenguo: ok, the notifications related patches have been landed, it is the basic frame and an "instance.create.start" notification example | 02:42 |
| liusheng | zhenguo: since we may will have some change about the process, I propose don't add more notifications now | 02:43 |
| liusheng | zhenguo: what do you think ? | 02:43 |
| liusheng | ask to all you guys, lol | 02:44 |
| zhenguo | oh, just came back | 02:44 |
| zhenguo | some thing wrong with my IRC client | 02:44 |
| wanghao | liusheng: agree, we could have mechanism | 02:44 |
| liusheng | it is easy to add more notifications | 02:45 |
| little | liusheng: it's OK | 02:45 |
| zhenguo | agree, so we should keep the task here, or move it to DONE? | 02:45 |
| liusheng | zhenguo: may change the task to basic notification fram and move it to done ? | 02:46 |
| little | If only provide a mechanism, it should be done | 02:46 |
| zhenguo | ok, | 02:46 |
| wanghao | new notification should be added in new process changes. | 02:47 |
| liusheng | wanghao: yes | 02:47 |
| zhenguo | liusheng: please help to update it | 02:47 |
| zhenguo | let's move on | 02:47 |
| liusheng | zhenguo: ok | 02:47 |
| liusheng | btw, the console task | 02:47 |
| zhenguo | ok, please continue | 02:47 |
| liusheng | I have do some tests | 02:48 |
| zhenguo | does it work? | 02:48 |
| liusheng | zhenguo: no :( | 02:48 |
| zhenguo | oh | 02:48 |
| zhenguo | what's the problem | 02:49 |
| zhenguo | I remeber you can get console with ssh driver | 02:49 |
| liusheng | zhenguo: I tested it with start a new console proxy service, but seems it cannot work to connect to the shellinabox | 02:49 |
| liusheng | zhenguo: yes | 02:49 |
| zhenguo | liusheng: oh, yes, we should not use shellinabox | 02:49 |
| zhenguo | no existing proxy can support shellinabox | 02:49 |
| liusheng | zhenguo: first step, I planed to add a console proxy without consoleauth | 02:50 |
| liusheng | zhenguo: really ? | 02:50 |
| zhenguo | yes, | 02:50 |
| liusheng | zhenguo: so how should Iｄｏ？ | 02:50 |
| zhenguo | so I create a httpproxy before, it works well, but refused by community | 02:50 |
| zhenguo | liusheng: we have another console driver | 02:50 |
| zhenguo | which can leverage serialconsole proxy | 02:51 |
| liusheng | zhenguo: or just redirect the "virsh console" ? | 02:51 |
| zhenguo | liusheng: not sure, | 02:51 |
| liusheng | zhenguo: yes, I almost copied a serialconsole proxy from nova | 02:52 |
| zhenguo | liusheng: will dig after this meeting | 02:52 |
| liusheng | zhenguo: ok, thanks | 02:52 |
| zhenguo | ok, let's move on | 02:52 |
| zhenguo | next is quotas | 02:52 |
| zhenguo | it's yours wanghao | 02:52 |
| wanghao | okay, now I have built the quota models. | 02:52 |
| wanghao | Like Quota, Quota Driver, Resource. | 02:53 |
| zhenguo | wanghao: yes, I saw the patch | 02:53 |
| wanghao | so next work will implement the db operation. | 02:53 |
| wanghao | zhenguo: yes | 02:53 |
| zhenguo | but seems some conflicts now, need to rebase | 02:53 |
| wanghao | zhenguo: yes, will do it today. | 02:53 |
| zhenguo | wanghao: ok, thanks | 02:53 |
| zhenguo | oh, seems not much time left | 02:54 |
| zhenguo | if not much problems we will move on to multi instances create | 02:54 |
| zhenguo | hey little | 02:54 |
| wanghao | okay | 02:54 |
| little | yes | 02:54 |
| little | I have summit a frame patch, but the jeknis failed | 02:54 |
| zhenguo | seems because merge conflict? | 02:55 |
| zhenguo | or others | 02:55 |
| little | Both | 02:55 |
| little | I am trying to solve it today | 02:55 |
| zhenguo | oh, need to modify to satisfy pep,py27, make all our gates happy | 02:56 |
| little | yes | 02:56 |
| zhenguo | ok | 02:56 |
| shaohe_feng | zhenguo: There's a problem on evn | 02:57 |
| zhenguo | shaohe_feng: what | 02:57 |
| shaohe_feng | zhenguo: I can not run tox, after I re-install linux and devstack | 02:57 |
| zhenguo | shaohe_feng: reinstall linux? | 02:57 |
| liusheng | shaohe_feng: what's the error ? | 02:58 |
| zhenguo | hah maybe discuss the fault info task a bit? | 02:59 |
| zhenguo | hey luyao, are you around? | 02:59 |
| shaohe_feng | zhenguo: yes. | 02:59 |
| shaohe_feng | liusheng: let try tox again | 02:59 |
| zhenguo | luyao's patch seems ready | 02:59 |
| zhenguo | but do we need to add the fault info like nova | 02:59 |
| zhenguo | nova has a seperate table and save not only the string | 03:00 |
| zhenguo | maybe we can discuss in the patch, seems luyao is absent now | 03:00 |
| *** Jeffrey4l_ has joined #openstack-mogan | 03:00 | |
| zhenguo | and the lock instances | 03:01 |
| zhenguo | oh, oh no time left | 03:01 |
| zhangjl | zhenguo:.... | 03:01 |
| zhenguo | zhangjl: haha | 03:01 |
| zhenguo | we can discuss in our channel | 03:01 |
| liusheng | it is ok, we don't use #openstack-meeting, hah | 03:01 |
| zhangjl | yes | 03:01 |
| shaohe_feng | liusheng: $ tox -egenconfig | 03:02 |
| zhenguo | liusheng: hah, ok so we can continue or end the meeting? | 03:02 |
| shaohe_feng | liusheng: error: command 'x86_64-linux-gnu-gcc' failed with exit status 1 | 03:02 |
| shaohe_feng | liusheng: gcc not install? | 03:02 |
| liusheng | zhenguo: we can end it, and the meeting log will be ended. lol | 03:03 |
| zhenguo | ok, everyone, let's discuss in our channel, will end the meeting | 03:03 |
| zhenguo | liusheng: no problems, hah | 03:03 |
| little | ok | 03:03 |
| zhenguo | ok thanks you all for joining the meeting | 03:03 |
| shaohe_feng | liusheng: $ which x86_64-linux-gnu-gcc | 03:03 |
| zhenguo | #endmeeting | 03:03 |
| shaohe_feng | /usr/bin/x86_64-linux-gnu-gcc | 03:03 |
| openstack | Meeting ended Thu Jan 19 03:03:39 2017 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 03:03 |
| openstack | Minutes: http://eavesdrop.openstack.org/meetings/mogan/2017/mogan.2017-01-19-02.01.html | 03:03 |
| openstack | Minutes (text): http://eavesdrop.openstack.org/meetings/mogan/2017/mogan.2017-01-19-02.01.txt | 03:03 |
| openstack | Log: http://eavesdrop.openstack.org/meetings/mogan/2017/mogan.2017-01-19-02.01.log.html | 03:03 |
| shaohe_feng | liusheng: fatal error: Python.h: No such file or directory | 03:04 |
| *** Jeffrey4l has quit IRC | 03:04 | |
| zhangjl | yum install python-devel -y | 03:04 |
| zhenguo | we can have a break, and then continue the lock instance task | 03:04 |
| liusheng | shaohe_feng: seems some thing wrong with gcc ? | 03:04 |
| liusheng | zhenguo: ok | 03:05 |
| shaohe_feng | liusheng: yes. why missing Python.h? | 03:05 |
| luyao | zhenguo, sorry,I'm back | 03:05 |
| zhangjl | zhenguo:ok, it`s my turn | 03:05 |
| zhangjl | #link:https://review.openstack.org/#/c/420486/ | 03:06 |
| liusheng | shaohe_feng: are you in Ubuntu ? | 03:06 |
| zhenguo | hah | 03:06 |
| zhangjl | i registed a specs for lock instance | 03:06 |
| zhenguo | zhangjl: yes, we need to discuss more about the difference of lock and maintenance as I commented | 03:07 |
| zhangjl | zhenguo:yes | 03:07 |
| zhenguo | luyao: hi, we can discuss fault info later after lock, haha | 03:07 |
| zhenguo | hi all, do you think we need both lock and maintenance | 03:07 |
| luyao | zhenguo, ok | 03:07 |
| zhangjl | zhenguo:i think so | 03:08 |
| liusheng | shaohe_feng: may try to apt-get dist-uprade ? | 03:08 |
| zhenguo | zhangjl: I read the nova's lock/unlock codes | 03:08 |
| zhenguo | zhangjl: it allows the owner and admin to lock/unlock the instances | 03:09 |
| shaohe_feng | liusheng: OK, let me try | 03:09 |
| zhenguo | but admin can do any operations to a locked instances but owner can't | 03:09 |
| zhenguo | in fact, Ironic node can move to maintenance automatically if there's something wrong with the node | 03:10 |
| zhenguo | I just thinking whether we can leverage that | 03:10 |
| zhenguo | like we have a periodic task to collect maintenace node, and move the instance to maintenance state? | 03:11 |
| openstackgerrit | Tao Li proposed openstack/mogan: [WIP] Support multi-instance creating (part 1) https://review.openstack.org/421064 | 03:11 |
| zhenguo | seems a lock can cover most cases, we don't need to expose maintain interface to users/admins | 03:12 |
| zhenguo | wdyt? | 03:12 |
| zhangjl | zhenguo: maybe it`s good.While, in our company ,most of all thinks lock an instance to disallow user to operate is better | 03:12 |
| zhenguo | zhangjl: including the owner? | 03:13 |
| zhangjl | zhenguo:yes | 03:13 |
| zhenguo | zhangjl: if it loked by an admin, all common users can't operate it indeed | 03:14 |
| zhenguo | zhangjl: but owner can also lock it to prevent other users in the same tenant operate it | 03:14 |
| zhangjl | zhenguo: emm. maybe, it`s so different between our options | 03:15 |
| zhenguo | zhangjl: hah | 03:16 |
| zhangjl | zhenguo: imo, lock and unlock should be only for admin | 03:16 |
| zhenguo | zhangjl: and then | 03:16 |
| zhenguo | zhangjl: ok, wen can ignore the owner now, just talk about admins | 03:16 |
| zhenguo | zhangjl: if the admin locked the instance, common users can't do any operations on it, how about admins | 03:17 |
| zhangjl | zhenguo:sorry for my slowly typing | 03:18 |
| zhenguo | zhangjl: np, infact there's many syntax with my typing, hah | 03:18 |
| zhenguo | oh, I mean problems with syntax | 03:18 |
| zhangjl | zhenguo: because there are more admin role user in openstack | 03:18 |
| zhangjl | zhenguo: so , for security, if an instance is been locked , any one of admin can only unlock | 03:20 |
| zhangjl | zhenguo: then , we can avoid other admin to delete server | 03:20 |
| zhenguo | zhangjl: so other admin can also unlock it and then delete it, right? | 03:20 |
| zhangjl | zhenguo:emm..yes | 03:21 |
| zhangjl | zhenguo: i need your advice | 03:21 |
| zhenguo | zhangjl: another question is do we need to use instance status for lock state | 03:22 |
| zhenguo | zhagnjl: or need to introduce a new field | 03:22 |
| zhenguo | zhangjl: as we have a state machine to control states transition now, and I think you can lock the instance in any status | 03:23 |
| zhangjl | zhenguo: use instance status maybe is better | 03:23 |
| zhangjl | zhenguo:only active can be locked | 03:23 |
| zhenguo | zhangjl: so if the instance is in building status, and you lock it, it will move on to locked | 03:23 |
| zhangjl | zhenguo:no, only active instance can be lock | 03:24 |
| zhenguo | zhangjl: when the building is done, we need to transition the state from building to active, but it's now in locked state | 03:24 |
| zhangjl | zhenguo:otherwise, 409 would be raise | 03:24 |
| zhenguo | zhangjl: in fact, we allow deleting instance in any state now :( | 03:25 |
| zhenguo | zhangjl: and lock should be a seperate state not related to instance status, | 03:25 |
| shaohe_feng | liusheng: auto reboot linux after upgrade, :'( | 03:27 |
| shaohe_feng | 环境太恶劣了，每次devstack，都得多次才能成功。 :'( | 03:37 |
| zhenguo | lol | 03:52 |
| zhenguo | shaohe_feng: it's ok now? | 03:52 |
| *** kevinz has quit IRC | 04:10 | |
| *** kevinz has joined #openstack-mogan | 04:10 | |
| openstackgerrit | Tao Li proposed openstack/mogan: [WIP] Support multi-instance creating (part 1) https://review.openstack.org/421064 | 04:35 |
| zhenguo | shaohe_feng: I have a question about the policy control of API, if I wan't to deny power on, but allow power off, how to make it | 05:09 |
| zhenguo | shaohe_feng: currently we use a decorator, but we can't get the argument like 'target' for set power state, so seems there are limitations, maybe that's why nova doesn't choose to use decorator, wdyt? | 05:13 |
| * zhenguo zZZZ.. | 05:14 | |
| shaohe_feng | zhenguo: yes. the we need to hardcode the "power off" | 05:18 |
| shaohe_feng | zhenguo: but we can think a more better to do it. | 05:18 |
| shaohe_feng | zhenguo: such as use filter to get the traget. | 05:20 |
| shaohe_feng | zhenguo: normally, such as "power off" or "power on", should be seperated api. | 05:21 |
| shaohe_feng | zhenguo: you can seen Lu Hongbin's email of zun. | 05:21 |
| *** yuntongjin has joined #openstack-mogan | 05:22 | |
| shaohe_feng | zhenguo: nova does mixture many action in one api. | 05:22 |
| *** yuntongjin has quit IRC | 05:26 | |
| *** yuntongjin has joined #openstack-mogan | 05:28 | |
| shaohe_feng | zhenguo: wo can hardcode. or decorator for every action. | 05:32 |
| shaohe_feng | zhenguo: such as: | 05:32 |
| *** wxy has quit IRC | 05:38 | |
| zhenguo | shaohe_feng: seems not easy to make it | 05:58 |
| zhenguo | shaohe_feng: like this https://github.com/openstack/mogan/blob/master/mogan/api/controllers/v1/instances.py#L122-L125 | 05:59 |
| zhenguo | shaohe_feng: I want to use different policy for different 'target' | 06:00 |
| shaohe_feng | zhenguo: we can use nova's hard code, but decorator first. anyway, nova's api define is not good. I really seen other project use nova style defination. so many developers question nova's style. | 06:02 |
| shaohe_feng | s/I really seen/I really never seen | 06:03 |
| zhenguo | shaohe_feng: yes, but we also face the same problem | 06:03 |
| shaohe_feng | zhenguo: I send your a picture on wechat. | 06:03 |
| zhenguo | shaohe_feng: like power state, we use target to indicate which operation should be taken | 06:03 |
| zhenguo | ok | 06:03 |
| shaohe_feng | zhenguo: and let me think a more better method. | 06:06 |
| zhenguo | shaohe_feng: thanks | 06:06 |
| zhenguo | shaohe_feng: I remember you said neutron uses another way | 06:06 |
| zhenguo | shaohe_feng: will be right back. | 06:07 |
| * zhenguo brb | 06:07 | |
| shaohe_feng | shaohe_feng: yes. they check it in hook phase. it is similor to our way, avoid hard code. | 06:08 |
| shaohe_feng | zhenguo: wo check it after hook, but before the app call | 06:08 |
| zhenguo | shaohe_feng: thanks | 06:18 |
| zhenguo | shaohe_feng: I want to introduce a new URI like /instances/instance_uuid/states/provision | 06:20 |
| zhenguo | shaohe_feng: we can PUT it with {'target': 'rebuild'} or other actions related to provision state | 06:20 |
| openstackgerrit | Tao Li proposed openstack/mogan: [WIP] Support multi-instance creating (part 1) https://review.openstack.org/421064 | 06:21 |
| zhenguo | seems our API defination makes more sense | 06:21 |
| shaohe_feng | zhenguo: yes. | 06:21 |
| zhenguo | shaohe_feng: and a new networks sub resouces for instances like /instances/instance_uuid/networks/floatingip or port or other neworks related thing | 06:22 |
| shaohe_feng | zhenguo: user is can easy to send their request in this way. | 06:23 |
| zhenguo | shaohe_feng: and console seems should be a new sub resouce like /instances/instance_uuid/console ? or add it under states? | 06:23 |
| zhenguo | shaohe_feng: yes, and they know what resouce their request | 06:23 |
| shaohe_feng | zhenguo: all sub-resouces in it's father resoucce, is not a good way. | 06:24 |
| shaohe_feng | zhenguo: do does Lu Hongbin use the api style at last? | 06:24 |
| shaohe_feng | what does? | 06:25 |
| zhenguo | shaohe_feng: not sure, be cause the use a action instead of resource | 06:25 |
| zhenguo | shaohe_feng: some one point out it's urgly | 06:25 |
| shaohe_feng | redhat use instances/instance_uuid/states/power/shutdown instances/instance_uuid/states/power/start | 06:26 |
| shaohe_feng | zhenguo: opensource can not satisfy | 06:26 |
| shaohe_feng | zhenguo: opensource can not satisfy every one. | 06:26 |
| shaohe_feng | some many people, different flavor. | 06:27 |
| shaohe_feng | in Chinese: It is difficult to cater for all tastes. | 06:28 |
| zhenguo | hah | 06:29 |
| zhenguo | shaohe_feng: but start is a verb | 06:29 |
| zhenguo | shaohe_feng: not sure if a verb should present in URI | 06:29 |
| shaohe_feng | zhenguo: yes. actually, redhat also discuss the verb for a long time. | 06:30 |
| zhenguo | shaohe_feng: yes, REST only have verbs like PUT/POST/DELETE | 06:31 |
| zhenguo | shaohe_feng: what's the meaning of PUT a start | 06:31 |
| shaohe_feng | zhenguo: maybe we let luyao introduce whether Fielding has metion the verb. | 06:31 |
| shaohe_feng | PUT means updata. | 06:32 |
| zhenguo | shaohe_feng: ok | 06:32 |
| shaohe_feng | zhenguo: so /instances/instance_uuid/states/provision, PUT it with {'target': 'rebuild'} | 06:32 |
| zhenguo | yes, but I don't think you can PUT a verb | 06:32 |
| zhenguo | shaohe_feng: yes | 06:32 |
| shaohe_feng | zhenguo: means you want to update the status of provision, the target status is 'rebuild'. | 06:33 |
| zhenguo | shaohe_feng: yes, just like power satte | 06:33 |
| shaohe_feng | zhenguo: here, the rebuid is a status, a noun. | 06:34 |
| zhenguo | shaohe_feng: yes, so I think a target seems more better | 06:34 |
| shaohe_feng | zhenguo: we want to change(PUT) the stats/provision to 'rebuild'. | 06:35 |
| zhenguo | shaohe_feng: yes, do you think it's better | 06:35 |
| shaohe_feng | Fielding is the author of RESTFUL, and he is aslo an main developer of Apachi. | 06:36 |
| shaohe_feng | zhenguo: yes. I think it looks resonable. | 06:37 |
| zhenguo | shaohe_feng: does luyao familiar with that, she read up the papter from Fielding? | 06:37 |
| shaohe_feng | but verb is always is a dispute | 06:38 |
| shaohe_feng | zhenguo: she has read the paper. | 06:38 |
| zhenguo | shaohe_feng: yes, I think our current way is more resonable | 06:43 |
| shaohe_feng | zhenguo: OK. let's follow this way. | 06:45 |
| zhenguo | shaohe_feng: Okay | 06:45 |
| shaohe_feng | zhenguo: our annual meeting at 4 o'clock | 06:55 |
| zhenguo | shaohe_feng: oh, have fun | 06:56 |
| shaohe_feng | zhenguo: good luck is more important than fun. | 06:56 |
| shaohe_feng | zhenguo: I want an iphone. | 06:56 |
| zhenguo | shaohe_feng: hah, good luck and have fun | 06:56 |
| shaohe_feng | zhenguo: Thank you very much. :-)) | 06:57 |
| zhenguo | shaohe_feng: iphone iphone iphone | 06:57 |
| zhenguo | oh, and good luck for luyao | 06:58 |
| wanghao | shaohe_feng: iphone iphone iphone | 06:58 |
| zhenguo | and yuntongjin, lol | 06:58 |
| shaohe_feng | everyone good luck, means I missing the chance of iphone. :-D | 06:59 |
| shaohe_feng | hopeful yuntongjin do not like iphone. | 07:00 |
| shaohe_feng | wanghao: thank you. iphone. | 07:00 |
| wanghao | shaohe_feng: haha | 07:01 |
| shaohe_feng | luyao and yuntongjin, android is also good. | 07:01 |
| zhenguo | huawei is good | 07:01 |
| wanghao | shaohe_feng: is there only one iphone? | 07:01 |
| shaohe_feng | huawei is good. luyao and yuntongjin. | 07:02 |
| shaohe_feng | wanghao: seem more than one. | 07:02 |
| wanghao | shaohe_feng: cool | 07:02 |
| shaohe_feng | wanghao: the special prize seems is imac pro | 07:02 |
| shaohe_feng | only one. | 07:03 |
| wanghao | shaohe_feng: wow, very cooooool | 07:03 |
| wanghao | shaohe_feng: wish you can get it, better than iphone more. | 07:03 |
| shaohe_feng | Thanks. seems too greedy. :-D | 07:04 |
| openstackgerrit | Tao Li proposed openstack/mogan: [WIP] Support multi-instance creating (part 1) https://review.openstack.org/421064 | 07:14 |
| openstackgerrit | wanghao proposed openstack/mogan: [WIP]Quota support in Mogan(part 1) https://review.openstack.org/417326 | 07:18 |
| *** yuntongjin has quit IRC | 07:34 | |
| zhenguo | hi all, do you think the current instance_type way is enough or we need to refactor it | 08:14 |
| zhenguo | I'd like to add filters for ports account and types when scheduling | 08:15 |
| zhenguo | liusheng, wanghao, little, zhangjl: hi | 08:16 |
| zhenguo | s/ports account/ports quantity | 08:16 |
| little | Add a new filter? | 08:17 |
| zhenguo | little: yes | 08:17 |
| zhangjl | zhenguo:sorry , i`m in meeting | 08:17 |
| zhenguo | little: currently we just schedule a node to satisfy, but not consider the ports | 08:17 |
| zhenguo | zhangj: np | 08:17 |
| *** wxy has joined #openstack-mogan | 08:17 | |
| liusheng | zhenguo: obviously we need to refactor/improve the current instance_type | 08:18 |
| liusheng | zhenguo: it is fake one, right ? | 08:18 |
| zhenguo | liusheng: hah | 08:18 |
| zhenguo | liusheng: how about use different types | 08:18 |
| zhenguo | liusheng: instance_type + port_type | 08:18 |
| little | zhenguo: can you give more details? | 08:19 |
| zhenguo | liusheng: and maybe other types | 08:19 |
| zhenguo | little: ok | 08:19 |
| liusheng | zhenguo: seems, the port type should be property of network paremeters | 08:19 |
| zhenguo | openstack baremetal server create --flavor H1 --image test --nic net-id=aa,port_type=10GE | 08:20 |
| zhenguo | liusheng: yes, but if we have a way to schedule it, seems it's enough | 08:20 |
| little | so mogan can select 10GE server to deploy? | 08:20 |
| liusheng | zhenguo: if we put the port_type in flavor, how to create a server with port specified ? | 08:21 |
| zhenguo | little: exactly the serve with 10GE port | 08:21 |
| little | If this is done, I think can we add ram, disk and cpu filters? | 08:22 |
| zhenguo | liusheng: not esay | 08:22 |
| zhenguo | s/esay/easy | 08:22 |
| zhenguo | little: seems no need to add that filters like nova | 08:22 |
| liusheng | zhenguo: we will support creating with port specified, right ? | 08:22 |
| zhenguo | little: as the specs for a server is fixed | 08:23 |
| zhenguo | liusheng: yes | 08:23 |
| zhenguo | liusheng: but need to specify the port type to specified network | 08:23 |
| zhenguo | liusheng: like the current way | 08:23 |
| liusheng | zhenguo: yes | 08:23 |
| zhenguo | liusheng: if it's in instance_type, not sure how to handle it | 08:23 |
| little | maybe this is very useful | 08:23 |
| zhenguo | little: for example | 08:24 |
| zhenguo | little: nova has that filters is because that want to select a node with enough properties | 08:24 |
| zhenguo | little: but we are not, we just want to select a node and use all resources | 08:24 |
| little | maybe user select specific spec server | 08:25 |
| zhenguo | little: hah but we can't support that | 08:25 |
| zhenguo | little: can you compose a server with specified specs? | 08:26 |
| zhenguo | little: we just have server types of servers, users can only select some one from it | 08:26 |
| little | so this is the differnet from nova,hah | 08:26 |
| zhenguo | little: hah, yes, | 08:26 |
| zhenguo | little: nova's flavor is not suit for us | 08:27 |
| little | if this is ok for users, i agree it | 08:27 |
| liusheng | zhenguo: if we create a port firstly with port_type specified, we don't need to select a host, right ? | 08:27 |
| zhenguo | little: and you can add details specs in the instance_type description | 08:27 |
| little | zhenguo: yeah | 08:27 |
| liusheng | zhenguo: only when bind the port ? | 08:28 |
| zhenguo | liusheng: no, we need to select a node with that port type | 08:28 |
| liusheng | zhenguo: oh, that is different with nova | 08:28 |
| zhenguo | liusheng: and you maybe specified 4 networks when creating a server | 08:28 |
| wanghao | zhenguo: for mogan, instance type like template, including cpu ,ram, disk, port, etc. | 08:28 |
| zhenguo | wanghao: yes, but we only filter the type | 08:29 |
| zhenguo | liusheng: and if we only filter instance type, a node may only have 2 ports, how to handle that? | 08:29 |
| liusheng | zhenguo: oh, we need to specifying mac address when create port | 08:29 |
| zhenguo | nova will just fail in this case | 08:29 |
| zhenguo | liusheng: seems just port type is enough | 08:30 |
| zhenguo | liusheng: users don't care about the MAC, just types like 1GE， 10GE | 08:30 |
| liusheng | zhenguo: yes, | 08:30 |
| little | maybe currently just filter port type | 08:31 |
| zhenguo | little: and the quntity | 08:31 |
| liusheng | zhenguo: so it is not a common use case that we creating a server with an exist port specified, right ? | 08:31 |
| zhenguo | quantity | 08:31 |
| wanghao | zhenguo: yes, I mean we put the port specs into instance_type, and filter it, if no nodes found, we failed. | 08:31 |
| zhenguo | wanghao: but this way, we can't specify with port bind to which neutron network | 08:32 |
| *** openstackgerrit has quit IRC | 08:33 | |
| little | I will go home now | 08:33 |
| zhenguo | little: oh, so early | 08:34 |
| zhenguo | wanghao: like --nic net-id=A, port-type=1GE; net-id=B,port-type=10GE | 08:34 |
| wanghao | zhenguo: if we use extra_spec in instance_type? we can specify those information in it. | 08:34 |
| zhenguo | wanghao: but you can't match the networks | 08:35 |
| wanghao | zhenguo: : or we build a dict for ports | 08:35 |
| wanghao | like this cpu=1, mem=10G, ports=[{port:id, port-type:1GE, net-id:B...}] | 08:36 |
| liusheng | zhenguo: seems it is more reasonable to put it in instance type | 08:36 |
| wanghao | zhenguo: even maybe cpu, mem is dict too, you can specify some cpu type, | 08:38 |
| *** little has quit IRC | 08:38 | |
| wanghao | zhenguo: if we can get those hardware information. | 08:38 |
| zhenguo | wanghao: net-id is not hardware info | 08:39 |
| zhenguo | wanghao: it's neutron networks | 08:39 |
| wanghao | zhenguo: yes, I see, just for example. | 08:39 |
| zhenguo | wanghao: yes, I mean if we put port type to instance_type, we can't match the networks | 08:40 |
| *** zhangjl has quit IRC | 08:40 | |
| wanghao | zhenguo: I mean, user can specify some hardware info and other resources in Openstack too. | 08:40 |
| zhenguo | wanghao: but for hardware, we not only care about the quantity of CPU, MEM, but also the type of it | 08:41 |
| zhenguo | wanghao: RedHat just add a feature to nova, | 08:41 |
| liusheng | wanghao: zhenguo may means we can support the a type of hardware ability in instance_type | 08:41 |
| zhenguo | they can tag the device | 08:41 |
| liusheng | wanghao: then scheduler will select nodes that can satisfy the ability | 08:42 |
| wanghao | I see. | 08:42 |
| zhenguo | liusheng: but we don't filter it | 08:42 |
| zhenguo | liusheng: that's would be more complex | 08:42 |
| zhenguo | liusheng: you can add more instance type if you want | 08:42 |
| zhenguo | liusheng: why we need to filter all hardware properties | 08:43 |
| zhenguo | just a description to list all properties to show users seems more easy | 08:43 |
| liusheng | zhenguo: we only need ti filter the properties only we want, right ? | 08:43 |
| liusheng | zhenguo: just use ronic node-list ? | 08:43 |
| zhenguo | liusheng: the users just want to select a node, right? | 08:44 |
| zhenguo | we can just think the node is a device | 08:44 |
| liusheng | zhenguo: yes, but if we have many nodes, it is not user-friend to select a node for users, and especially after we supportting multiple creation | 08:46 |
| liusheng | zhenguo: seems it the works schduler should do | 08:46 |
| zhenguo | liusheng: if we have many nodes, we can just divide them into seperate groups | 08:47 |
| zhenguo | liusheng: admin can do that | 08:47 |
| liusheng | zhenguo: you mean like ag ? | 08:47 |
| zhenguo | liushegn: no, just use instance type | 08:48 |
| zhenguo | liusheng: like Gold, Silver... | 08:48 |
| liusheng | zhenguo: lol | 08:48 |
| zhenguo | liusheng: no need to make us more complex | 08:48 |
| liusheng | zhenguo: hmmm, but that seems the scheduler is not intelligent enough | 08:49 |
| zhenguo | liusheng: hah, the new scheduler for ironic is just follow this way | 08:50 |
| zhenguo | liusheng: they maybe copy us | 08:50 |
| zhenguo | liusheng: I mean the nova's palcement | 08:50 |
| zhenguo | liusheng: as if we have more properties in instance_type, the scheduler would be like P1=V1, P2=V2, P3=V3,P4=V4... | 08:51 |
| zhenguo | why we don't just use P1+P2+P3+P4=V1+V2+V3+v4 | 08:51 |
| zhenguo | hah, it's fixed | 08:52 |
| liusheng | zhenguo: make seense | 08:55 |
| liusheng | zhenguo: may we need to support the basic properties in instance_type firstly, that is a hight priority task, hah | 08:57 |
| zhenguo | liusheng: hah, this question is asked by anyone contributed to Mogan, especially at the beginning of this project. | 08:57 |
| liusheng | zhenguo: btw, do you think we need to replace the *instance* word in mogan ? | 08:57 |
| zhenguo | liusheng: ... | 08:57 |
| zhenguo | liusheng: I also don't want to put basic properties to instance_type | 08:58 |
| zhenguo | liusheng: seems the current way is enough | 08:58 |
| zhenguo | as we don't filter it | 08:59 |
| zhenguo | we just match the types, instance_type and port_type and maybe disk_type in future | 08:59 |
| zhenguo | every type is specified with the resources | 09:00 |
| liusheng | zhenguo: so the instance type will include nothing ? | 09:00 |
| zhenguo | server->instance_type, network->port_type, raid->disk_type | 09:00 |
| zhenguo | liusheng: yes, lol | 09:00 |
| liusheng | zhenguo: oh, i see | 09:00 |
| liusheng | zhenguo: hmm, that is werid!!! | 09:01 |
| zhenguo | liusheng: just think the server is a device, we don't care about what's inside | 09:01 |
| liusheng | zhenguo: why we need instance_type | 09:01 |
| zhenguo | liusheng: if we don't have instance_type, how to choose a server | 09:01 |
| zhenguo | we use instance_type to choose a server | 09:01 |
| zhenguo | port_type to choose a ehternet | 09:01 |
| liusheng | zhenguo: the instance type is more like a tag on the server | 09:01 |
| zhenguo | liusheng: yes | 09:02 |
| zhenguo | liusheng: the difference is just the name | 09:02 |
| liusheng | zhenguo: so we just need a tag than instance type | 09:02 |
| zhenguo | liusheng: yes, the difference is just the name | 09:02 |
| liusheng | zhenguo: the instance type make things more complex | 09:02 |
| zhenguo | liusheng: why | 09:02 |
| liusheng | zhenguo: so how to deprecate it ? haha | 09:03 |
| zhenguo | lol | 09:03 |
| liusheng | zhenguo: we need to support a set of CURD api on it | 09:03 |
| liusheng | zhenguo: and extra | 09:03 |
| zhenguo | you mean instance_type? | 09:03 |
| liusheng | zhenguo: tag on a resource is a common usage | 09:03 |
| liusheng | zhenguo: yes | 09:03 |
| zhenguo | liusheng: you mean just specify a tag name when creating a instance? | 09:04 |
| liusheng | zhenguo: yes | 09:04 |
| zhenguo | liusheng: it's unacceptable, as we also want to show users, the specs of the server they request | 09:04 |
| liusheng | zhenguo: if instance type include nothing, and it actually same with tag | 09:05 |
| liusheng | zhenguo: you mean the extra specs ? | 09:05 |
| zhenguo | liusheng: it has a description | 09:05 |
| liusheng | zhenguo: so it only for a description ? | 09:06 |
| liusheng | zhenguo: ... | 09:06 |
| zhenguo | liusheng: and extra specs | 09:07 |
| zhenguo | liusheng: and for tag, how to manage it | 09:07 |
| zhenguo | liusheng: you may also need to add CURD api on it | 09:07 |
| liusheng | zhenguo: tag is a attribute of server | 09:07 |
| liusheng | zhenguo: yes, set/unset | 09:08 |
| zhenguo | liusheng: before requesting a server, users should know how many types of instances you provide, and which one he wants to choose, how to handle that | 09:08 |
| liusheng | zhenguo: seems that will easy mislead users | 09:08 |
| liusheng | zhenguo: so it is more like a tag on server | 09:11 |
| zhenguo | liusheng: no | 09:11 |
| liusheng | zhenguo: to filter different type of instances | 09:11 |
| zhenguo | liusheng: yes, | 09:11 |
| zhenguo | liusheng: but that's enough | 09:11 |
| zhenguo | liusheng: you can use capabilites to math with intance type extras if you want, you can specify any key/value there | 09:12 |
| liusheng | zhenguo: but that won't be used by scheduler, right ? | 09:12 |
| zhenguo | liusheng: we have a capabilities filter | 09:13 |
| zhenguo | liusheng: ironic node also has a capabilities field for this | 09:13 |
| zhenguo | liusheng: so extra specs can satisfy everything you want to filter | 09:14 |
| liusheng | zhenguo: so it also indicate some actual properties of a ironic node | 09:14 |
| zhenguo | liusheng: but we only support a name for filter in instance_type like volume_type | 09:14 |
| zhenguo | liusheng: which properties you mean | 09:15 |
| liusheng | zhenguo: something that represent th capabilities of nodes | 09:15 |
| zhenguo | liusheng: we only use extra specs for special properties | 09:15 |
| zhenguo | liusheng: yes, just special properties | 09:15 |
| liusheng | zhenguo: yes, I agree taht | 09:15 |
| zhenguo | liusheng: cinder's volume type also includes nothing | 09:16 |
| liusheng | zhenguo: I don't object to saperate the properties into type, disk type, network type | 09:16 |
| liusheng | zhenguo: but I don't think it reasonable that the the functionality is jsut like a tag on a server | 09:17 |
| zhenguo | liusheng: as we can't combine the together i think | 09:17 |
| zhenguo | liusheng: hah, it's really enough | 09:17 |
| liusheng | zhenguo: or only for a "description" | 09:17 |
| zhenguo | liusheng: name, description, extra_specs | 09:17 |
| liusheng | zhenguo: yes, the extra_epecs we can spcify some properties to indicate the capabilities | 09:18 |
| liusheng | zhenguo: I know about a bit of cinder | 09:18 |
| zhenguo | liusheng: yes, why you want to put cpu, mem in the instance_type? | 09:18 |
| liusheng | zhenguo: no, I don't think so | 09:19 |
| zhenguo | liusheng: so you agree we don't need those things? | 09:19 |
| liusheng | zhenguo: I have said I don't object seperate them into disk type, net-type, instance_tyupe | 09:19 |
| zhenguo | liusheng: ok, hah | 09:20 |
| liusheng | zhenguo: no... we can add disk_type, net_type, instance_type | 09:20 |
| liusheng | zhenguo: but I think the instance_type is not only for a description and works like a tag | 09:20 |
| zhenguo | seems it's ok, not sure why you don't like it, hah | 09:22 |
| liusheng | zhenguo: :P, I never don't like it. hah | 09:22 |
| wanghao | sorry guys, it's time to leave | 09:22 |
| wanghao | bye | 09:22 |
| zhenguo | liusheng: hah, just like when you create a volume, you just choose a type | 09:23 |
| liusheng | wanghao: bye | 09:23 |
| zhenguo | wanghao: bye | 09:23 |
| *** wanghao has quit IRC | 09:23 | |
| liusheng | zhenguo: yes, but the volume type can include some properties that indicate the capabilities of storage backends | 09:23 |
| zhenguo | but the volume type is just a name, there's no IOPS, speed, or other properties | 09:23 |
| liusheng | zhenguo: no, that is just a volume type you don't have specific requirement | 09:24 |
| zhenguo | liusheng: why I don't have a specified requirement | 09:25 |
| zhenguo | liusehng: I may want to create a volume with high speed and high IOPS | 09:25 |
| liusheng | zhenguo: I mean you just want a volume, and you don't have specific requirement of the volume | 09:25 |
| zhenguo | no | 09:25 |
| liusheng | zhenguo: hat need the storage backend support | 09:26 |
| liusheng | zhenguo: that | 09:26 |
| zhenguo | liusheng: yes, that's just like us | 09:26 |
| liusheng | zhenguo: and you need to create a volume type for the storage backend with some properties specified | 09:26 |
| zhenguo | liusheng: volume type also just has a name and description | 09:27 |
| liusheng | zhenguo: no | 09:30 |
| liusheng | zhenguo: the extra_spec is important for volume type | 09:31 |
| zhenguo | liusheng: hah, we also has a extra_spec | 09:31 |
| liusheng | zhenguo: if we have a high performance volume backend, and we want to create volumes on it, we need create a volume type with a key-value in volume type's extra specs | 09:32 |
| liusheng | it is like {"volume_backend_name": "xxx"} | 09:32 |
| zhenguo | liusheng: you can also do this for instance_type | 09:32 |
| liusheng | zhenguo: that is important | 09:32 |
| liusheng | zhenguo: yes, so it absolutely not only a name and description | 09:32 |
| zhenguo | hah, ye | 09:32 |
| *** little has joined #openstack-mogan | 09:33 | |
| * zhenguo brb | 09:35 | |
| liusheng | zhenguo: maybe we have same ideas but misunderstand :P | 09:35 |
| zhenguo | liusheng: yes, I think we can just keep the current way, and add new filter for ports, maybe need more discuss if new requiremetns come | 09:36 |
| liusheng | zhenguo: yes, but seem we need to support real type firstly now, I mean the real disk size, cpu num, ram size. | 09:37 |
| liusheng | zhenguo: that is hight proirty | 09:38 |
| zhenguo | liusheng: again, why we need support that | 09:54 |
| liusheng | zhenguo: we only need to create server always be 10G disk, 10G ram, 1 cpu ? | 09:57 |
| zhenguo | liusheng: no, instance type is indicate that | 09:58 |
| zhenguo | we don't need to filter that things | 09:58 |
| zhenguo | like type1 is for 10Gdisk, 10Gram | 09:58 |
| zhenguo | i have to go for dinner | 09:58 |
| zhenguo | ttyl | 09:58 |
| liusheng | zhenguo: ok, got it | 09:58 |
| liusheng | zhenguo: if so, we also need to support than hardcoded, right ? | 09:59 |
| liusheng | zhenguo: you have type1, type2, typen, but they all 10G disk, 10G ram, 1 cpu | 10:00 |
| *** kevinz has quit IRC | 10:30 | |
| *** little has quit IRC | 10:38 | |
| zhenguo | liusheng: no, type1-10Gdisk-10Gram-1cpu, type2-100Gdisk-100Gram-100cpu | 10:39 |
| liusheng | zhenguo: how to support that ? | 10:40 |
| zhenguo | liusheng: ? | 10:40 |
| zhenguo | liusheng: we already supported that, you can set instance_type to different nodes according this | 10:41 |
| liusheng | zhenguo: I mean we have hardcoded the 10G disk, 10G ram, 1 cpu in code, how to support different size with different type ? | 10:41 |
| zhenguo | liusheng: why we have hardcoded that | 10:41 |
| liusheng | zhenguo: https://github.com/openstack/mogan/blob/master/mogan/engine/baremetal/ironic.py#L76-L98 | 10:42 |
| liusheng | zhenguo: am I wrong ? | 10:42 |
| zhenguo | liusheng: oh, sorry | 10:42 |
| liusheng | zhenguo: the current instance type has no effect | 10:42 |
| zhenguo | liusheng: it's just used for schedule | 10:42 |
| zhenguo | liusheng: and instance info to ironic node is useless | 10:43 |
| zhenguo | liusheng: we just use it for partition | 10:43 |
| *** little has joined #openstack-mogan | 10:43 | |
| zhenguo | liusheng: so we need to add partition support but not related to instance type | 10:43 |
| liusheng | zhenguo: oh, I thought that is size of the all instances, I was wrong | 10:44 |
| zhenguo | liusheng: hah, cpu and mem is useless | 10:44 |
| liusheng | zhenguo: I am not familiar with that in ironic :( | 10:44 |
| zhenguo | liusheng: but root_disk, swap, and ephermeral is used to partition | 10:45 |
| zhenguo | liusheng: we need to consider this | 10:45 |
| liusheng | zhenguo: ok, things become clear, lol | 10:45 |
| zhenguo | liusheng: haha, but not sure how to handle paritions | 10:45 |
| liusheng | zhenguo: that seems another task | 10:46 |
| zhenguo | liusheng: hah, yes | 10:46 |
| zhenguo | liusheng: let me first check whether instance_info is useless like what I just said, hah | 10:47 |
| zhenguo | liusheng: thanks for pointing it out :P | 10:48 |
| liusheng | zhenguo: so we actually cannot set the cpu num and mem size for an instance, right ? the instance will use the all the resource of the node | 10:48 |
| zhenguo | liusheng: yes | 10:48 |
| liusheng | zhenguo: OK, I get it! | 10:49 |
| zhenguo | liusheng: let me remove unneeded info to see whether it works | 10:50 |
| liusheng | zhenguo: ok | 10:50 |
| zhenguo | liusheng: I think instance info is used by nova to show the instance properties to users although it's wrong | 10:51 |
| liusheng | zhenguo: maybe, but it can be set to ironic node | 10:52 |
| zhenguo | liusheng: yes | 10:52 |
| *** little has quit IRC | 10:53 | |
| *** openstackgerrit has joined #openstack-mogan | 11:03 | |
| zhenguo | liusheng: just send a patch to clean it https://review.openstack.org/#/c/422551/ | 11:04 |
| liusheng | zhenguo: cool | 11:05 |
| zhenguo | liusheng: not sure if local_gb is also useless, let me test | 11:06 |
| liusheng | zhenguo: ok | 11:06 |
| zhenguo | liusheng: but it's really ugly that we hardcode the partitions | 11:07 |
| liusheng | zhenguo: hah, I always thought it is related with the instance_type | 11:07 |
| zhenguo | liusheng: hah, sorry for the misleading | 11:07 |
| liusheng | zhenguo: np, it is just because I am not familiar with ironic | 11:08 |
| zhenguo | liusheng: hah | 11:09 |
| zhenguo | liusheng: currently ironic only support disk_layout like [root_gb, swap_mb, ephermeral_gb] | 11:24 |
| zhenguo | liusheng: so we can't support more advanced partition, only can specify these fields | 11:24 |
| liusheng | zhenguo: it is only for root partition ? | 11:25 |
| zhenguo | liusheng: root_gb is for roto partition, also can specify swap and ephermeral | 11:25 |
| zhenguo | liusheng: in fact, ephermeral is other partition rather than root parition | 11:26 |
| zhenguo | liusheng: these partitions will in one disk | 11:26 |
| liusheng | zhenguo: oh, so may we will support user disk ? | 11:26 |
| zhenguo | liusheng: seems not | 11:27 |
| zhenguo | liusheng: all partitions in one disk | 11:27 |
| zhenguo | liusheng: and we may support raid later | 11:27 |
| liusheng | zhenguo: I mean does the bm can support attaching cinder volume ? | 11:27 |
| zhenguo | liusheng: currently not | 11:28 |
| zhenguo | liusheng: someone is working on boot from volume | 11:28 |
| liusheng | zhenguo: may will support :) | 11:28 |
| zhenguo | liusheng: yes | 11:28 |
| openstackgerrit | Zhenguo Niu proposed openstack/mogan: Remove unused instance info https://review.openstack.org/422551 | 11:29 |
| zhenguo | liusheng: I remember nova has an option about auto disk config | 11:30 |
| zhenguo | liusheng: only leave a root_gb in instance_info now, will add another patch to remove the hardcoded 10, maybe can use node.properties.local_gb and ram for swap | 11:31 |
| liusheng | zhenguo: yes, I can search the words in Nova's code, it is a attribute of Instance object | 11:38 |
| zhenguo | liusheng: yes, seems can leave partition to users, only create a root part | 11:39 |
| *** little has joined #openstack-mogan | 12:00 | |
| openstackgerrit | Zhenguo Niu proposed openstack/mogan: Set node local_gb to instance_info root_gb https://review.openstack.org/422586 | 12:26 |
| openstackgerrit | Merged openstack/mogan: Remove unused instance info https://review.openstack.org/422551 | 12:45 |
| *** kevinz has joined #openstack-mogan | 13:00 | |
| *** little has quit IRC | 13:07 | |
| *** kevinz has quit IRC | 13:14 | |
| *** kevinz has joined #openstack-mogan | 13:15 | |
| *** kevinz has quit IRC | 13:43 | |
| *** kevinz has joined #openstack-mogan | 13:43 | |
| *** kevinz has quit IRC | 15:35 | |
| openstackgerrit | Zhenguo Niu proposed openstack/mogan: Remove ironic node states request for instance https://review.openstack.org/422737 | 16:26 |
| *** liusheng has quit IRC | 17:18 | |
| *** liusheng has joined #openstack-mogan | 17:18 | |
| openstackgerrit | Merged openstack/mogan: Remove ironic node states request for instance https://review.openstack.org/422737 | 18:14 |
| *** openstackgerrit has quit IRC | 19:33 | |
| *** Jeffrey4l_ has quit IRC | 21:35 | |
| *** Jeffrey4l_ has joined #openstack-mogan | 21:36 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!