Monday, 2017-01-16

*** wanghao has joined #openstack-mogan00:41
*** liusheng has quit IRC01:10
*** liusheng has joined #openstack-mogan01:18
*** little has joined #openstack-mogan01:21
*** kevinz has joined #openstack-mogan01:30
*** zhangjl has joined #openstack-mogan01:34
openstackgerritMerged openstack/python-moganclient: Updated from global requirements
*** l4yerffeJ_ has joined #openstack-mogan01:53
*** l4yerffeJ has quit IRC01:53
*** liusheng has quit IRC01:54
*** yuntongjin has joined #openstack-mogan01:56
*** liusheng has joined #openstack-mogan02:12
*** yuntongjin has quit IRC02:36
*** yuntongjin has joined #openstack-mogan02:45
*** yuntongjin has quit IRC02:49
*** yuntongjin has joined #openstack-mogan03:16
*** Jeffrey4l_ has quit IRC03:21
*** Jeffrey4l has joined #openstack-mogan03:21
*** yuntongjin has quit IRC03:31
openstackgerritzhangjl proposed openstack/mogan-specs: Adds specification for lock instances
*** yuntongjin has joined #openstack-mogan04:15
*** wanghao has quit IRC05:58
*** wanghao has joined #openstack-mogan05:58
openstackgerritwanghao proposed openstack/mogan: [WIP]Quota support in Mogan(part 1)
openstackgerritwanghao proposed openstack/mogan: [WIP]Quota support in Mogan(part 1)
openstackgerritZhenguo Niu proposed openstack/mogan: Introduce a state machine for instance status
zhenguozhangjl: thanks for adding the first spec, hah06:53
zhangjlzhenguo:well, thanks06:53
zhenguohi all, if there's no objection for , I will land it, then we can add more states later07:18
zhenguoliusheng, shaohe_feng: are you around?07:20
liushengzhenguo: yes07:20
zhenguoliusheng: for state machine patch, I will land it by today if there's no objection from you guys07:21
liushengzhenguo: ok, I will take a review now07:21
zhenguoliusheng: thanks07:21
liushengzhenguo: that is really a big change. lol07:22
zhenguoliusheng: btw, does the ipmitool work now?07:22
liushengzhenguo: no :(07:22
zhenguoliusheng: hah, in fact just the concept of state machine is a big thing07:22
zhenguoliusheng: the logic is really simple, just check the orginal status before transition07:22
zhenguoliusehng: maybe something wrong with the fake ipmi I think07:23
liushengzhenguo: ok, le me check07:23
liushengzhenguo: yes, but I cannot find it07:23
zhenguoliusheng: will help to check whether it work07:24
liushengzhenguo: ipmitool -H -p 6231 -I lanplus -U admin -f /tmp/ -v sol activate07:24
zhenguoliusheng: It can must work in a real physical node07:24
liushengzhenguo: really ?07:24
zhenguoliusheng: yes, I used it many times with our blue env in HangZhou07:25
zhenguoliusheng: but now, I can't connect it :(07:25
liushengzhenguo: you mean the ironic node need to be a physical node ? or the devstack should be installed in a physical node ?07:25
zhenguoliusheng: I'm not sure the current devstack node, but for ssh_driver, I think it can work07:26
liushengzhenguo: currenly, we use pxe_ipmitool driver right ?07:28
zhenguoliusheng: for virtual machine, ipmitool should not work, but ironic have a sub project to help simulate that, but not sure if console can work07:29
liushengzhenguo: I have a devstack installed on a physical server, but someting wrong that it cannot create mogan server, if you mean we need a devstack installed on physical server, I am going to try to re-isntall the devstack07:29
zhenguoliusheng: but now, I can't even use ipmitool for powering, that's weird, should dig out why07:30
liushengzhenguo: haha07:30
zhenguoliusheng: not a devstack in physical server, but use mogan and ironic to manage real node07:30
zhenguoliusheng: instead of a virtual machine07:30
liushengzhenguo: I guess so07:31
liushengzhenguo: I can run the command like this: ipmitool -I lanplus -H -L ADMINISTRATOR -p 6232 -U admin -R 12 -N 5 -P password power status07:31
zhenguoliusheng: really?07:31
liushengzhenguo: yes07:31
zhenguoliusheng: you can change to use  ipmitool -I lanplus -H -L ADMINISTRATOR -p 6232 -U admin -R 12 -N 5 -P password -v sol activate07:32
liushengzhenguo: it cannot work07:32
liushengRunning Get PICMG Properties my_addr 0x20, transit 0, target 0x2007:33
liushengError Response 0xc1 from Get PICMG Properities07:33
liushengNo PICMG Extenstion discovered07:33
liushengError activating SOL payload: Invalid command07:33
liushengzhenguo: it return these errors07:33
zhenguoliusheng: oh, seems there's something wrong with the upstream work07:33
liushengzhenguo: I have googled this error, but cannot find useful info07:34
zhenguoliusheng: Will check code07:34
liushengzhenguo: do you think there should be a "-p 6231" in ipmitool comand include by the shellinaboxd process ?07:35
liushengzhenguo: currently the shellinaboxd process started by ironic don't include this, but I think it should include this07:36
*** yuntongjin has quit IRC07:36
zhenguoliusheng: seems yes, but it still not work with that in ipmitool command07:36
liushengzhenguo: the 6231, 6232, 6233 it 3 impi_port set to the 3 ironic nodes by ironic07:37
zhenguoliusheng: you don't need to  care about shellinabox now, just make sure ipmitool can work07:37
liushengzhenguo: yes07:37
liushengzhenguo: yes07:37
zhenguoliusheng: seems there is a virtualbmc log07:38
liushengzhenguo: where ?07:38
zhenguoliusheng: in ironic vm log dir07:38
liushengzhenguo: I found it, but it don't record log when I run ipmitool comands07:41
liushengzhenguo: may it only record the logs about operations by ironic07:41
zhenguoliusheng: is there any context there07:41
liushengzhenguo: yes, many lines of: 2017-01-04 23:23:07,070.070 51044 DEBUG VirtualBMC [-] Get power state called for domain node-107:42
liushengzhenguo: 2017-01-04 23:11:03,059.059 51200 INFO VirtualBMC [-] Virtual BMC for domain node-2 started07:42
zhenguoliusheng: you can use mogan to power a node to see whether there's some thing recorded07:43
zhenguoliusheng: and use ipmitool directly to see07:43
liushengzhenguo: ok07:43
liushengzhenguo: yes, it recorded: 2017-01-16 15:43:39,895.895 51044 DEBUG VirtualBMC [-] Power off called for domain node-1  when I stop a server07:45
zhenguoliusheng: with mogan or ipmitool07:45
liushengzhenguo: with mogan07:46
zhenguoliusheng: but there's nothing with ipmitool?07:46
liushengzhenguo: how to run directly with ipmitool ?07:47
liushengzhenguo:  ipmitool -I lanplus -H -L ADMINISTRATOR -p 6232 -U admin -R 12 -N 5 -P password power start ?07:47
zhenguoliusheng: yes07:47
liushengzhenguo: ok let me try07:47
liushengzhenguo: 2017-01-16 15:47:29,398.398 51044 DEBUG VirtualBMC [-] Power on called for domain node-107:48
liushengzhenguo: it can also work and record this log07:48
zhenguoliusheng: so for console command?07:48
liushengzhenguo: it cannot work07:50
zhenguoliusheng: nothing recorded?07:50
liushengzhenguo: yes07:50
zhenguoliusheng: maybe virtualbmc doesn't support this yet, will confirm07:51
liushengzhenguo: yes, maybe07:51
zhenguoliusheng: seems it can only support power on/off, get/set boot device,07:57
zhenguoliusheng: so for console testing, I would suggest you use agent_ssh driver07:59
liushengzhenguo: oh, thanks for your info, I will try07:59
zhenguoliusheng: you need to restack with IRONIC_DEPLOY_DRIVER=agent_ssh in you local config file08:00
liushengzhenguo: don't need any other options ?08:01
zhenguoliusheng´╝Ü yes08:01
zhenguoliusehng: wait08:02
openstackgerritwanghao proposed openstack/mogan: [WIP]Quota support in Mogan(part 1)
openstackgerritzhangjl proposed openstack/mogan-specs: Adds specification for lock instances
zhenguoliusheng: seems only that option can work08:10
liushengzhenguo: ok, let me try, thanks for you help :)08:11
zhenguoliusheng: np, feel free to ping me if there's still something wrong with that08:11
liushengzhenguo: hah, thanks, you are a ironic export!08:11
zhenguoliusheng: lol, just because the ssh console support is added by me08:14
liushengzhenguo: hah, awesome!08:14
zhenguoliusheng: so if there's still something wrong, we can hack ti08:16
liushengzhenguo: ok08:17
zhenguoliusheng: ok, good luck08:17
liushengzhenguo: lol08:17
*** wanghao_ has joined #openstack-mogan08:47
*** wanghao has quit IRC08:47
littlehi guys, i encountered a issue about multi-instance support: should we get a flow per instance or create all instances in a flow?08:50
zhenguolittle: how about nova?08:51
littlenova does not use task flow08:51
zhenguolittle: but if in separate flow, it may takes long time08:51
littleso it will process it much easier08:52
zhenguolittle: I think we should do every create it different flow, but they should be in parallel, wdyt?08:53
littleso you mean get a flow per instance?08:54
zhenguowanghao_: does cinder support creating multi volume?08:55
zhenguolittle: yes, maybe we should follow cinder if it supports creating multi volume08:55
zhenguolittle: let me check first, but yes I think we should create aflow for every instance08:56
wanghao_zhenguo: no, you just can create one volume in API08:56
zhenguowanghao_: oh08:56
wanghao_zhenguo:  Do we need to support multi creation,  or  just leave it to mistral.08:57
littlezhenguo: so i think one flow per instnce is good08:57
zhenguowanghao_: not sure, but nova supports that08:57
zhenguolittle: thanks08:57
wanghao_zhenguo: yes, from nova side,  it should be since multi instance with same flavor is making sense to users.08:58
zhenguowanghao_: yes, and for us, multi instance with same flavor for baremetal seems also makes sense, not sure, in fact, you are customer, lol08:59
wanghao_zhenguo: I think it should be.  many baremetals may be needed to deploy in one time.09:00
zhenguowanghao_: yes, thanks09:00
wanghao_zhenguo: so support multi instances in one taskflow, that will be nice.09:01
zhenguowanghao_: why not multi flow?09:01
littleIf in one flow, the revert process is much complex09:02
wanghao_zhenguo: isn't that be a long time with multi flow?09:02
zhenguowanghao_: every flow includes tasks like scheduler, reschedule, set_instance_info, deploy_instance09:02
zhenguowanghao: we can create multi flows, they will run in parallel,09:02
wanghao_zhenguo: little: I see, that's will be great.09:03
wanghao_parallel way.09:03
littleindependent process for every instance09:04
zhenguowanghao, little: yes, that will be more clear09:04
wanghao_zhenguo: oh,  btw,   I think a useful feature that do we want to mange a baremetal?09:05
wanghao_zhenguo: I mean, OS has be installed, everything is ok, and mogan can mange it to itself.09:06
zhenguowanghao_: you mean adopt an existing baremetal node?09:06
wanghao_zhenguo: yes09:06
zhenguoliusheng: yes, we can do that, and ironic already support it09:06
wanghao_zhenguo: cool.09:06
zhenguowanghao_: maybe need to add before O release09:07
zhenguowanghao_: that's a good example for mogan is better than nova in baremetal management, lol09:08
wanghao_zhenguo:  agree,  Nova doesn't to mange an existing instance.09:08
zhenguowanghao_: yes,09:10
* zhenguo brb09:21
*** wanghao_ has quit IRC09:24
openstackgerritliusheng proposed openstack/mogan: Remove the unused notifier and engine api initialization
*** kevinz has quit IRC10:06
*** little has quit IRC10:08
*** zhangjl has quit IRC10:29
openstackgerritZhenguo Niu proposed openstack/mogan: Introduce a state machine for instance status
zhenguoliusheng: is your ssh driver env ready? I remember we use agent_ssh on our tempest gate11:16
liushengzhenguo: no, I need to restack again11:19
liushengzhenguo: will try after my env get ready11:19
zhenguoliusheng: ok11:23
openstackgerritliusheng proposed openstack/mogan: Remove the unused notifier and engine api initialization
liushengzhenguo: I have finished my env installation11:43
zhenguoliusheng: is that work?11:44
liushengzhenguo: yes, can create server11:44
zhenguoliusheng: hah, you can try to enable console11:44
liushengzhenguo: ok, the same steps ?11:45
zhenguoliusheng: yes11:45
zhenguoliusheng: hi, liudong proposed this bp , wdyt?11:51
liushengzhenguo: I think we need to consider the posssibilities various drivers11:52
liushengzhenguo: but seems we only consider ironic, right ?11:54
liushengzhenguo: I have tried the console11:54
liushengzhenguo: seems the shellinabox still started with wrong paremeter11:55
zhenguoliusheng: yes, we should consider to support more drivers like nova, because there are many baremetal provisioning project11:56
liushengzhenguo: yes11:56
zhenguoliusheng: currently we use ironic is just like a driver, that's all, so maybe it's easy to split it out11:57
liushengzhenguo: so we may need to keep the concept in mind now, before we starting to change our architecture11:57
zhenguoliusheng: yes, liudong said he like to do this, hah11:57
zhenguoliusheng: they don't use ironic, but want to use mogan11:58
liushengzhenguo: hah11:58
liushengzhenguo: that make sense11:58
zhenguoliusheng: yes11:58
liushengzhenguo: about the console, the ssh driver use command like "virsh console node-1"11:58
zhenguoliusheng: yes11:58
liushengzhenguo: but I cannot directly use that command to login the node11:58
zhenguoliusheng: oh11:58
liushengConnected to domain node-111:59
liushengEscape character is ^]11:59
liushengerror: internal error: character device <null> is not using a PTY11:59
liushengzhenguo: it return this11:59
zhenguoliusheng: so seems the devstack plugin is not support this11:59
zhenguoliusheng: but we can change the vm xml file to make it suport PTY12:00
liushengzhenguo: but seems libvirt support this12:00
zhenguoliusheng: yes, but you need to add PTY console in the xml file12:00
liushengzhenguo: i have googled this, there is a suggestion way, but it still cannot work, the error dissapared, but it will be stucked12:01
liushengzhenguo: virsh console --devname serial1 node-112:01
liushengzhenguo: because the PTY in the xml is "serial1" port12:01
liushengzhenguo: it will be stucked, and only show: Escape character is ^]12:02
zhenguoliusheng: I tried with my env, seems yes12:04
zhenguoliusheng: you can try to access shellinabox console url12:06
zhenguoliusheng: via browser12:06
liushengzhenguo: it is same with "virsh console node-1"12:07
zhenguoliusheng: you mean on the it shows the same context in the page?12:07
liushengzhenguo: yes12:07
zhenguoliusheng: you can try to remove other serial and console in the xml12:08
zhenguoliusheng: just leaving pty12:08
zhenguoand serial012:08
zhenguoliusheng: like this
liushengzhenguo: let me try12:11
liushengzhenguo: do you know where the xml file stored in ?12:11
liushengzhenguo: I want to backup12:11
zhenguoliusheng: not sure, but no need to backup it as you can copy other node's xml if needed, lol12:12
liushengzhenguo: lol12:12
zhenguoliusheng: oh, just find that, there no pty console in the xml file ony a serial dev12:14
liusheng    <serial type='pty'>12:14
liusheng      <source path='/dev/pts/51'/>12:14
liusheng      <target port='1'/>12:14
liusheng      <alias name='serial1'/>12:14
liusheng    </serial>12:14
liushengzhenguo: isn't this ?12:15
zhenguoliusheng: yes, this just a serial dev, you also need to add a console dev with pty12:15
liushengzhenguo: oh, how to add that ?12:15
zhenguoliusheng: see
liushengzhenguo: oh, I found it12:16
liushengzhenguo: thanks12:16
liushengzhenguo: I just found there is a console tag that is the console log12:17
zhenguoliusheng: yes, you can change it directly to pty and change the serial port num12:18
*** kevinz has joined #openstack-mogan12:20
liushengzhenguo: when I save the xml, it says: error: unsupported configuration: Only the first console can be a serial port12:20
*** kevinz has quit IRC12:21
zhenguoliusheng: can you paste your config here?12:21
liushengzhenguo: oh may I need to delete the console log device12:22
zhenguoliusheng: just delete it to make things simple12:22
liushengzhenguo: it is weird, libvirt can reset it after I editted12:27
liushengzhenguo: need to stop firstly ?12:27
zhenguoliusheng: seems need to restart libvirt12:27
zhenguoliusheng: does that work?12:35
liushengzhenguo: I still cannot edit the xml12:35
zhenguoliusheng: even if restart libvirtd?12:35
liushengzhenguo: it will automatically reset after aditted12:35
liushengzhenguo: yes12:35
zhenguoliusheng: need to as some libvirt guy, hah12:36
liushengzhenguo: you mean 1. stop libvirtd 2. edit 3. start libvirtd ?12:36
zhenguoliusheng: you can google it12:36
liushengzhenguo: let me randomly try, lol12:36
zhenguoliusheng: a libvirt maintainer is here, but he is absent now12:37
liushengzhenguo: how ?12:37
liushengzhenguo: who12:37
zhenguoliusheng: renguannan12:37
liushengzhenguo: oh, maybe some periodic task in ironic do this ?12:38
zhenguoliusheng: no,12:38
zhenguoliusheng: I can't remember how to do that, but seems need extra steps12:38
liushengzhenguo: let me ask google12:39
zhenguoliusheng: hah, good choice12:39
zhenguoliusheng: I have to go home now, ttyl?12:39
liushengzhenguo: ok12:39
liushengzhenguo: bye12:39
zhenguoliusheng: bye12:43
*** l4yerffeJ_ has quit IRC13:19
*** l4yerffeJ_ has joined #openstack-mogan13:19
*** little has joined #openstack-mogan13:25
*** little has quit IRC13:41
*** l4yerffeJ_ has quit IRC13:58
*** l4yerffeJ_ has joined #openstack-mogan13:59
*** l4yerffeJ_ has quit IRC14:00
*** l4yerffeJ_ has joined #openstack-mogan14:00
openstackgerritZhenguo Niu proposed openstack/mogan: Remove the unused notifier initialization
openstackgerritZhenguo Niu proposed openstack/mogan: [WIP] Save power_state to instance and sync with ironic
*** l4yerffeJ__ has joined #openstack-mogan15:12
*** l4yerffeJ_ has quit IRC15:12
*** l4yerffeJ__ has quit IRC15:14
*** l4yerffeJ__ has joined #openstack-mogan15:14
*** l4yerffeJ__ has quit IRC15:35
*** l4yerffeJ__ has joined #openstack-mogan15:36
*** l4yerffeJ__ has quit IRC15:37
*** l4yerffeJ__ has joined #openstack-mogan15:37
openstackgerritMerged openstack/mogan: Remove the unused notifier initialization
openstackgerritMerged openstack/mogan: Introduce a state machine for instance status
*** l4yerffeJ has joined #openstack-mogan21:34
*** l4yerffeJ__ has quit IRC21:34
*** Jeffrey4l has quit IRC21:34
*** Jeffrey4l has joined #openstack-mogan21:35

Generated by 2.14.0 by Marius Gedminas - find it at!