Wednesday, 2017-07-12

*** wanghao has joined #openstack-mogan00:16
zhenguomorning mogan!00:54
zhenguoliusheng: cool00:56
*** liujiong has joined #openstack-mogan00:56
liushengzhenguo: lol00:56
zhenguoliusheng: gate broken again?00:57
liushengzhenguo: not sure00:58
liushengzhenguo: seems devstack installed failed00:58
zhenguoliusheng: yes00:59
liushengzhenguo: this situation occured sometimes00:59
liushengzhenguo: but don't know shy00:59
liushengzhenguo: why00:59
zhenguoliusheng: seems not related to us01:00
*** openstackgerrit has joined #openstack-mogan01:02
openstackgerritliusheng proposed openstack/mogan master: Support running api under uWSGI
wanghaohi guys,  patch for managing existing bm is ready, welcome to review,  part-one:
zhenguowanghao: awesome, thanks for the hard work!01:21
wanghaozhenguo: you're welcom01:23
*** wanghao_ has joined #openstack-mogan01:25
*** wanghao has quit IRC01:25
openstackgerritZhenguo Niu proposed openstack/mogan master: Fix circular reference error when attaching interface
*** wanghao has joined #openstack-mogan01:56
Xinranzhenguo, liusheng  sorry for the delay. I was not in company yestoday. and thanks for you guy's effort01:56
Xinranzhenguo, liusheng very appreciate01:56
liushengXinran: you're welcome :)01:56
*** wanghao_ has quit IRC01:58
zhenguoliusheng: np02:03
zhenguowrong person02:03
zhenguoliusheng, Xinran: recheck still got failed02:06
zhenguoliusheng: but another error02:06
zhenguoliusheng: let me recheck again02:06
zhenguoliusheng: how's the placement patch02:09
liushengzhenguo: modifying02:10
zhenguoliusheng: ok, let's try to land it by today02:11
openstackgerritliusheng proposed openstack/mogan master: Clean the compute_node and compute_port objects and db interfaces
openstackgerritliusheng proposed openstack/mogan master: Get rid of listing availability zone api and clean some legacy code
openstackgerritliusheng proposed openstack/mogan master: Clean the methods about updating node resources to Mogan db
openstackgerritliusheng proposed openstack/mogan master: Get rid of node listing api of Mogan
openstackgerritliusheng proposed openstack/mogan master: Refactor the scheduler to use placement service
zhenguoluyao, Xinran: on my tests, there are still some problems with attaching interfaces, like when ironic attach vif failed, the vif and neutron port info will still associated the server, you can test with some corner cases.02:15
zhenguoluyao, Xinran: and the new devstack will provide two ports for each ironic node, so you can test attach without detaching the port.02:15
Xinranzhenguo,  I will check it02:19
zhenguoXinran: thanks02:19
*** litao__ has joined #openstack-mogan02:22
openstackgerritliusheng proposed openstack/mogan master: Support running api under uWSGI
luyaozhenguo, ok, so just clean the info if  attach vif failed?02:23
zhenguoluyao: not sure, maybe better to do some cleanup02:24
zhenguoluyao: you finished your thesis?02:24
luyaozhenguo, not yet, do some improvement now :)02:25
zhenguoluyao: hah02:26
luyaozhenguo, I will fix this bug asap02:27
zhenguoluyao: thanks, and found this bug
openstackLaunchpad bug 1703592 in Mogan "Circular reference error when attaching interface" [Undecided,In progress] - Assigned to Zhenguo Niu (niu-zglinux)02:28
luyaozhenguo, you have fixed this bug?02:32
zhenguoluyao: I think so02:32
luyaozhenguo, hah : )02:32
zhenguoluyao: need you to confirm with it, hah02:33
luyaozhenguo, OK, : D02:33
openstackgerritZhenguo Niu proposed openstack/mogan-specs master: Node aggregates
*** wanghao_ has joined #openstack-mogan02:43
*** wanghao has quit IRC02:43
*** wanghao has joined #openstack-mogan02:54
*** wanghao__ has joined #openstack-mogan02:55
*** wanghao has quit IRC02:55
*** wanghao_ has quit IRC02:58
zhenguoliusheng: the detaching interface patch seems still got failed with test_attach_detach_interface03:08
*** wanghao has joined #openstack-mogan03:10
zhenguoliusheng, shaohe_feng: please help to +1 this
*** wanghao__ has quit IRC03:14
openstackgerritZhenguo Niu proposed openstack/mogan master: Clean up unneeded server=True in model_query
openstackgerritZhenguo Niu proposed openstack/mogan master: Clean up unneeded server=True in model_query
openstackgerritZhenguo Niu proposed openstack/mogan master: Adds aggregates DB model and API
shaohe_fengzhenguo: OK03:47
openstackgerritZhenguo Niu proposed openstack/mogan master: Trivial-Fix: Correct KeypairExists exception message
openstackgerritXinran WANG proposed openstack/mogan master: Add filters to server list API
openstackgerritwanghao proposed openstack/mogan master: Refactor exception raise by using mogan exception
*** liujiong has quit IRC06:19
*** liujiong has joined #openstack-mogan06:19
*** openstackgerrit has quit IRC06:48
*** liujiong_lj has joined #openstack-mogan06:50
-openstackstatus- NOTICE: FYI, is down, review.o.o is slow - please have patience until this is fixed06:52
*** liujiong has quit IRC06:54
*** openstackgerrit has joined #openstack-mogan07:11
openstackgerritliusheng proposed openstack/python-moganclient master: Remove log translations
openstackgerritliusheng proposed openstack/python-moganclient master: Remove log translations
openstackgerritZhenguo Niu proposed openstack/mogan master: Adds aggregates DB model and API
openstackgerritZhenguo Niu proposed openstack/mogan-specs master: Node aggregates
-openstackstatus- NOTICE: Gerrit is going to be restarted, due to low performance07:49
zhenguohi all, please help to review the placement patch
-openstackstatus- NOTICE: Gerrit has been successfully restarted07:56
zhenguoliusheng: hi, do you have a plan to add socat console support?07:57
liushengzhenguo: not yet, I am not very fimilar with it :(07:58
zhenguoliusheng: which will also make it possible for console logs support for us07:58
zhenguoliusheng: haha, just an ironic api to get the console url like shellinabox07:58
zhenguoliusheng: the logic is done by ironic07:59
liushengzhenguo: will learn it07:59
zhenguoliusheng: need to use physical env for testing08:00
liushengzhenguo: our physical env need to be restack08:00
zhenguoliusheng: yes08:00
zhenguoliusheng: no rush, just want to introduce it before Pike08:00
zhenguoliusheng: but I think it's easy08:01
liushengzhenguo: hah, will try08:01
zhenguoliusheng: thanks08:01
zhenguoliusheng: will this break our gate ?08:01
liushengzhenguo: you can help to take a look at the uwsgi support patch if you get time08:01
zhenguoliusheng: sure08:02
zhenguoliusheng: but after the placement one08:02
zhenguoliusheng: i'm testing it08:02
liushengzhenguo: actually, I am not very sure, just believe you, hah08:02
*** liujiong_lj is now known as liujiong08:02
zhenguoliusheng: I think our devstack plugin will override the resource_class08:03
zhenguoliusheng: which flavor you are using for tempest test08:03
liushengzhenguo: ok, thanks, but I assume it need a matched client side patch of the uwsgi support patch08:03
liushengzhenguo: the "small"08:03
zhenguoliusheng: I will get rid of the small flavor08:03
liushengzhenguo: yes, I have a TODO in the tempest, that we'd better to get rid of the "small" flavor in tempest08:04
zhenguoliusheng: with this
zhenguoliusheng: we will got a flavor like baremtal_16cpu_1024ram_1024disk08:05
zhenguoliusheng: as ironic will set the resource class itsself like this08:05
liushengzhenguo: but the flavor name can be simple, right ?08:05
liushengzhenguo: we just need to match the resource class08:05
zhenguoliusheng: yes, but 'small' seems not that specific08:06
zhenguoliusheng: I like the pattern what ironic used, hah08:06
liushengzhenguo: yes, I don't mean "small" haha08:06
zhenguoliusheng: so, you will get the 'small' on tempest?08:06
zhenguoliusheng: or just using what devstack created08:07
liushengzhenguo: may you can try to clean something08:08
liushengzhenguo: sorry, ignore above words08:08
zhenguoliusheng: will try to test the patch with new project-config08:08
zhenguoliusheng: btw, the detaching patch seems still got failed08:08
liushengzhenguo: yes, it is easy to create a specific flavor for tempest only08:08
liushengzhenguo: yes, I have no idea about our unstable gate08:09
liushengzhenguo: :(08:09
zhenguoliusheng: lol08:09
zhenguoliusheng: seems match a recheck bug, let me recheck it08:09
liushengzhenguo: need to dig and improve it08:09
zhenguoliusheng: seems most are not related ot us, right?08:10
liushengzhenguo: yes, only one situation that the console tempest fail sometimes08:11
zhenguoliusheng: ironic is now removing the ssh driver08:12
liushengzhenguo: how it effect us08:12
zhenguoliusheng: how to we test console08:13
zhenguoliusheng: seems fake ipmidriver can't support console now08:13
liushengzhenguo: I remember it can only work in physical environment, right ?08:14
zhenguoliusheng: you may need to update the doc about how to test console with ssh driver enabled08:14
zhenguoliusheng: yes08:14
zhenguoliusheng: but ssh driver can provide shellinabox console08:14
zhenguoliusheng: if ssh driver got removed, I think we should make socat as default console, wdyt? and we need to add console log support08:15
liushengzhenguo: not sure how Nova test it, if there any console tempest based on Nova+Ironic ?08:15
zhenguoliusheng: seems not,  I think the tempest is with very low coverage08:15
liushengzhenguo: yes08:15
zhenguoliusheng: socat console need websocket proxy, we can just copy from nova08:16
liushengzhenguo: so the socat console can avoid these problems ?08:16
zhenguoliusheng: no08:16
zhenguoliusheng: we can't test console without physical env...08:17
zhenguoliusheng: both for shellinabox and socat08:17
liushengzhenguo: hmm...08:17
zhenguoliusheng: as we only left a virtualbmc for vm which can't provide console08:17
* liusheng seems Mogan develop is expensive..08:18
zhenguoliusheng: hah08:19
zhenguoliusheng: maybe virtualbmc can add that support in future08:19
liushengzhenguo: it has been implemented in Ironic ?08:20
zhenguoliusheng: a sub project of ironic08:20
liushengzhenguo: oh, need to learn more08:20
zhenguoliusheng: no need to learn, just do it, hah08:20
zhenguoliusheng, shaohe_feng: I added some trivalfix patches, please help to review, which are simple08:28
zhenguoliusheng: for the keypair patch, I want' to correct the wrong 'KeyPaire' work in error message08:37
liushengzhenguo: yes, just replied you08:37
zhenguolusheng: and seems nova itself it's not consistent for the exception name08:38
zhenguoliusheng: I find all exceptions for keypair in nova use Keypair other than that one08:38
liushengzhenguo: hah08:38
liushengzhenguo: so maybe you can just fix typo or change all them to consistent08:39
zhenguoliusheng: on exception side, with the patch, it got consistency, lol08:40
liushengzhenguo: ... alright08:41
zhenguoliusheng: hah08:41
liushengzhenguo: we may need to confirm if gate got broken after project-config changes08:41
zhenguoliusheng: yes, I just recheked the patch08:42
zhenguoliusheng: but seems need to wait for some time to make project-config take effect08:42
liushengzhenguo: doesn't it effect immediately ?08:43
zhenguoliusheng: no08:43
zhenguoliusheng: maybe half hour?08:43
liushengzhenguo: oh08:43
*** liujiong has quit IRC08:47
*** liujiong has joined #openstack-mogan08:47
openstackgerritliusheng proposed openstack/python-moganclient master: Fix the gate broken by osc-lib updates
zhenguoliusheng: see my comments for the placement patch,09:03
zhenguoliusheng: if we just delete allocations, the rp can be scheduled again, but the ironic node is in deleting progress, which will cause error.09:04
liushengzhenguo: I have replied you in previous patch09:04
liushengzhenguo: I agree we'd better to improve that, but I think it should be done in another patch to add the waiting mechanism in server deletion09:04
liushengzhenguo: wdyt ?09:05
zhenguoliusheng: how to wait?09:05
liushengzhenguo: add a time to periodically query the node state09:05
zhenguoliusheng: as I understand the simple way is to get rid of the rp, and wait it come back when the node is available again09:05
zhenguoliusheng: why we need that09:05
liushengzhenguo: why need to get rid of rp ? I think we should release the rp after deleting09:06
zhenguoliusheng: but only if the ironic node got available the rp should be release09:07
liushengzhenguo: we need to add a timer to wait the state of node to become "available" and then delete the allocation to release the rp09:08
zhenguoliusheng: a periodic task wait for that costs too much,09:08
liushengzhenguo: no09:08
liushengzhenguo: cost what ?09:08
zhenguoliusheng: it's a loop09:08
zhenguoliusheng: and for physical env it may take many hours to be done as the disk erase09:09
liushengzhenguo: even though you also need to delete the allocation09:10
zhenguoliusheng: yes, I  propose to also delete the rp09:10
liushengzhenguo: the resources updating cannot delete the allocation09:10
liushengzhenguo: ...09:10
zhenguoliusheng: yes, delete rp and allocations09:11
liushengzhenguo: that looks weird..09:12
zhenguoliusheng: why09:12
zhenguoliusheng: but I can't find a way to handle it gracefully09:12
liushengzhenguo: the allocation in placement was degined to handle the resource consumation09:12
zhenguoliusheng: I really take many hours to delete a node in our productions09:12
zhenguoliusheng: yes its is09:13
liushengzhenguo: similar situation, the deployment may also take many hours, a waiting task I don't think cost much resourcs09:14
zhenguoliusheng: why delpyment will take that long09:14
shaohe_fengzhenguo: OK09:14
zhenguoliusheng: it will be less than 20 mins09:14
*** liujiong has quit IRC09:15
*** liujiong has joined #openstack-mogan09:15
zhenguoliusheng: deleting node will erase all disks, which may take a day or long to finish09:15
zhenguoliusheng: even if you don't delete the rp when deleting server ,it will get deleted when update resources, right?09:17
liushengzhenguo: hmm.. another side, the resource updating won't only update the nodes in "available" state09:17
zhenguoliusheng: sure, you will delete rp if it's not 'available'09:18
zhenguoliusheng: but no handle allocations there09:18
liushengzhenguo: no, the current implemenation won't delete the not available nodes09:19
liushengzhenguo: it only delete the abnormal state nodes09:20
zhenguoliusheng: which is abnormal state09:20
liushengzhenguo: that I think is the designe intention of placement09:21
liushengzhenguo: e.g. state of a node is available, but instance_uuid of the node is not None09:21
liushengzhenguo: or nodes in error state09:22
zhenguoliusheng: seems yes09:22
liushengzhenguo: the rps or inventories in placement don't represent only the resources can be consumed09:23
zhenguoliusheng: what others it for?09:23
liushengzhenguo: the also includes the resources have been consumed09:24
zhenguoliusheng: yes, I know,09:24
zhenguoliusheng: so we can provide node list for aggregates with that09:24
zhenguoliusheng: but wait for status is too long, that's why nova don't wait that09:26
liushengzhenguo: hmm, if so how aobut improve the resource updating to also update allocations09:28
zhenguoliusheng: you mean check the node status?09:28
liushengzhenguo: yes, and also update the allocations in placement09:29
zhenguoliusheng: if the node is avaialble, then delete the allocations09:29
liushengzhenguo: if state of node is available, need to check if there is allocation items assinged to this node09:29
zhenguoliusheng: seems a good choice09:29
liushengzhenguo: yes,09:29
zhenguoliusheng: please go ahead, hah09:30
liushengzhenguo: ok, will try09:30
zhenguoliusheng: thanks09:30
* zhenguo brb09:31
*** wanghao has quit IRC09:32
liushengzhenguo: please take a look at if you get time09:33
zhenguoliusheng: ok09:42
zhenguoliusheng: when deleting server, should we delete the allocations if the rp is gone?09:42
zhenguoliusheng: seems the project-config patch really break the gate, hah09:44
liushengzhenguo: :(09:44
liushengzhenguo: placement didn't provide an api to delete the allocations of a provider :(09:45
liushengzhenguo: only provide list the allocations of a provider09:45
zhenguoliusheng: hah09:45
liushengzhenguo: I guess it will provide that api09:45
zhenguoliusheng: if so, let's wait for that happen09:45
openstackgerritMerged openstack/python-moganclient master: Fix the gate broken by osc-lib updates
*** liujiong has quit IRC10:23
openstackgerritMerged openstack/mogan master: Trivial-Fix: Correct KeypairExists exception message
openstackgerritMerged openstack/mogan master: Clean up unneeded server=True in model_query
openstackgerritMerged openstack/mogan master: Fix circular reference error when attaching interface
zhenguoliusheng: seems the gate isn't broken11:21
zhenguoliusheng: need your +A for this
openstackgerritZhenguo Niu proposed openstack/mogan master: Use IRONIC_USE_RESOURCE_CLASSES in devstack plugin
* zhenguo away11:54
*** litao__ has quit IRC11:56
*** zhenguo_ has joined #openstack-mogan12:40
*** zhenguo has quit IRC12:41
*** zhenguo_ is now known as zhenguo12:41
*** liusheng has quit IRC15:42
*** liusheng has joined #openstack-mogan15:43
*** zhuli_ has joined #openstack-mogan16:00
*** zhuli_ has quit IRC18:40
*** harlowja has quit IRC19:17
*** harlowja has joined #openstack-mogan20:02
*** wanghao has joined #openstack-mogan23:54
*** zhuli_ has joined #openstack-mogan23:55

Generated by 2.15.3 by Marius Gedminas - find it at!