Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-03-18
19:30:20 sean-k-mooney noonedeadpunk: most flavor fileds are imuntable after its created
19:31:02 sean-k-mooney the update endpoint only supports updating the description
19:31:04 noonedeadpunk I can understand about flavor properties, as it will mix up isntances
19:31:27 sean-k-mooney https://docs.openstack.org/api-ref/compute/?expanded=update-flavor-description-detail#update-flavor-description
19:31:45 noonedeadpunk but is_public is kinda visability thing, which should be safe to update in that terms
19:32:03 sean-k-mooney in theory yes.
19:32:38 sean-k-mooney although updating it wont change the status of the is_public flag in the embded flavor in an instnace
19:32:42 gmann but it might cause restriction access to few tenants which are not in access list.
19:33:31 openstackgerrit melanie witt proposed openstack/nova master: DNM: try to get some debug info for bug 1844929 https://review.opendev.org/701478
19:33:31 openstack bug 1844929 in OpenStack Compute (nova) "grenade jobs failing due to "Timed out waiting for response from cell" in scheduler" [High,Confirmed] https://launchpad.net/bugs/1844929
19:34:03 sean-k-mooney gmann: well this is not using the os-flavor-acces api
19:34:17 sean-k-mooney this is setting the is_public flag in the flavor itself
19:34:34 sean-k-mooney the os-flavor-acces api is normally what peopel shoudl use
19:34:53 sean-k-mooney so to make a flavor private you just add acess to only the admin/service project
19:35:00 sean-k-mooney then no other tenats can see it
19:35:01 gmann yeah but we restrict the public flavor to process in flavor-access api
19:35:30 gmann yeah, removing all other tenants make it private
19:36:14 sean-k-mooney noonedeadpunk: does ^ makes sense
19:37:00 sean-k-mooney noonedeadpunk: is_public on the flavor is a hold over from before we had per tenant contol over flaovr visablity
19:37:06 noonedeadpunk hm, let me check that
19:38:02 noonedeadpunk (actully deploy latest sandbox and test)
19:38:40 sean-k-mooney lyarwood: looking at https://github.com/openstack/devstack-plugin-ceph/blob/master/devstack/lib/ceph#L1213-L1251 it looks like ceph is not in the devstack slice
19:38:50 sean-k-mooney so stop devstack@* should be ok
19:39:34 sean-k-mooney you would still have to do "killall qemu" or use virish stop on all active domains too to stop the rbd device in use error
19:39:37 lyarwood sean-k-mooney: right and I don't think we deploy anything on the subnode in this case anyway
19:39:46 gmann only thing is you would not be able to add projects or list access info on public flavor
19:40:23 sean-k-mooney gmann: you can use the flavor acess api to do that
19:40:40 sean-k-mooney horizon allows you to take a standard public flavor and add tenants too it
19:41:08 gmann sean-k-mooney: no, after 2.7 add project is restricted on non-public and list access is for non-public only from stating
19:41:19 sean-k-mooney gmann: i tought the os-flaovr-acess api basically ignored the is_public atribte
19:41:29 gmann https://github.com/openstack/nova/blob/c9f5b583b6072f542d1757e35fd6305b9698496a/nova/api/openstack/compute/flavor_access.py#L50-L75
19:41:44 sean-k-mooney gmann: right so if you use an older microversion then what happens
19:42:13 gmann with older than 2.7 yes you can add.
19:42:39 sean-k-mooney i suspect horizon is using older then 2.7
19:42:52 sean-k-mooney why did we make that change?
19:43:27 gmann well, public means for everyone so we do not really need to add access things right
19:44:05 sean-k-mooney gmann: basicaly i tought we effectivly deprecated is_public and ignored it
19:44:21 sean-k-mooney gmann: so if you use falvor acess it effectivly became private
19:44:32 sean-k-mooney excpet to the tenants listed
19:46:00 sean-k-mooney gmann: if that is not the case then i agree that this should be mutable in the flavor api
19:46:34 sean-k-mooney you cant really gracefully retire a flavor othersize.
19:47:22 sean-k-mooney it looks like it was change as a result of this bug https://bugs.launchpad.net/nova/+bug/1361476
19:47:22 openstack Launchpad bug 1361476 in OpenStack Compute (nova) "flavor access create should check public/private first" [Low,Fix released] - Assigned to Sergey Nikitin (snikitin)
19:47:51 gmann humm, but will change in is_public on embedded instances effect like make that flavor stale. the reason we do not allow flavor modification on other parameters.
19:48:14 sean-k-mooney my expectation is that if a flaovr is public and i add a tenatn via os-flavor-access then any tenant other then those listed via flavor access would not see it
19:48:48 sean-k-mooney gmann: is public on the embeded instance has no meaning
19:49:03 gmann and keep is_public same ?
19:49:03 sean-k-mooney so i dont think its an issue
19:49:13 gmann listing flavor?
19:49:17 sean-k-mooney yes
19:49:30 sean-k-mooney so my expectation is is_public would still be public/true
19:49:35 sean-k-mooney and flavor list would not list
19:49:37 sean-k-mooney it
19:49:53 sean-k-mooney unless you were in the tenant on the access list or an admin
19:50:30 sean-k-mooney so basicaly i think https://bugs.launchpad.net/nova/+bug/1361476 was invalid
19:50:30 openstack Launchpad bug 1361476 in OpenStack Compute (nova) "flavor access create should check public/private first" [Low,Fix released] - Assigned to Sergey Nikitin (snikitin)
19:51:05 gmann i mean we have is_public in list flavor filters if anyone listing by their instance's flavor is_public flag
19:51:09 sean-k-mooney well actully maybe not
19:51:49 sean-k-mooney gmann: can you say that again
19:52:45 sean-k-mooney looking at the bug i would expect "nova flavor-access-list --flavor 1" to work but i would not expect the flavor to be in "openstack flavor list" unless you were in a tenant with acess or an admin
19:52:47 gmann i mean will is_public change the flavor signature or not? even that does not make any change in real configueation but still a attribute in flavor user facing dict
19:53:27 sean-k-mooney is_public only changes if you can see the flaovr or not
19:53:48 sean-k-mooney so i think that is just metadata about the flaovr
19:53:51 sean-k-mooney not a part of it
19:54:04 gmann yeah. kind of.
19:54:14 sean-k-mooney kind of like the description wich we allow to be udpated
19:54:15 gmann or we just remove this flag.
19:54:49 sean-k-mooney i think the only benift to it is the defualt polciy
19:55:04 sean-k-mooney e.g. should it be visable by default or not
19:55:26 sean-k-mooney but i certenly dont think we should be blocking the use of the flaovr acess api based on it
19:56:00 sean-k-mooney i would be ok with removing it too but im not sure how others would feel
19:56:35 sean-k-mooney e.g. do operators use it frequently. if soo i think it should be kept and mutable. if not remove and just use flavor acess api for this
19:56:42 gmann ok, even list public flavor on list access is not wrong which is why that bug did the change in add access
19:57:18 gmann sorry, need to go for lunch, ttyl
19:57:27 sean-k-mooney no worries
19:57:36 sean-k-mooney im going for food too
19:59:06 noonedeadpunk it was pretty interesting discussion. Like I thought it is a bit more simple than it is
19:59:43 sean-k-mooney well the code is simple to change but it has several other implications obvirously
20:00:09 sean-k-mooney noonedeadpunk: this is why we generally require a spec for api changes as this type of discussion normally happens when we dig into it
20:00:57 noonedeadpunk yeah and it is fair
20:01:47 noonedeadpunk will try to write down it as it seems that some patching is required anyway
20:36:16 gmann yeah, spec can be better idea to discuss if something we miss. sean-k-mooney proposal looks ok to me for now.
20:50:56 openstackgerrit Merged openstack/nova master: Lowercase ironic driver hash ring and ignore case in cache https://review.opendev.org/711680
21:00:23 openstackgerrit melanie witt proposed openstack/nova stable/train: Lowercase ironic driver hash ring and ignore case in cache https://review.opendev.org/713739
21:36:52 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP gate: Ensure n-cpu is stopped on the subnode during evacuation https://review.opendev.org/713674
23:36:56 alex_xu dansmith: artom, thanks for the review https://review.opendev.org/#/c/687856/15/nova/compute/manager.py@8346, I reply that, but yes, we tried differnt options, and back and forward many times ourselve, looking for suggestion :)
#openstack-nova - 2020-03-19
00:08:09 artom alex_xu, sure, I'll take another look tomorrow
00:18:47 openstackgerrit Ghanshyam Mann proposed openstack/nova master: Add new default roles in os-flavor-access policies https://review.opendev.org/713697
01:05:18 openstackgerrit Merged openstack/nova master: Refine and introduce correct parameters for test_get_guest_config_numa_host_instance_topo_cpu_pinning https://review.opendev.org/713351
01:18:24 openstackgerrit melanie witt proposed openstack/nova master: DNM: try to get some debug info for bug 1844929 https://review.opendev.org/701478
01:18:24 openstack bug 1844929 in OpenStack Compute (nova) "grenade jobs failing due to "Timed out waiting for response from cell" in scheduler" [High,Confirmed] https://launchpad.net/bugs/1844929
01:30:08 openstackgerrit Brin Zhang proposed openstack/nova master: Add new default roles in os-instance-actions policies https://review.opendev.org/706470
01:36:41 openstackgerrit Brin Zhang proposed openstack/nova master: Add new default roles in os-instance-actions policies https://review.opendev.org/706470
04:39:48 openstackgerrit melanie witt proposed openstack/nova master: Synchronize sqlalchemy models with migrations for alembic 1.4.1 https://review.opendev.org/713778
07:10:11 openstackgerrit Kevin Zhao proposed openstack/nova master: fix scsi disk unit number of the attaching volume when cdrom bus is scsi https://review.opendev.org/712607
08:02:18 gibi good morning nova
08:03:45 gibi stephenfin: hi! dansmith +2 all over the qos remaining patches, could you check back to those? https://review.opendev.org/#/q/topic:bp/support-move-ops-with-qos-ports-ussuri
08:04:32 gibi stephenfin: the major change since you looked at is a compute service version check in the API
08:05:16 gibi to ensure the computes are on Ussuri version before we start moving the servers around as the feauture needs support from the compute service
08:05:37 gibi due to the PCI claim magic

Earlier   Later