Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-06-06
15:43:49 mriedem pksingh: you don't
15:44:01 mriedem the nova db tables for network resources are only ever used for nova-network
15:44:14 mriedem if you use neutron, regardless of microversion, those tables weren't used
15:44:36 mriedem if you had some out of tree code that was relying on those tables getting updated, that out of tree code was incorrect
15:45:02 pksingh mriedem: ok, great, so to get floating IPs, any how we need to use neutron APIs
15:45:10 mriedem yes
15:45:25 mriedem or use a microversion<2.44
15:45:53 mriedem but it's in your best long-term interest to use the networking API directly (neutron)
15:46:55 mriedem gibi: i'll apologize ahead of time for not making any progress on notification stuff while you were out for 2 weeks
15:49:27 pksingh mriedem: thanks a lot for your help, one more question
15:49:49 pksingh mriedem: so nova will not show floating IP in server detail get request
15:50:16 mriedem pksingh: floating ips will still show up in the server 'addresses' details
15:50:39 pksingh mriedem: ok, from where do you get it?
15:50:54 mriedem i'd like to know what you're doing with the nova database...
15:51:05 mriedem but they come from the instance_info_cache table
15:51:09 mriedem in the network_info column
15:51:17 pksingh mriedem: i work in openstack container team
15:51:31 mriedem why do you need to know the internal data model then?
15:51:36 pksingh mriedem: we were also thinking to assign floating IP to containers
15:51:37 cdent dansmith: I think you've said in the past that you'd prefer to have your attention drawn to mailing list posts instead of having to pay constant attention: ed started a thread on "complex allocations". your input would be most appreciated. I followed up a bit later with a lot of explicit questions. http://lists.openstack.org/pipermail/openstack-dev/2017-June/117913.html
15:51:47 mriedem pksingh: you should probably not copy nova
15:52:03 pksingh mriedem: but we came to know that nova has deprecated this api
15:52:27 gibi mriedem: hi! no problem, I made no progress either ;)
15:52:31 pksingh mriedem: we are not copying the nova, just was curious how nova does that
15:52:31 mriedem pksingh: as the spec says, the compute api is deprecated because (1) nova-network is deprecated and (2) it's a proxy to neutron
15:53:21 mriedem pksingh: i don't mean "don't copy nova" because nova is great and shouldn't be copied, i'm saying don't copy it because of technical debt and cruft and just legacy stuff that you probably don't want to consider when building a new system
15:53:38 dansmith cdent: yeah I saw, but haven't read it yet...
15:54:11 cdent ✔
15:54:47 pksingh mriedem: yes, we decide not to develop any such api in zun, but we were thinking how to show Floating IPs assigned to containers
15:55:12 pksingh mriedem: so we tried to understand the nova way, and decide if that suits us
15:55:22 mriedem probably also don't want to copy server['addresses'] format from nvoa
15:55:32 mriedem mordred would probably say the same
15:55:47 pksingh mriedem: yes, we have different approach
15:56:28 mriedem if there are users asking for the feature, i'd ask them how they'd like the information represented or what's most useful to them
15:56:43 mordred yes please. don't copy the server['addresses']
15:56:45 mordred format
15:56:58 mriedem i'd also caution against proxying too much to neutron to fulfill your rest api response
15:58:18 mordred pksingh: also keep in mind that it's perfectly valid to use both floating ips and fixed ips ... "floating ip" is not an alias for "get me an ip that has external access"
15:58:57 mordred pksingh: and that especially with ipv6, tying too much thinking around the specific construct of floating_ip when a NAT construct is not specifically desired makes things harder on users
15:59:41 pksingh mordred: ok
16:00:04 openstackgerrit Stephen Finucane proposed openstack/nova master: Convert floating IP tests to ddt https://review.openstack.org/471360
16:00:04 openstackgerrit Stephen Finucane proposed openstack/nova master: Catch neutronclient.NotFound on floating deletion https://review.openstack.org/410815
16:00:05 openstackgerrit Stephen Finucane proposed openstack/nova master: Convert additional disassociate tests to mock https://review.openstack.org/471408
16:02:26 sfinucan mriedem: ...and reworked ^
16:04:20 sdague mriedem: https://review.openstack.org/#/c/468356/ you can un WIP this because the neutronclient changes are now merged
16:07:31 mriedem sdague: which neutronclient changes?
16:07:36 mriedem for global request id? that wasn't the reason for the wip
16:07:46 sdague mriedem: what was the reason?
16:07:55 mriedem those went into neutronclient 6.3.0 which was blacklisted from g-r
16:08:04 sdague it was unblacklisted 33 hours ago
16:08:27 mriedem https://github.com/openstack/requirements/commit/df081c07c68ee2b529291c8fec6f5686ab97b002
16:08:28 mriedem i see
16:08:30 sdague https://review.openstack.org/#/c/470047/
16:08:34 mriedem pretty sparse commit message
16:09:01 sdague yeh, well, armax just responded on my status email and said they reverted it
16:09:02 mriedem heh, change the api is the answer?
16:09:34 sdague and if you look at the requirements in nova, we're now back at that minimum
16:09:56 armax mriedem: sparse the commit message of the requirement revert?
16:10:02 mriedem yeah, i see the depends-on
16:10:15 armax mriedem: we like people to follow the crumbs
16:10:22 mriedem so anyone with neutronclient 6.3.0 hitting a server that doesn't have that api workaround is still broken client-side
16:10:23 armax like a good TV show that leaves you hanging
16:10:33 mriedem which kind of sucks
16:10:47 armax mriedem: the server change has been backported for all stable branches
16:10:58 armax mriedem: your point was taken into account
16:10:59 mriedem yeah i saw
16:11:18 armax surely 6.3.0 might work prior to newton
16:11:26 armax server-side I mean
16:11:35 sdague mriedem: yeh, well, regardless, it's good enough for the nova patch to move forward
16:11:40 armax but if you ain’t test it you ain’t say it for soor
16:11:42 armax sure
16:11:55 sdague armax: this was mostly fighting against pecan magic?
16:12:00 mriedem sdague: i know this isn't nova's problem
16:12:15 mriedem i just feel bad for the puppies and kittens that now have to die
16:12:26 mriedem more bad for the kittens because, you know, dogs...
16:12:58 armax sdague: I don’t believe so, the issue seemed to be latent and just surface because of the client change that dropped the .json suffix on the API URIs
16:20:04 efried mriedem FYI PowerVM support matrix adds are ready: https://review.openstack.org/#/c/470999/
16:20:21 mriedem oh i know
16:20:24 mriedem it's starred
16:20:44 efried mriedem On another topic, thinking it may be a good idea to blast the ML for the InstanceInfo change, since it would affect OOT drivers.
16:21:33 mriedem go head
16:21:35 mriedem *ahead
16:26:34 openstackgerrit Dan Smith proposed openstack/nova master: Fix and optimize external_events for multiple cells https://review.openstack.org/445142
16:32:49 openstackgerrit Merged openstack/nova master: Remove DeviceIsBusy exception https://review.openstack.org/465723
16:33:28 openstackgerrit Merged openstack/nova master: libvirt: Use config types to parse XML for root disk https://review.openstack.org/411941
16:34:05 openstackgerrit Merged openstack/nova master: Remove oslo.config deprecated parameter enforce_type https://review.openstack.org/462364
16:34:41 openstackgerrit Merged openstack/nova master: Modify incorrect comment on return_reservation_id https://review.openstack.org/471278
16:34:58 mhenkel_ mriedem: would you mind to check https://blueprints.launchpad.net/openstack/?searchtext=opencontrail-nova-vif-plugin and approve?
16:35:21 openstackgerrit Merged openstack/nova master: Remove incorrect comments in multiple_create https://review.openstack.org/471273
16:51:37 openstackgerrit Merged openstack/nova master: Support tag instances when boot(2/4) https://review.openstack.org/459569
16:52:16 openstackgerrit Merged openstack/nova master: [placement] adjust resource provider links by microversion https://review.openstack.org/468923
16:52:56 openstackgerrit Merged openstack/nova master: make route and controller in alpha sequence https://review.openstack.org/469380
16:53:31 openstackgerrit Merged openstack/nova master: ComputeDriver.get_info not limited to inst name https://review.openstack.org/471106
16:54:07 openstackgerrit Merged openstack/nova master: Refactor a test method including 7 test cases https://review.openstack.org/395436
16:54:44 openstackgerrit Merged openstack/nova master: Remove usage of parameter enforce_type https://review.openstack.org/469745
16:55:23 openstackgerrit Merged openstack/nova master: Pass a list of instance UUIDs to scheduler https://review.openstack.org/465171
16:55:29 openstackgerrit Matt Riedemann proposed openstack/nova master: Send out notifications when instance tags changed https://review.openstack.org/459923
16:55:30 openstackgerrit Matt Riedemann proposed openstack/nova master: Move notifications/objects/test_base.py https://review.openstack.org/471425
17:11:59 mlavalle mriedem: so this https://bugs.launchpad.net/nova/+bug/1696125
17:11:59 openstack Launchpad bug 1696125 in OpenStack Compute (nova) "Detach interface failed - Unable to detach from guest transient domain (pike)" [High,Confirmed]

Earlier   Later