Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-10-09
15:00:16 dansmith mriedem: ah okay I had looked
15:00:36 mriedem they don't send you 10 emails for a month before that happens?
15:00:50 mriedem most of my email now is school spam
15:00:53 efried dansmith To be clear, by "hamstringing" you mean settling on the param being an allocations object versus some other thing we would have more time to think about and design?
15:01:19 dansmith efried: no, passing allocations into the virt driver is not a huge deal,
15:01:22 bauzas mriedem: just one name : strike
15:01:35 dansmith efried: but if libvirt gets an rp uuid and a count into spawn right now, what does it do?
15:01:38 mriedem bauzas: oh i see
15:01:56 alex_xu efried: jaypipes thanks for taking care the trait spec
15:01:59 bauzas grrrrrr
15:02:08 efried dansmith Well, it *can* ignore the new param and just use flavors.
15:02:24 efried dansmith Or, if it has the time and resources, it can implement the thing that consumes the allocations.
15:02:26 dansmith efried: sure, but what purpose have you served by doing that for no reason?
15:02:41 efried Because of the ability to do that second thing.
15:02:49 bauzas efried: dansmith: saw your conversation, yeah I'll just use the flavor for that
15:02:56 efried dansmith PowerVM also has use for it, FWIW.
15:03:15 efried Because we're hamstrung by the PCI manager, and the sooner we can move to generic RPs, the better.
15:03:20 dansmith efried: fine, so add that thing, I'm fine with that, I'd just like not to linearize the libvirt stuff behind it
15:03:25 dansmith efried: go for it
15:03:48 efried dansmith Oh, totally. I don't think it serializes anything. I'll see if I can hit it. Thanks.
15:05:34 mriedem http://logstash.openstack.org/#dashboard/file/logstash.json?query=message%3A%5C%22device-mapper%3A%20reload%20ioctl%20on%5C%22%20AND%20%20%20message%3A%5C%22failed%3A%20Invalid%20argument%5C%22%20AND%20%20%20tags%3A%5C%22screen-c-vol.txt%5C%22&from=7d
15:05:56 openstackgerrit Stephen Finucane proposed openstack/nova master: utils: Replace 'utf8' with oslo.utils equivalent https://review.openstack.org/466799
15:07:52 sean-k-mooney mriedem: just an fyi, i mentioned this to jay and dan earlier but i have asked for the intel nfv ci to be swapped to silent mode.
15:08:31 sean-k-mooney mriedem: ssh to gerrit seams to have been restored but logs are not being uploaded and jobs are failing so i am guessing we are still being blocked form connecting to some sites
15:09:08 openstackgerrit Stephen Finucane proposed openstack/nova master: doc: Rework man pages https://review.openstack.org/502105
15:09:16 sean-k-mooney mriedem: i have asked for the vpn and ssh logins so i can review the local copy of the log and see if i can determin what is needed to restore the ci to a working state
15:10:07 sean-k-mooney mriedem: but ya just an fyi i did not want to have it spaming every patchset when i know its currently broken
15:13:06 mriedem sean-k-mooney: ok
15:17:20 dansmith jaypipes: are you rebasing your nrp set?
15:17:47 jaypipes dansmith: on to the no-orm-resource-providers series, yes.
15:18:03 jaypipes dansmith: and in the process I'll be fixing up requested things on the n-r-p patches
15:18:08 dansmith jaypipes: okay, does it have to go at the end?
15:18:21 dansmith there's a ton of stuff to do still in the de-orm set
15:18:22 bauzas jaypipes: I'm tempted to +2 eventually your no-orm bottom change :)
15:18:46 dansmith and it's kinda unfortunate to have to chew all of that before we can make any progress on nrp :/
15:18:53 jaypipes dansmith: lemme see if there's a middle patch I could go from... one sec
15:19:14 dansmith jaypipes: or maybe a reasonable re-ordering of the lower set so you could tack on earlier or something
15:21:36 jaypipes dansmith: https://review.openstack.org/#/c/509035 is where I could tie on n-r-p. I suppose I could re-order to have that one land earlier in the series.
15:22:12 dansmith might be good if it's not too much work
15:23:20 openstackgerrit Balazs Gibizer proposed openstack/nova master: Fix sending legacy instance.update notification https://review.openstack.org/510603
15:25:24 sean-k-mooney johnthetubaguy: o/
15:25:35 johnthetubaguy sean-k-mooney: hi
15:26:10 sean-k-mooney johnthetubaguy: QQ im just looking at the compute manager class in the pre live migration fuction https://github.com/openstack/nova/blob/664322cae7fcb76ed27b3afd7a7b6434e6e23229/nova/compute/manager.py#L5542-L5557
15:26:39 sean-k-mooney johnthetubaguy: this is where we will have to consume the new neutron port bining api above the virt drivers correct
15:26:45 jaypipes dansmith: ok, I'll see what I can do
15:27:11 dansmith jaypipes: regardless, I've +2d a bunch of the early part of that de-orm set
15:27:31 dansmith so if our gate was working at all, we could merge those soonish
15:28:01 sean-k-mooney johnthetubaguy: so instead of getting current bindings with self.network_api.get_instance_nw_info(context, instance) i will need to create the destingation port bindings and the activate them later
15:28:22 sean-k-mooney johnthetubaguy: just trying to make sure https://review.openstack.org/#/c/375580/5/specs/queens/approved/neutron-new-port-binding-api.rst is accurate
15:28:33 johnthetubaguy sean-k-mooney: it calls through to here today: https://github.com/openstack/nova/blob/master/nova/network/neutronv2/api.py#L327
15:28:56 johnthetubaguy https://github.com/openstack/nova/blob/664322cae7fcb76ed27b3afd7a7b6434e6e23229/nova/network/neutronv2/api.py#L327
15:30:13 openstackgerrit Balazs Gibizer proposed openstack/nova master: Remove dead code of api.fault notification sending https://review.openstack.org/505164
15:30:47 sean-k-mooney johnthetubaguy: right so that si called on line https://github.com/openstack/nova/blob/664322cae7fcb76ed27b3afd7a7b6434e6e23229/nova/compute/manager.py#L5556
15:31:12 gibi stephenfin, mriedem: could you check back to https://review.openstack.org/#/c/505164 I updated the reno based on stephenfin's excellent suggestions
15:31:17 johnthetubaguy sean-k-mooney: yes, that's the one
15:31:41 johnthetubaguy sean-k-mooney: but should be no changes to the compute manager, only changes in the network layer
15:33:07 sean-k-mooney johnthetubaguy: well what i wanted to check is where is the xml generated for the remote node is it is the prelivemigration on line 5547 or after the networking call on 5556
15:34:03 sean-k-mooney i had the perception that i might need to change the order of those calls also im currently trace through the code to confim though
15:35:49 johnthetubaguy sean-k-mooney: for the XML stuff, its all about adding the correct details into migrate_data so this method gets it correct: https://github.com/openstack/nova/blob/664322cae7fcb76ed27b3afd7a7b6434e6e23229/nova/virt/libvirt/driver.py#L6357
15:37:41 sean-k-mooney ah cool thanks that is what i was looking for. i can trace out how that gets populated now that i know where to look
15:39:18 johnthetubaguy sean-k-mooney: have a look at this bit: https://github.com/openstack/nova/blob/664322cae7fcb76ed27b3afd7a7b6434e6e23229/nova/virt/libvirt/driver.py#L6930
15:58:14 openstackgerrit Balazs Gibizer proposed openstack/nova master: use already loaded BDM in instance. https://review.openstack.org/483324
15:58:15 openstackgerrit Balazs Gibizer proposed openstack/nova master: use already loaded BDM in instance. (2) https://review.openstack.org/483955
15:58:15 openstackgerrit Balazs Gibizer proposed openstack/nova master: use already loaded BDM in instance.create https://review.openstack.org/483969
15:58:38 sean-k-mooney johnthetubaguy: right so that is part of the virt diriver prelive migration function which is called before setup_networks_on_host which is why i taught i might need to swap the order of those calls
16:00:24 gibi mriedem, stephenfin: I had to rebase https://review.openstack.org/#/c/483324 to adapt some new test cases from master. You were both +2 before so could you check it again?
16:01:05 sean-k-mooney johnthetubaguy: any way you confimed im looking more or less in the right part of the code so ill trace through it and confim the current workflow and what we need for the new workflow
16:03:44 johnthetubaguy sean-k-mooney: I guess you want to send the updated port bindings to os-vif on the destination node, before handing control back to the source node?
16:05:46 sean-k-mooney johnthetubaguy: am well i want to create the binding for the destionation before i create destination xml so that if the interfaces defintion change i can update that section of xml. e.g. linux bridge to ovs or iptables to contrack firewall
16:05:53 jaypipes dansmith: I don't get it... I cherry-pick this patch locally: https://review.openstack.org/#/c/509030/ onto master (trying to re-order the de-orm series) and the cherry-pick applies code that is from the patch underneath it (https://review.openstack.org/#/c/509029/) :(
16:05:53 openstackgerrit Eric Berglund proposed openstack/nova master: WIP(5): PowerVM driver: ovs vif https://review.openstack.org/422512
16:06:30 johnthetubaguy sean-k-mooney: agreed
16:06:47 dansmith jaypipes: if you want to pastebin a list in the right order, I can try to do it
16:07:06 jaypipes dansmith: I thought cherry-pick was supposed to only apply the diff of the specific patch and nothing else?
16:07:08 sean-k-mooney johnthetubaguy: in the current flow i generate teh xml and then bind which mean today when i live migrate form a lb host to ovs it work but livirt still plugs the tap into linux brige and neutron wait for the port to show up on ovs
16:07:30 dansmith jaypipes: not really, but it's only supposed to do the one you ask for, which is what you're saying yeah
16:08:51 sean-k-mooney johnthetubaguy: works was ment to be in qoutes because it live migrate but you networking is screwed untill you hard reboot the vm
16:09:04 jaypipes dansmith: well, this is what I'm *trying* to do... http://paste.openstack.org/show/623122/
16:09:44 dansmith jaypipes: k, lemme try
16:09:58 jaypipes dansmith: and what I tried was git checkout master && git pull && checkout -b de-orm-rp && git cherry-pick 0c1a5bc
16:10:16 jaypipes dansmith: but that applied changes from ca983c1 for some reason
16:13:32 openstackgerrit Dan Smith proposed openstack/nova master: rp: de-ORM ResourceProvider.get_by_uuid() https://review.openstack.org/509025
16:13:32 openstackgerrit Dan Smith proposed openstack/nova master: rp: Move RP._get|set_aggregates() to module scope https://review.openstack.org/509026
16:13:33 openstackgerrit Dan Smith proposed openstack/nova master: rp: remove CRUD operations on Inventory class https://review.openstack.org/509029
16:13:33 openstackgerrit Dan Smith proposed openstack/nova master: rp: streamline InventoryList.get_all_by_rp_uuid() https://review.openstack.org/509030
16:13:34 openstackgerrit Dan Smith proposed openstack/nova master: rp: remove dead code in Allocation._create_in_db() https://review.openstack.org/509031
16:13:34 openstackgerrit Dan Smith proposed openstack/nova master: rp: remove ability to delete 1 allocation record https://review.openstack.org/509032
16:13:35 openstackgerrit Dan Smith proposed openstack/nova master: rp: fix up AllocList.get_by_resource_provider_uuid https://review.openstack.org/509033
16:13:35 openstackgerrit Dan Smith proposed openstack/nova master: rp: rework AllocList.get_all_by_consumer_id() https://review.openstack.org/509035
16:13:36 openstackgerrit Dan Smith proposed openstack/nova master: rp: Remove RP.get_traits() method https://review.openstack.org/509027
16:13:36 openstackgerrit Dan Smith proposed openstack/nova master: rp: move RP._set_traits() to module scope https://review.openstack.org/509028
16:13:37 openstackgerrit Dan Smith proposed openstack/nova master: rp: remove _HasAResourceProvider mixin https://review.openstack.org/509036
16:13:37 openstackgerrit Dan Smith proposed openstack/nova master: rp: break functions out of _set_traits() https://review.openstack.org/509908
16:13:39 dansmith jaypipes: that look right? ^
16:14:03 dansmith had to re-order one thing from your list, but otherwise that ordering was fine
16:15:04 jaypipes dansmith: yep, that'll work. thanks man, really appreciated.
16:15:09 dansmith np
16:15:19 jaypipes dansmith: I see you just slid the CRUD one before the Streamline one. k.
16:15:30 dansmith had to else huge esplosion
16:15:46 jaypipes yes, that's the espolosion I was dealing with..

Earlier   Later