Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-09-11
21:43:07 dansmith mriedem: where are you right now? we're about done here and this might be easier in person
21:43:26 mriedem i'm cranky now and hiding
21:43:29 openstackgerrit OpenStack Proposal Bot proposed openstack/nova master: Updated from global requirements https://review.openstack.org/502700
21:43:32 dansmith alright
21:46:26 openstackgerrit Balazs Gibizer proposed openstack/nova master: doc: note that custom resources are not fully supported https://review.openstack.org/501252
21:46:38 openstackgerrit OpenStack Proposal Bot proposed openstack/os-vif master: Updated from global requirements https://review.openstack.org/502708
21:47:24 mriedem i'm on a couch in a corner by the EMC business area
21:49:09 openstackgerrit OpenStack Proposal Bot proposed openstack/python-novaclient master: Updated from global requirements https://review.openstack.org/502745
21:54:16 openstackgerrit Balazs Gibizer proposed openstack/nova stable/pike: doc: note that custom resources are not fully supported https://review.openstack.org/500521
22:05:16 openstackgerrit Balazs Gibizer proposed openstack/nova master: Refactor ServerMovingTests for non-move tests https://review.openstack.org/498596
22:38:28 rybridges Hello. I had a question or 2 about vendordata in the Ocata release
22:39:39 rybridges Are we still able to write our vendordata logic stuff into a driver? I am noticing this line in the ocata code base -> https://github.com/openstack/nova/blob/stable/ocata/nova/api/metadata/base.py#L745 Is vendordata not doable through a driver anymore?
23:07:20 openstackgerrit Ghanshyam Mann proposed openstack/nova master: Merge extended availability zone controller https://review.openstack.org/502859
#openstack-nova - 2017-09-12
00:06:50 mwynne Hey guys. I'm looking at the rabbit queues (rabbitmqctl list_queues | awk '$2 != 0{print}'), and noticed that the conductor queue is always increasing. Is this normal? Can I debug this somehow?
00:12:37 mwynne netstat -tunlp
00:12:47 mwynne (whoops)
04:09:16 openstackgerrit Merged openstack/nova master: Refactor ServerMovingTests for non-move tests https://review.openstack.org/498596
04:21:31 openstackgerrit OpenStack Proposal Bot proposed openstack/nova master: Updated from global requirements https://review.openstack.org/502700
07:23:04 rabel can someone review https://review.openstack.org/#/c/494169/ ? i think except from minor changes it is ready for weeks.
08:02:11 openstackgerrit Christian Berendt proposed openstack/nova master: Choose VMware datastore by the least provisioned space https://review.openstack.org/402632
08:51:25 rabel https://review.openstack.org/#/c/402632/10 seems to be ready for final review for months. could someone have a look at it?
10:07:45 mewald Hey guys, is there a chance the block_device_mapping table can be truncated / cleared without any problems? Nova should be able to get this info from Cinder via API, right?
10:28:38 mewald ok forget it, it was bullshit
10:34:01 mewald I have a wrong / incomplete connection_info column in block_device_mapping table. New volumes are added with different connection_info JSON objects than: for example they include the secret_uuid. The lack of this uuid is currently breaking my set up . Any advice on how to fix this?
11:58:45 bhagyashris Hi all, Need guidance of How to write the upgrade script?
13:29:18 boolman I'm having problems adding iolimits on my vms. I tried adding quota on flavor,image,volume but it won't add iotune to the virsh DOM. output from image/volume/flavor show : http://ix.io/zMX
13:52:53 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/nova-specs master: Intel Fortville Dynamic Device Personalization (DDP) https://review.openstack.org/503001
14:06:15 openstackgerrit Merged openstack/python-novaclient master: Updated from global requirements https://review.openstack.org/502745
14:07:57 openstackgerrit OpenStack Proposal Bot proposed openstack/nova master: Updated from global requirements https://review.openstack.org/502700
14:09:05 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/nova-specs master: Intel Fortville Dynamic Device Personalization (DDP) https://review.openstack.org/503001
14:11:07 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/os-vif master: Migration from ``ip`` commands to ``pyroute2`` https://review.openstack.org/484386
14:18:58 openstackgerrit Balazs Gibizer proposed openstack/nova master: Transform instance.resize.error notifications https://review.openstack.org/417882
14:20:05 openstackgerrit Balazs Gibizer proposed openstack/nova master: fake_notifier: Refactor wait_for_versioned_notification https://review.openstack.org/489637
14:38:44 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/nova-specs master: Intel Fortville Dynamic Device Personalization (DDP) https://review.openstack.org/503001
15:00:53 openstackgerrit Takashi NATSUME proposed openstack/nova master: Enable cold migration with target host(1/2) https://review.openstack.org/408955
15:06:12 openstackgerrit Takashi NATSUME proposed openstack/nova master: Enable cold migration with target host(2/2) https://review.openstack.org/408964
15:06:29 openstackgerrit Takashi NATSUME proposed openstack/nova master: api-ref: Add parameters in cold migrate action https://review.openstack.org/410042
15:14:14 mriedem bauzas: are you ok with this? https://review.openstack.org/#/c/499399/ - you said you were happy after the reno got updated
15:14:24 mriedem i'd like to move this series along to get it backported to stable/pike
15:17:51 openstackgerrit Matt Riedemann proposed openstack/nova master: Remove dest node allocation if evacuate MoveClaim fails https://review.openstack.org/499878
15:17:52 openstackgerrit Matt Riedemann proposed openstack/nova master: Add a test to make sure failed evacuate cleans up dest allocation https://review.openstack.org/499877
15:17:52 openstackgerrit Matt Riedemann proposed openstack/nova master: Pass migration from API to conductor for evacuate https://review.openstack.org/500176
15:17:53 openstackgerrit Matt Riedemann proposed openstack/nova master: Create allocations against forced dest host during evacuate https://review.openstack.org/499399
15:17:53 openstackgerrit Matt Riedemann proposed openstack/nova master: Add recreate test for evacuate claim failure https://review.openstack.org/499874
15:17:54 openstackgerrit Matt Riedemann proposed openstack/nova master: Modernize set_vm_state_and_notify https://review.openstack.org/499799
15:31:19 mriedem Kevin_Zheng: yes if you go overquota after the build_request is created when we recheck in conductor, the instance will be put into ERROR state https://github.com/openstack/nova/blob/master/nova/conductor/manager.py#L1084
15:31:20 openstackgerrit Balazs Gibizer proposed openstack/nova master: factor out compute service start in ServerMovingTest https://review.openstack.org/503037
15:31:20 openstackgerrit Balazs Gibizer proposed openstack/nova master: Moving more utils to ServerResourceAllocationTestBase https://review.openstack.org/499539
15:31:36 mriedem Kevin_Zheng: that's because we can't just delete the instance at that point in conductor
15:31:53 mriedem because the use could be listing the instance via the BuildRequest before it goes to ERROR state
15:31:56 mriedem *user
15:31:59 Kevin_Zheng mriedem: I understand, but isn't that a big change to users?
15:32:49 Kevin_Zheng If large number of requests comes at the same time, in the past, I got OverQuota error
15:32:57 mriedem well, depends on what happens before counting quotas if you passed the initial check in the api and then failed when committing the reservation down in the compute
15:33:03 Kevin_Zheng But now, I got huge number of instances
15:33:12 Kevin_Zheng in Error States
15:33:30 mriedem Kevin_Zheng: are you requesting multiple instances in the same request or concurrent requests?
15:33:39 Kevin_Zheng concurrent requests
15:34:51 mriedem so in the past we'd create a reservation in the api and the quota check would include reservations
15:35:04 mriedem now we don't have reservations, and we're counting based on the instances in the cells,
15:35:23 mriedem i think what might be missing, and this might have come up in reviewing the series, was we don't count the build requests in the api db
15:35:26 mriedem but i need to look
15:35:31 mriedem unless melwitt is around
15:35:58 Kevin_Zheng Yeah, that's also what I thought as a solution :)
15:37:10 mriedem melwitt: dansmith: do you remember talking about including a count of build_requests in the API during the quota check?
15:37:43 dansmith mriedem: I think I remember talking about it
15:39:03 mriedem right so if i create 20 instances in concurrent requests, i'll probably get 10 in ERROR state
15:39:10 mriedem rather than 10 plus 10 failed requests with 409s
15:40:12 mriedem Kevin_Zheng: we should start with a bug report, can you open one?
15:40:27 Kevin_Zheng mriedem: sure, I will
15:40:51 mriedem check_num_instances_quota is just iterating the cells looking for instances, but at this point we don't have instances in the cells, we have build_requests in the api db
15:41:01 mriedem so we'd need something similar
15:42:44 mriedem and that method is called from the api and conductor, and in the case of conductor we'd have to not count build_requests because at that point that conductor calls check_num_instances_quota we have the instances created in the cells
15:43:11 openstackgerrit Ghanshyam Mann proposed openstack/nova-specs master: Spec to remove the hide server address config options Partial implement blueprint remove-configurable-hide-server-address-feature https://review.openstack.org/502516
15:47:21 mriedem i see one issue with counting via build_requests is that when we count instances in the api, we also count cores and ram, which are fields on the instance in the db,
15:47:44 mriedem with a build_request, we have a serialized instance record, so we can count total instances via build_requests, but not cores/ram...unless we used placement
15:48:09 mriedem we want to use placement for counting quotas anyway...which we are going to talk about tomorrow morning
15:50:15 rybridges Hello. I had a question or 2 about vendordata in the Ocata release
15:50:21 rybridges Are we still able to write our vendordata logic stuff into a driver? I am noticing this line in the ocata code base -> https://github.com/openstack/nova/blob/stable/ocata/nova/api/metadata/base.py#L745 Is vendordata not doable through a driver anymore?
15:51:31 mriedem rybridges: you should be able to use the old style driver stuff
15:51:35 mriedem in ocata
15:58:19 Kevin_Zheng https://bugs.launchpad.net/nova/+bug/1716706
15:58:20 openstack Launchpad bug 1716706 in OpenStack Compute (nova) "Should count instances in build requests when check quotas" [Undecided,New] - Assigned to Zhenyu Zheng (zhengzhenyu)
16:01:45 openstackgerrit Balazs Gibizer proposed openstack/nova master: Test resource allocation during soft delete https://review.openstack.org/495159
16:03:12 mriedem Kevin_Zheng: thanks, added it here https://etherpad.openstack.org/p/nova-ptg-queens-cells
16:03:15 mriedem L22
16:03:26 mriedem also, i realized we can't use placement in the api to count cores/ram
16:03:32 mriedem since those allocations aren't created by then
16:04:46 Kevin_Zheng could use flavor instead
16:09:35 melwitt mriedem: we talked about counting build_requests but the consensus was to not because they're in a separate DB from instances and it would be possible for both to exist at the same time and result in a wrong count (doubled)
16:11:20 mriedem melwitt: when we check in the API they shouldn't exist at the same time, although that might be hard to know...
16:11:33 mriedem couldn't we exclude by uuid or something?
16:11:42 openstackgerrit Merged openstack/nova master: fake_notifier: Refactor wait_for_versioned_notification https://review.openstack.org/489637
16:12:10 Kevin_Zheng mriedem, records of other existing instances, might be doubled
16:12:16 mriedem select count(build_requests) where user/project; build a list of uuids; then iterate cells getting a count of instances where uuid is not in that list?
16:12:20 melwitt mriedem: we check in conductor though (check in API is for cells v1 I believe)
16:12:32 mriedem melwitt: yeah in conductor we wouldn't include the check for build_requests
16:12:40 mriedem because we know they both exist at the time of that check
16:12:53 mriedem melwitt: i'm not talking about the recheck
16:12:56 melwitt then how is this happening in the API?

Earlier   Later