Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-11-21
21:42:01 mriedem artom: can you tl;dr the diff from the approved spec in rocky to stein?
21:42:25 mriedem because this essentially sounds like, do the RT/claim stuff like cold migration already does (and what nikola's old patch did)
21:42:54 mriedem was the rocky version of the spec much more dependent on actually modeling numa in placement?
21:44:07 openstackgerrit Zane Bitter proposed openstack/nova stable/rocky: Don't use private interface in oslo.service https://review.openstack.org/619360
21:44:21 sean-k-mooney the stine one has been refined somewhat to not depend on plamcennt at all
21:44:36 openstackgerrit melanie witt proposed openstack/nova-specs master: Propose counting quota usage from placement and API database https://review.openstack.org/509042
21:45:54 sean-k-mooney mriedem: it will prefrom claims in the RT on the destination node and pass back the topology info as ovo form the dest to the source the same way we did for the multi port bindings spec
21:47:41 mriedem sean-k-mooney: i just read the stein one
21:47:52 mriedem i was looking for the summary on the diff from the rocky spec
21:48:36 sean-k-mooney ah ok stephen just proposed tthe rocky one as is so the delta should just be comparing to v1 of the stein one
21:50:08 sean-k-mooney there was no content change when stepen reproposed it so https://review.openstack.org/#/c/599587/1..6/specs/stein/approved/numa-aware-live-migration.rst will show you the delta
21:51:05 sean-k-mooney basicaly we added detail for the resouce claimes section and how the nume info on the destinaiton woudl be generated
21:53:20 artom mriedem, whoa, thanks dude
21:53:43 artom What sean-k-mooney said. And also how the destination will communicate said NUMA stuff to the source
21:54:30 mriedem right, ok so no placement
21:54:31 mriedem got it
21:54:32 artom And more explicitly spelled out why we can't use just placement (even if it was ready, which it isn't) and still need the resource tracker
21:54:53 mriedem given that, i'm going to f off for thanksgiving now
21:55:05 artom Fair enough, r right off then :)
21:55:12 artom *f
21:55:13 sean-k-mooney mriedem: enjoy the break o/
21:55:52 artom I haven't fully parsed your feedback then, but I'll have a new version for you when you're back
21:55:57 artom s/then/yet
21:55:58 artom wtf brain
22:00:04 openstackgerrit Matt Riedemann proposed openstack/nova stable/ocata: Default embedded instance.flavor.is_public attribute https://review.openstack.org/619366
22:38:13 sean-k-mooney cfriesen: by the way were you aware we approved a spec for vTPM on hyperv back in liberty https://specs.openstack.org/openstack/nova-specs/specs/liberty/approved/hyper-v-vtpm-devices.html
22:39:34 openstackgerrit Merged openstack/nova master: Make supports_direct_io work on 4096b sector size https://review.openstack.org/616580
22:42:22 sean-k-mooney cfriesen: it also got appoved in mitaka but i dont think they ever implmented it
22:44:19 sean-k-mooney actully they do http://git.openstack.org/cgit/openstack/compute-hyperv/tree/compute_hyperv/nova/vmops.py#n1334
22:45:39 cfriesen sean-k-mooney: hadn't seen it. now I'm obligated to go check it out. :)
22:48:32 cfriesen what git tree is that? I don't see any use of "vtpm" in nova, so it seems like it was never implemented on the nova side.
22:48:58 sean-k-mooney its in openstack/compute-hyperv
22:49:52 sean-k-mooney https://github.com/openstack/compute-hyperv its still active aparently
22:50:24 cfriesen looks like it was superceded by https://blueprints.launchpad.net/nova/+spec/hyper-v-shielded-vms
22:52:57 sean-k-mooney im going to add thi to the reply i was sending to fungi on the ml
22:53:55 cfriesen was proposed late for ocata, but doesn't seem to have been re-proposed for pike
22:54:53 sean-k-mooney well hyperv had an out of tree driver just like power so i guess they did not need to reporpose to support it
23:14:10 fungi thanks for the details, sean-k-mooney!
23:18:14 fungi and yeah, in the future where hosts which have legit tpms can pass them through the hypervisor to guests, as long as it's apparent to the guest whether it's getting an actual tpm or an emulated one i think that's probably fine
23:19:08 fungi at least running the emulated tpm on the hypervisor is still a step up from running an emulated tpm inside the guest itself
23:22:16 fungi certainly an attacker with access to the host can clone and impersonate the tpm (unlike with a proper hardware one) but that's a higher barrier than compromising the guest
23:51:16 openstackgerrit Merged openstack/nova master: Use long_rpc_timeout in select_destinations RPC call https://review.openstack.org/607735
23:55:45 openstackgerrit Merged openstack/nova stable/pike: Handle HostMappingNotFound when deleting a compute service https://review.openstack.org/583567
#openstack-nova - 2018-11-22
02:11:25 openstackgerrit Fan Zhang proposed openstack/nova master: Log disk transfer stats in live migration monitor. https://review.openstack.org/619395
02:18:23 openstackgerrit Jason SUN proposed openstack/nova master: For instances end up in cell0, set its availability_zone to None. https://review.openstack.org/618737
03:36:18 openstackgerrit Yikun Jiang proposed openstack/nova master: Add live migration timeout action https://review.openstack.org/619143
03:36:18 openstackgerrit Yikun Jiang proposed openstack/nova master: Remove live_migration_progress_timeout config https://review.openstack.org/619142
04:32:45 openstackgerrit Fan Zhang proposed openstack/nova master: Log disk transfer stats in live migration monitor. https://review.openstack.org/619395
05:13:57 openstackgerrit Takashi NATSUME proposed openstack/nova master: Add a bug tag for nova doc https://review.openstack.org/619434
05:17:55 openstackgerrit Merged openstack/python-novaclient master: Fix a type of block_device_mapping_v2 in a comment https://review.openstack.org/619145
06:15:59 openstackgerrit Jason SUN proposed openstack/nova master: For instances end up in cell0, set its availability_zone to None. https://review.openstack.org/618737
06:20:04 openstackgerrit Yikun Jiang proposed openstack/nova master: Add live migration timeout action https://review.openstack.org/619143
06:20:05 openstackgerrit Yikun Jiang proposed openstack/nova master: Remove live_migration_progress_timeout config https://review.openstack.org/619142
06:50:31 openstackgerrit Zhenyu Zheng proposed openstack/nova master: Bump compute service to indicate attach/detach root volume is supported https://review.openstack.org/614750
07:17:27 openstackgerrit OpenStack Proposal Bot proposed openstack/nova stable/rocky: Imported Translations from Zanata https://review.openstack.org/614757
08:51:06 openstackgerrit Yikun Jiang proposed openstack/nova master: Add live migration timeout action https://review.openstack.org/619143
08:51:06 openstackgerrit Yikun Jiang proposed openstack/nova master: Remove live_migration_progress_timeout config https://review.openstack.org/619142
10:28:50 openstackgerrit Sylvain Bauza proposed openstack/nova master: libvirt: implement reshaper for vgpu https://review.openstack.org/599208
11:38:01 openstackgerrit Jason SUN proposed openstack/nova master: For instances end up in cell0, set its availability_zone to None. https://review.openstack.org/618737
11:40:45 gary_perkins Hi folks! General question about spinning up instances and storage: I have compute nodes with very little spare storage. Is it possible to confine all storage operations to go to Cinder? Even ephemeral storage and spinning up image-backed instances where normally local storage is allocated? Is this something flavors could handle or should handle? TIA
11:41:16 Miouge Is there a way to expose the compute node an instance is running on in metadata? I would have to do with DynamicJSON vendordata?
11:42:47 gary_perkins mriedem: Thanks for your help yesterday. The issue is still not resolved, but I'll have to return to it another time.
11:46:26 openstackgerrit Balazs Gibizer proposed openstack/nova master: Calculate RequestGroup resource provider mapping https://review.openstack.org/616239
11:46:28 openstackgerrit Balazs Gibizer proposed openstack/nova master: Pass resource provider mapping to neutronv2 api https://review.openstack.org/616240
11:46:30 openstackgerrit Balazs Gibizer proposed openstack/nova master: Send RP uuid in the port binding https://review.openstack.org/569459
11:46:31 openstackgerrit Balazs Gibizer proposed openstack/nova master: Test boot with more ports with bandwidth request https://review.openstack.org/573317
11:46:33 openstackgerrit Balazs Gibizer proposed openstack/nova master: Extend RequestGroup object for mapping https://review.openstack.org/619527
11:46:35 openstackgerrit Balazs Gibizer proposed openstack/nova master: Fill the RequestGroup mapping during schedule https://review.openstack.org/619528
11:46:37 openstackgerrit Balazs Gibizer proposed openstack/nova master: Recalculate request group - RP mapping during re-schedule https://review.openstack.org/619529
12:00:43 openstackgerrit Balazs Gibizer proposed openstack/nova master: Transfer port.resource_request to the scheduler https://review.openstack.org/567268
12:00:44 openstackgerrit Balazs Gibizer proposed openstack/nova master: Extend RequestGroup object for mapping https://review.openstack.org/619527
12:00:46 openstackgerrit Balazs Gibizer proposed openstack/nova master: Calculate RequestGroup resource provider mapping https://review.openstack.org/616239
12:00:48 openstackgerrit Balazs Gibizer proposed openstack/nova master: Fill the RequestGroup mapping during schedule https://review.openstack.org/619528
12:00:50 openstackgerrit Balazs Gibizer proposed openstack/nova master: Pass resource provider mapping to neutronv2 api https://review.openstack.org/616240
12:00:52 openstackgerrit Balazs Gibizer proposed openstack/nova master: Recalculate request group - RP mapping during re-schedule https://review.openstack.org/619529
12:00:55 openstackgerrit Balazs Gibizer proposed openstack/nova master: Send RP uuid in the port binding https://review.openstack.org/569459
12:00:57 openstackgerrit Balazs Gibizer proposed openstack/nova master: Test boot with more ports with bandwidth request https://review.openstack.org/573317
12:37:05 awalende Does anyone know how I can troubleshoot oslo_timeouts exceptions in pre-live migrations? Only 1/5 migration attempts work, all others run in somekind of oslo messaging timeout:
12:37:22 awalende Pre live migration failed at blies: MessagingTimeout: Timed out waiting for a reply to message ID 359a5c51e98f4474ac
12:46:33 mhen awalende, we're experiencing the exact same thing in one of our infrastructures - we're still investigating but had no luck identifying the cause so far
12:51:36 awalende We tried to cram up some migration timeout values in nova.conf but I believe it has nothing to do with the migration itself.
12:52:52 awalende Don't know what exactly is happening in the pre-migration. Probably networking and maybe cinder stuff...
13:01:52 sean-k-mooney awalende: pre-migration has a number of check and we do wire up some of the networking on the destination
13:04:06 sean-k-mooney awalende: the timeout could be as a result of messages getting lost or connection issue btween the compute nodes and amqp
13:05:04 sean-k-mooney awalende: have you seen any log messages related to services reconnecting to rabbitmq that could indicate network issues
13:06:56 awalende sean-k-mooney, I have enabled debug logging on the compute nodes, but the only amqp related stuff is the timeOut-Exception itself. We are running HA rabbit on 3 Controllers, I'll check there for any further infos
13:17:40 mhen awalende, are you using persistent queues?
13:18:20 awalende we don't
13:18:49 awalende because we had often problems with split-brain issues on rabbit
13:20:36 mhen good to know, then we can rule that out as the culprit for now
15:07:23 gary_perkins Hi folks! General question about spinning up instances and storage: I have compute nodes with very little spare storage. Is it possible to confine all storage operations to go to Cinder? Even ephemeral storage and spinning up image-backed instances where normally local storage is allocated? Is this something flavors could handle or should handle? TIA
15:11:25 mnaser simple and clean backport if anyone is around -- https://review.openstack.org/#/c/619349/
15:11:51 artom gary_perkins, I think you want boot from volume instances
15:22:26 gary_perkins artom: yes, but I want to transparently enforce the use of cinder volumes for all storage so that the compute nodes don't run out of their limited capacity
15:24:46 gary_perkins and by, transparently, I mean, I don't want users to have to create a bootable cinder volume prior to spinning up an instance
15:41:59 gibi gary_perkins: have you thought about mounting a cinder volume to each compute and use that space as nova's instance directory?
15:42:58 gary_perkins gibi: hmmm... that's an idea!
15:45:54 gary_perkins gibi: thanks! I wasn't sure one could do that. I'll give that a try!
16:25:17 artom gibi, can you do that? I would have thought that Cinder volumes (control plane) wouldn't be accessible from the data plane (tenants)
16:25:40 artom gary_perkins, there was a spec for flavor-enforced boot from volume

Earlier   Later