| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-07-06 | |||
| 16:54:10 | sean-k-mooney | the corrent nova behavior is to just not wait for ovn in prelive migration | |
| 16:54:26 | ralonsoh | exactly and wait for the port plug event | |
| 16:54:28 | sean-k-mooney | so once your neutron patch is up we can add a nova one that will make it start waiting | |
| 16:55:24 | sean-k-mooney | ping me when the neutorn patch is up if you would like me to write the nova one. its should not be that hard but i probaly will not get to it until monday | |
| 16:55:45 | ralonsoh | sean-k-mooney, sure, tomorrow | |
| 16:58:26 | sean-k-mooney | ralonsoh: no worries its getting late for you | |
| 17:26:53 | stephenfin | sean-k-mooney: https://review.opendev.org/c/openstack/os-vif/+/798055 looks pretty good. If you've time to address my comments tonight I'll +2 in the AM | |
| 17:28:19 | sean-k-mooney | stephenfin: am i want to poc something else this evening but ill adress it tomorow | |
| 17:35:43 | opendevreview | Merged openstack/placement master: Fix oslo policy DeprecatedRule warnings https://review.opendev.org/c/openstack/placement/+/799418 | |
| 18:16:30 | opendevreview | Merged openstack/nova master: Fix oslo policy DeprecatedRule warnings https://review.opendev.org/c/openstack/nova/+/799422 | |
| 18:52:39 | opendevreview | Merged openstack/os-vif master: update os-vif ci to account for devstack default changes https://review.opendev.org/c/openstack/os-vif/+/798038 | |
| #openstack-nova - 2021-07-07 | |||
| 03:31:52 | opendevreview | wangzhenmeng proposed openstack/nova master: Three CPU parameters, mode, model and vendor_id, are added to flavor, some guest os do not support new CPU, so you need to specify a specific CPU when starting. https://review.opendev.org/c/openstack/nova/+/796986 | |
| 03:34:35 | opendevreview | wangzhenmeng proposed openstack/nova master: Add CPU parameters to flavor https://review.opendev.org/c/openstack/nova/+/796986 | |
| 09:24:55 | ralonsoh | hi folks, do you know what I need to set in libvirt for live-migration? | |
| 09:25:00 | ralonsoh | I'm having this error: | |
| 09:25:01 | ralonsoh | libvirt.libvirtError: Secret not found: no secret with matching uuid '70447a0f-443f-4b30-a154-c75706973e08' | |
| 09:25:38 | sean-k-mooney[m] | are you using ceph | |
| 09:25:49 | sean-k-mooney[m] | normally you do not need to set anything in libvirt | |
| 09:26:00 | ralonsoh | yes I configured ceph | |
| 09:26:07 | ralonsoh | I'll check it again | |
| 09:26:27 | ralonsoh | (maybe not in the second compute node...) | |
| 09:26:27 | sean-k-mooney[m] | you need to configure the ceph key ring as a secret but the devstack plugin does that for you | |
| 09:27:18 | ralonsoh | yeah, I think I didn't configure ceph in the second compute node | |
| 09:27:20 | ralonsoh | thanks! | |
| 09:28:17 | sean-k-mooney[m] | https://opendev.org/openstack/devstack-plugin-ceph/src/branch/master/devstack/lib/ceph#L244 | |
| 09:28:41 | sean-k-mooney[m] | that is what adds the user key ring as a secret | |
| 09:28:59 | sean-k-mooney[m] | so you could just do that or restack the compute | |
| 09:29:41 | sean-k-mooney[m] | you also need to copy over the ceph.confirm | |
| 09:29:49 | sean-k-mooney[m] | *ceph.conf | |
| 09:30:01 | ralonsoh | sean-k-mooney[m], btw, I think slaweq made a good point on https://review.opendev.org/c/openstack/neutron-specs/+/799198/3/specs/xena/ovn-auxiliary-port-bridge-live-migration.rst#51 | |
| 09:30:06 | ralonsoh | I'm checking this now | |
| 09:31:07 | sean-k-mooney[m] | ovn must handel that or we cannot support live migration with ovn | |
| 09:31:50 | ralonsoh | I know | |
| 09:33:34 | sean-k-mooney[m] | so from my point of view we should either ignore that and if it cause any issue file it as a bug with ovn or we disable live migration when you use ovn. we are not actually changing the behaviour although we are extending the time its in that state | |
| 09:33:52 | sean-k-mooney[m] | today the port exits on the source and dest for a period of time its just shorter | |
| 09:34:10 | ralonsoh | sean-k-mooney[m], in OVN we don't have, for example, the concept of multiple port binding | |
| 09:34:23 | ralonsoh | and could be difficult to introduce it | |
| 09:35:03 | ralonsoh | could be possible, just an idea, to monitor this multiple port binding | |
| 09:35:16 | ralonsoh | and provide routing for the older one | |
| 09:35:28 | ralonsoh | when the scr port is deleted, the destination port will prevail | |
| 09:35:41 | ralonsoh | this will also prevent errors if the migration fails | |
| 09:35:47 | ralonsoh | anyway, this is not supported | |
| 09:37:28 | sean-k-mooney[m] | this is not related to multiple port bindings | |
| 09:37:46 | ralonsoh | I was making a parallelism | |
| 09:37:46 | sean-k-mooney[m] | this willl happen with just libvirt and ovn without openstack | |
| 09:37:51 | ralonsoh | with OVN backend | |
| 09:38:12 | ralonsoh | this is what OVN should support or something similar | |
| 09:38:30 | sean-k-mooney[m] | ovn needs to monitor the packets | |
| 09:38:47 | ralonsoh | what for? | |
| 09:38:53 | sean-k-mooney[m] | and when it receives new packets on the dest host swap that to to the primary port | |
| 09:39:13 | ralonsoh | could be an idea yes | |
| 09:39:18 | sean-k-mooney[m] | basically if need to activate the dest port when it receives the mac learning RARP packets | |
| 09:39:25 | ralonsoh | but should be aware of the multiple port locations | |
| 09:39:32 | sean-k-mooney[m] | yes | |
| 09:39:42 | ralonsoh | in any case, this is very similar to what we have now | |
| 09:39:58 | sean-k-mooney[m] | anyway in post-copy mode the vm today is potentially in this state for several minutes | |
| 09:40:03 | ralonsoh | that means OVN will configure the backend when the VM is unpaused | |
| 09:40:07 | sean-k-mooney[m] | so this is not new to this spec | |
| 09:40:53 | ralonsoh | in any case, I didn't think about this problem | |
| 09:41:21 | ralonsoh | and now I don't see what's the benefit using this feature | |
| 09:41:38 | sean-k-mooney[m] | ya but its not related to any of the change we are making and it happens already today without them so while we should bring it up with the ovn team | |
| 09:41:48 | sean-k-mooney[m] | its not a blocker for this change | |
| 09:42:16 | ralonsoh | but it is a relevant OVN behavior change we need to document in the spec | |
| 09:42:26 | sean-k-mooney[m] | its not | |
| 09:42:32 | sean-k-mooney[m] | this already happens today | |
| 09:42:54 | sean-k-mooney[m] | so its not a behaviour change | |
| 09:43:01 | ralonsoh | right | |
| 09:43:20 | ralonsoh | in any case, to make this work properly, we need this OVN change | |
| 09:43:26 | ralonsoh | of course | |
| 09:43:32 | ralonsoh | the feature is still valid | |
| 09:43:35 | ralonsoh | and the spec | |
| 09:43:57 | sean-k-mooney[m] | perhaps. if ovn is installing teh correct flows on the dest in this sate | |
| 09:44:11 | sean-k-mooney[m] | then the only issue we have is with packets getting sent to the vm right | |
| 09:44:34 | ralonsoh | yes | |
| 09:44:42 | sean-k-mooney[m] | ovn proably should be mirroring the packet flow to both ports | |
| 09:45:09 | ralonsoh | let me ask again: how much time could it take to copy the VM to dest? | |
| 09:45:20 | sean-k-mooney[m] | multiple minutes | |
| 09:45:40 | ralonsoh | so we can have this network disconnection time, right? | |
| 09:46:00 | sean-k-mooney[m] | with ml2 ovs it wont have downtime | |
| 09:46:13 | sean-k-mooney[m] | the mac learning packets fix this issue | |
| 09:46:19 | sean-k-mooney[m] | i’m not sure how ovn will work | |
| 09:47:14 | ralonsoh | OVN will block any traffic from src port | |
| 09:47:25 | ralonsoh | because the port_binding is changed | |
| 09:47:44 | sean-k-mooney[m] | that is ok | |
| 09:48:10 | sean-k-mooney[m] | i assume it is blocking packets form the dest until the port binding is activated on the dest | |
| 09:48:49 | ralonsoh | do you have 5 mins via meet? | |
| 09:48:53 | ralonsoh | or this afternoon | |
| 09:49:00 | ralonsoh | whenever you can | |
| 09:49:14 | sean-k-mooney[m] | are you free at the top of the hour | |
| 09:49:25 | ralonsoh | yes | |
| 09:49:33 | sean-k-mooney[m] | i was just going to grab coffee but i’m free then | |
| 09:49:33 | ralonsoh | ping me then | |
| 09:49:37 | ralonsoh | perfect | |
| 09:56:20 | opendevreview | Lee Yarwood proposed openstack/nova master: zuul: Add CentOS 8 stream integrated compute tempest job to gate https://review.opendev.org/c/openstack/nova/+/797616 | |
| 10:03:23 | sean-k-mooney | ralonsoh: https:/meet.google.com/izr-tezp-qab whenever you are free | |
| 10:56:26 | opendevreview | Stephen Finucane proposed openstack/nova master: Remove the redundant code for attach interface https://review.opendev.org/c/openstack/nova/+/796208 | |
| 10:56:37 | opendevreview | Merged openstack/os-traits master: setup.cfg: Replace dashes with underscores https://review.opendev.org/c/openstack/os-traits/+/791133 | |
| 11:36:15 | viks_ | hi, suppose if i have a preconfigured physical system which already have an os installed, can i make it as a openstack instance so that i can control that machine... i.e. assign an ip via neutron, boot, shutdown, restart, inject ssh etc ? | |
| 11:40:30 | lyarwood | viks_: https://docs.openstack.org/ironic/latest/ | |
| 11:43:57 | viks_ | lyarwood: Thanks... but In ironic it installs OS on the given baremetal... but i do not need to install anything, just add a machine and control it... | |