| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-06-18 | |||
| 13:18:41 | hamdyk | yes, libvirt waits for event that plugging was successful | |
| 13:18:53 | sean-k-mooney | so we fully complete plug before we ever define the domain in libvirt | |
| 13:18:58 | yonglihe | mriedem: You have all context about this, It may easier for you than other cores. Hope you could have a time slot. Thanks, https://review.opendev.org/#/c/658716/ | |
| 13:19:06 | jangutter | sean-k-mooney: but on reboot? The domain is already defined? | |
| 13:19:32 | sean-k-mooney | oh hard reboot we destroy and redefine the domain | |
| 13:19:53 | sean-k-mooney | on soft reboot we do not unplug the vf for qemu or ovs | |
| 13:20:43 | hamdyk | so we are good with the patch? | |
| 13:20:58 | sean-k-mooney | not really | |
| 13:21:11 | sean-k-mooney | i would like to check with the ovs folks first | |
| 13:21:48 | sean-k-mooney | unless we can do it in a vendor neutral way i would prefer to do it in libvirt but im open to being conviced | |
| 13:22:00 | jangutter | sean-k-mooney, hamdyk: what happens on upgrade? This would cause an interface type=hostdev to morph to unmanaged PCI passthrough. | |
| 13:22:32 | sean-k-mooney | if the ovs or dpdk folks are open to adding a way to set the mac via ovs in the future the im oke with the patch | |
| 13:22:53 | sean-k-mooney | jangutter: only for this once case correct | |
| 13:23:02 | sean-k-mooney | this code path will not be taken by standard sriov | |
| 13:23:10 | sean-k-mooney | i have not read the nova patch yet | |
| 13:23:31 | sean-k-mooney | but while i want to add a generic sriov driver in os-vif i have not sat down an written it yet | |
| 13:23:35 | jangutter | sean-k-mooney, hamdyk: there's another key point here, OpenStack is not the only user of this type of plugging, it makes sense to put the common code in libvirt. | |
| 13:24:30 | jangutter | sean-k-mooney, hamdyk: If it's in Nova/os-vif, I'd really like to see documentation, (it's just me), so that the reference design prevents other people from making 100 other ways of doing the same. | |
| 13:25:04 | sean-k-mooney | i can live with it being in os-vif infact it solves some race condition that libvirt creates | |
| 13:25:29 | sean-k-mooney | jangutter: we shoudl avoid libvirt ever plugging vif if we can | |
| 13:25:55 | sean-k-mooney | it creates a race between neutron and qemu that cannot be fixed in libvirt by design | |
| 13:26:33 | jangutter | sean-k-mooney: I agree. However, this smells more like a "port-on-the-VM" option than a "port-on-the-Switch". | |
| 13:26:47 | sean-k-mooney | the mac address | |
| 13:27:10 | sean-k-mooney | yes perhaps but in that case we would be looking at a qemu change right to set the mac | |
| 13:27:38 | jangutter | sean-k-mooney: yep. The confusion comes in that a "let's configure the switch" thing is conflated with "let's configure the virtual thing connected to the switch". | |
| 13:27:45 | hamdyk | so, if OVS/DPDK guys are planning to set the VF MAC in the near future that would be great | |
| 13:27:48 | sean-k-mooney | libvirt is not going to embed a dpdk driver and i dont really want libvirt talking to ovs | |
| 13:28:03 | jangutter | sean-k-mooney: agree 100% | |
| 13:29:05 | sean-k-mooney | hamdyk: im going to ask them on #openvswitch but could you send an email to the openvswich mailing list and ask? im not currently subsribed to it | |
| 13:29:40 | hamdyk | so we need to take care of the edge cases (upgrade, soft reboot if it is a problem) | |
| 13:29:41 | jangutter | sean-k-mooney: I'm definitely not opposed to this with -2, (and I'm definitely not the official gatekeeper for this decision either). | |
| 13:30:11 | sean-k-mooney | hamdyk: soft reboot is fine. | |
| 13:30:32 | sean-k-mooney | hamdyk: upgrades will not modify the xml until the guest reboots | |
| 13:30:39 | hamdyk | sean-k-mooney: will do send the email | |
| 13:31:46 | sean-k-mooney | if they come back and say yes we can add that feature then im ok with proceeding with your patch as a workaound in the short term if we also add a todo to consume the ovs/dpdk feature when available | |
| 13:32:01 | hamdyk | sean-k-mooney: so after the guest reboots, os-vif will do the plugging again then setting the MAC, correct ? | |
| 13:32:03 | sean-k-mooney | hamdyk: i need to revew the nova patch too however | |
| 13:32:28 | sean-k-mooney | hamdyk: on soft reboot no but it does not need too on hard reboot yes | |
| 13:32:54 | hamdyk | I see | |
| 13:35:55 | sean-k-mooney | soft reboot we just send a achi command to the guest os and dont teardown the vm | |
| 13:36:51 | sean-k-mooney | hard reboot we kill teh runnign vm. unplug all the ports and disk redeine the vm and plug all the port/disk/devices again from scratch as if it was a first boot | |
| 13:37:32 | jangutter | hamdyk: You could split out the _set_config_VIFHostDevice section from https://review.opendev.org/#/c/658785 | |
| 13:38:31 | jangutter | hamdyk: That should be pretty non-controversial. | |
| 13:55:22 | openstackgerrit | Artom Lifshitz proposed openstack/nova master: Revert resize: wait for events according to hybrid plug https://review.opendev.org/644881 | |
| 13:55:23 | openstackgerrit | Artom Lifshitz proposed openstack/nova master: [DNM] testing bug/1813789 revert resize events https://review.opendev.org/664442 | |
| 13:55:23 | openstackgerrit | Artom Lifshitz proposed openstack/nova master: WIP: Include direct-physical in compute manager events check https://review.opendev.org/664431 | |
| 13:55:59 | openstackgerrit | Josephine Seifert proposed openstack/nova-specs master: Spec for the Nova part of Image Encryption https://review.opendev.org/608696 | |
| 14:07:31 | hamdyk | jangutter: will do | |
| 14:35:40 | mdbooth | lyarwood: LGTM, but missing unit test I think. | |
| 14:38:19 | ohwhyosa | Hello peoples! | |
| 14:38:37 | ohwhyosa | What is the best way to change the keymap in spice / novnc per instance? | |
| 14:38:56 | ohwhyosa | I saw the keymap param in nova.conf and I saw also that it's deprecated | |
| 14:39:05 | ohwhyosa | But that is more global | |
| 14:39:33 | ohwhyosa | Say I have a french person, a german person and an american person accesing their instances, surely they can change the keympa? | |
| 14:40:06 | sean-k-mooney | ohwhyosa: the best way is for them to do it in the os when they log in | |
| 14:40:34 | lyarwood | mdbooth: ack thanks | |
| 14:40:40 | lyarwood | is gerrit dying for anyone else? | |
| 14:40:50 | sean-k-mooney | ohwhyosa: that or they can try and set it in there spice client | |
| 14:40:58 | ohwhyosa | sean-k-mooney, But then, for example, the cirros pass with a :) at the end becomes complicated, and if the only way they had of accessing the machine was via console they'd be screweeeed | |
| 14:41:04 | ohwhyosa | sean-k-mooney, Thanks anyway! | |
| 14:41:24 | ohwhyosa | Hmmm sean-k-mooney the spice client, the web one? | |
| 14:41:34 | sean-k-mooney | ohwhyosa: no a desktop one | |
| 14:42:01 | sean-k-mooney | you can install a spice client locally and use that to conect to the vm directly | |
| 14:42:21 | sean-k-mooney | that client should be able to set the keymap i think | |
| 14:42:48 | ohwhyosa | sean-k-mooney, great! thanks! | |
| 14:46:48 | sean-k-mooney | stephenfin: you worked on ^ a bit do we have any documenation that could help | |
| 14:48:57 | artom | sean-k-mooney, errr, don't think support anything besides the webclient for SPICE | |
| 14:49:03 | artom | Because of the websocket proxy thing | |
| 14:49:18 | artom | The native client would need to understand websockets | |
| 14:50:23 | sean-k-mooney | im not sure about that i think you can get the direct socket url via the api | |
| 14:50:51 | artom | IIRC you get the websocket HTTP URL | |
| 14:50:55 | sean-k-mooney | but i think you do need to configure you hosts to support it | |
| 14:51:15 | stephenfin | sean-k-mooney: afaik it should be configurable via the VNC/Spice client, if said client supports such a thing | |
| 14:51:20 | sean-k-mooney | maybe im not sure i have used it with novnc in the past | |
| 14:51:25 | stephenfin | noVNC does since version 1.0, I think | |
| 14:52:35 | sean-k-mooney | artom: i would not be surprised if there are spice clients tha tsupprot websocket streams too by they way | |
| 14:53:30 | artom | sean-k-mooney, maybe, but back when I tried https://review.opendev.org/#/c/442040/ there weren't | |
| 14:54:58 | sean-k-mooney | oh ok. is that just a limitation for spice? im pretty sure i have used remote viewer and other clinets to connect to no vnc in the past | |
| 14:55:15 | artom | All consoles, I think | |
| 14:55:37 | artom | Because you can't get to the compute host directly, you need to go through a proxy on the publicly accessible controller | |
| 14:55:54 | artom | And all of our proxies use websockets | |
| 14:55:55 | sean-k-mooney | i could have been using virt-manager to connect via the host... | |
| 14:56:06 | artom | To be able to authenticate the connection first | |
| 14:56:29 | sean-k-mooney | well that the thing i thought we supported directly connecting to the host without the proxy and that was required for non web clients | |
| 14:56:44 | sean-k-mooney | anyway got to grab coffee before a meeting | |
| 15:03:09 | ohwhyosa | thanks sean-k-mooney artom stephenfin !! | |
| 15:12:26 | Nick_A | We are trying to inject passwords and keys with libguestfs-tools. keys work fine but not passwords. tried with config drive on and off. any ideas? | |
| 15:38:07 | openstackgerrit | Lee Yarwood proposed openstack/nova master: libvirt: flatten rbd images when unshelving an instance https://review.opendev.org/457886 | |
| 15:46:52 | Nick_A | virt-customize works, so it seems libguestfs isn't the issue | |
| 15:52:34 | openstackgerrit | Stephen Finucane proposed openstack/nova-specs master: Additional upgrade clarifications for cpu-resources https://review.opendev.org/666032 | |
| 16:06:25 | dansmith | mriedem: can you circle back to this soonishly? https://review.opendev.org/#/c/644881/ | |
| 16:08:56 | mriedem | https://www.youtube.com/watch?v=yKNQzFxV-qg | |
| 16:09:02 | mriedem | mm | |
| 16:09:03 | mriedem | buh | |
| 16:09:04 | mriedem | snuf | |
| 16:09:57 | dansmith | https://www.youtube.com/watch?v=9MOAhnQbC6s | |
| 16:10:17 | mriedem | NSFW | |
| 16:21:23 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Fix GET /servers/detail host_status performance regression https://review.opendev.org/666042 | |
| 16:24:38 | Nick_A | Is this the place to ask about nova-lxd? | |
| 16:24:44 | dansmith | no. | |