| *** ykarel__ is now known as ykarel | 03:20 | |
| opendevreview | melanie witt proposed openstack/nova master: libvirt: Add conf option for live migration monitor interval https://review.opendev.org/c/openstack/nova/+/937223 | 03:53 |
|---|---|---|
| opendevreview | melanie witt proposed openstack/nova master: libvirt: Add conf option for live migration monitor interval https://review.opendev.org/c/openstack/nova/+/937223 | 03:57 |
| opendevreview | Shalini Srivastava proposed openstack/nova stable/2025.2: Filter invalid kwargs in volume_api.create to prevent TypeError https://review.opendev.org/c/openstack/nova/+/992849 | 06:50 |
| opendevreview | Shalini Srivastava proposed openstack/nova stable/2025.2: Filter invalid kwargs in volume_api.create to prevent TypeError https://review.opendev.org/c/openstack/nova/+/992849 | 07:38 |
| lajoskatona | sean-k-mooney, sean-k-mooney[m]: Hi, for the bug https://bugs.launchpad.net/nova/+bug/1854844 (libvirt: tx/rx queue length and max queues are not updated on live migration) | 07:58 |
| lajoskatona | sean-k-mooney, sean-k-mooney[m]: I pushed a reproducer: https://review.opendev.org/c/openstack/nova/+/991772 as a first step | 07:58 |
| lajoskatona | I plan to push patch on top of it (I hope this week, but who knows....) | 07:59 |
| sean-k-mooney[m] | lajoskatona: its technically not a bug | 08:04 |
| sean-k-mooney[m] | or rather it was a known issue with using config optins to contol it. | 08:05 |
| sean-k-mooney[m] | also in older version of qemu you could not change it during live migraton | 08:05 |
| sean-k-mooney[m] | the same way you cant always change the mtu | 08:06 |
| sean-k-mooney[m] | you can update the mtu if the mtu element is already but adding/removing ti change the vhost-user feature flags form 32 bit format to 64 bit format whic you canoot do during a live migration | 08:07 |
| sean-k-mooney[m] | lajoskatona: with all that said ya that is someitng i wold liek to fix if tis somethign we can fix today | 08:09 |
| sean-k-mooney[m] | the fix requires use to pass the queue lenghts back from the dest to the source | 08:09 |
| sean-k-mooney[m] | which requires adding new fileds to the live migration data object or some other hacks | 08:10 |
| sean-k-mooney[m] | thanks for creating the repoducer | 08:10 |
| sean-k-mooney[m] | do you plan on creating a fix? | 08:10 |
| sean-k-mooney[m] | lajoskatona: just to be clear changign the tx/rx quee lenaght is changing the emulated hardware to report a diffent quene lenght which is not somehting real hardware would do. today what happens is you live migrate with the old value and they update the next time you reboot. the nuber of ques is also not really somethign that should change as the number fo tx/rx queue on a real nic wont magicly change at runtime like that | 08:15 |
| sean-k-mooney[m] | either. | 08:15 |
| sean-k-mooney[m] | so ideally we would schdule to a host that supprot hte same number and lenght of queues as the real fix | 08:15 |
| sean-k-mooney[m] | rather then adjusgting the geust hardware live | 08:16 |
| sean-k-mooney[m] | lajoskatona: assuming the plan is not to add the number of queues and there lenght if not alreay in the xml like we did for mtu that would be a good firsst step. | 08:25 |
| sean-k-mooney[m] | eventually we should also schdule on this in some way but the first step would at least help with the upgrade impact | 08:26 |
| lajoskatona | sean-k-mooney[m]: yes I plan to push a fix also, thanks for the details, I consider them before pushing anything | 08:26 |
| sean-k-mooney[m] | we also dont want to remove them from the xml if tis already there | 08:26 |
| sean-k-mooney[m] | no worrise im glad some one has had time to actully look at fixing it | 08:26 |
| sean-k-mooney[m] | i treid to explain in the bug that i did carea bout fixing it but just didnt have the time to do it | 08:27 |
| sean-k-mooney[m] | so im happy to review | 08:27 |
| sean-k-mooney[m] | ill also admit i forgot this was still a bug | 08:28 |
| opendevreview | Ashish Gupta proposed openstack/nova master: [DNM][Experimental] tests: Use per-database write locks instead of global lock https://review.opendev.org/c/openstack/nova/+/992862 | 08:45 |
| cw0306-lee[m] | Hi, gibi! I added comment on review.opendev.org/c/openstack/nova-specs/+/991636 | 09:25 |
| cw0306-lee[m] | Could you please check it? | 09:25 |
| opendevreview | Kamil Sambor proposed openstack/nova master: Change nova-alt-configurations job https://review.opendev.org/c/openstack/nova/+/983179 | 09:25 |
| opendevreview | Kamil Sambor proposed openstack/nova master: Restore 'fork' start method for daemon mode on Python 3.14 https://review.opendev.org/c/openstack/nova/+/987815 | 09:25 |
| opendevreview | Kamil Sambor proposed openstack/nova master: [dnm] testing python 3.14 with restored fork https://review.opendev.org/c/openstack/nova/+/987821 | 09:26 |
| cw0306-lee[m] | * Hi, | 09:26 |
| cw0306-lee[m] | @_oftc_gibi:matrix.org! I added comment on https://review.opendev.org/c/openstack/nova-specs/+/991636 Could you please check it? | 09:26 |
| sean-k-mooney | how do we feel about https://review.opendev.org/c/openstack/nova-specs/+/471815, we are at spec freeze today and i would really like to finally see that adance | 09:41 |
| sean-k-mooney | the implemtion might need some addtional docs and or test but it exists https://review.opendev.org/c/openstack/nova/+/941227 | 09:42 |
| sean-k-mooney | so to me this is small and worth doing | 09:43 |
| seyeongkim | sean-k-mooney: Hi. I've addressed your comments on https://review.opendev.org/c/openstack/nova/+/963665 and also updated the LP bug with more details on the use case — when you have a moment, could you take another look and let me know if anything else is needed? Thanks! | 10:00 |
| sean-k-mooney | seyeongkim: sure but again this is really a feature nto a bug | 10:05 |
| sean-k-mooney | we have never offcially supproted the usecase your trying to enable | 10:06 |
| opendevreview | Seyeong Kim proposed openstack/nova master: libvirt: Support boot_index for multiple block devices https://review.opendev.org/c/openstack/nova/+/963665 | 10:10 |
| opendevreview | Joan Gilabert proposed openstack/nova master: Add mtty support to nova-next https://review.opendev.org/c/openstack/nova/+/922140 | 10:16 |
| opendevreview | Konrad Gube proposed openstack/nova-specs master: Propose assisted-volume-extend API https://review.opendev.org/c/openstack/nova-specs/+/949504 | 10:24 |
| seyeongkim | sean-k-mooney: Thanks Sean, and fair point on it being closer to a feature. Could I ask you to take one more detailed look though? My thinking is mostly from the user's side: once someone sets boot_index on a volume, it feels reasonable to expect that device to at least show up to the grub, rather than being silently skipped. Right now Nova accepts the boot_index but the extra disks never get a per-device boot order, so SeaBIOS skips them entire | 10:26 |
| sean-k-mooney | i jsut did i have not -2'd it but i dont think this is approate to treat as a bug | 10:28 |
| sean-k-mooney | seyeongkim: if you removed all the logic related to attch/detach | 10:28 |
| sean-k-mooney | it might eb accpable to set the index staticly via server create | 10:28 |
| sean-k-mooney | that is the only thing the api ahs ever allowed | 10:29 |
| sean-k-mooney | but it was documetned as driver spcric | 10:29 |
| sean-k-mooney | and im not sure this was ever supproted in the libvirt driver | 10:29 |
| sean-k-mooney | i think that was only ever properly suprpoted in xen or one of the other drivers | 10:29 |
| sean-k-mooney | ah found it it was implemt by the hyperv driver | 10:33 |
| sean-k-mooney | https://github.com/openstack/nova/commit/d68c04299aa2c04aea16e881d93076236cc64d7b | 10:34 |
| sean-k-mooney | seyeongkim: so this wa never supprot by libvirt, vmware, xen or powervm | 10:38 |
| seyeongkim | sean-k-mooney: Thanks for the review. Our customer relies on a multiple-disk boot scenario that worked before this SeaBIOS change (https://github.com/coreboot/seabios/commit/2f4d068645c211e309812372cd0ac58c9024e93b), and I haven't been able to find a workaround yet(it turned out SATA is not viable workaround although I mentioned it in review comment). Since libvirt provide options that per-disk boot order in the XML, I thought this Nova patch co | 10:41 |
| sean-k-mooney | sure but nova does nto provide the libvirt api | 10:46 |
| seyeongkim | right thank you so much! | 10:47 |
| sean-k-mooney | you cant assume that jsut ebcause libvirt supprot soemthign that nova supprot it even if it worked by acindent before | 10:47 |
| seyeongkim | agreed | 10:48 |
| cw0306-lee[m] | sean-k-mooney @_oftc_sean-k-mooney:matrix.org: Thank you for comment. I replied to it. | 10:52 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!