Thursday, 2026-04-23

opendevreviewmelanie witt proposed openstack/nova master: DNM test ceph ipv6  https://review.opendev.org/c/openstack/nova/+/98462300:58
opendevreviewTakashi Kajinami proposed openstack/nova-specs master: libvirt: AMD SEV-SNP support  https://review.opendev.org/c/openstack/nova-specs/+/98337602:09
dougszusean-k-mooney - morning, the bug about ephemeral volume corruption during VM migration (https://bugs.launchpad.net/nova/+bug/2095173) has that been fixed in Nova by dropping the creation of a filesystem in the base image? 09:01
lajoskatonaUggla: Hi, today I have a topic for os-external-events vs Neutron notifier vs SDK (https://etherpad.opendev.org/p/nova-2026.2-ptg#L771), shall I ask to have it discussed before 1530UTC (1730 CEST) as I have to drop around that time ?12:12
UgglaHi lajoskatona, it should happen around 1400 / 1430UTC. So yes.12:16
lajoskatonaUggla: thanks12:19
jkulikHow come n-api and n-scheduler require a cell DB? Is it just to be able to register as a compute service?12:53
sean-k-mooneyjkulik: its used for 2 thing12:59
sean-k-mooneyso cell 0 is where all the service entries for the global serviece live i..e the schdluer/api ecta13:00
jkulikah. cell0! thanks, I thought I'd have to use a productive cell13:00
sean-k-mooneyits also used to sotre instance that cannot be schduled ot "real" "comptue" cells13:00
sean-k-mooneyjkulik: ya no you put those in cell 013:00
jkulikthank you. helped a lot13:01
sean-k-mooneyright os the api need a plce to stre the full instnace object for an instance that failed to go to a real cell so we just use that for the service records as well13:01
jkulikyeah. makes sense. thanks again13:01
sean-k-mooneyi dont think it woudl breake anything to have the scuder use the cell1 db isntead but it buys you nothing either because you still need cell0 and the schduelr does not work per cell so we recommend cell0 because it alwasy extis and it shoudl be under less load13:02
sean-k-mooneyjkulik: no worries13:02
opendevreviewLajos Katona proposed openstack/nova master: Use SDK for Neutron security-groups  https://review.opendev.org/c/openstack/nova/+/98114113:47
opendevreviewTakashi Kajinami proposed openstack/nova-specs master: libvirt: AMD SEV-SNP support  https://review.opendev.org/c/openstack/nova-specs/+/98337613:55
opendevreviewAnton Iacobaeus proposed openstack/nova-specs master: Intel TDX support in libvirt driver  https://review.opendev.org/c/openstack/nova-specs/+/97960814:35
bauzasUggla: planning any break after the current discussion ? I'm afraid to say I'm a little fried and I need coffee14:58
UgglaYes I'd like to have a short 5mn break if TheJulia and dansmith are ok.14:58
TheJuliafine with me15:01
UgglaTheJulia thx15:06
tkajinamjust in case someone can answer this question ... In case we remove a trait from a compute host RP, it does not care any existing instance requires that trait, right ?15:07
TheJuliaThat is an interesting question and relates to the next topic :)15:09
dansmithtkajinam: no, nothing will break15:10
dansmithmeaning, there's no checking like that, nor any implication that traits now mean traits in the future or past15:10
TheJuliaThat is what I was thinking based upon what was in my recall of nova code, because its far past the use of scheduling15:10
tkajinamok15:10
tkajinamand that's what I thought15:11
opendevreviewStephen Finucane proposed openstack/os-vif master: pre-commit: Bump versions  https://review.opendev.org/c/openstack/os-vif/+/97635816:23
opendevreviewStephen Finucane proposed openstack/os-vif master: Migrate setup configuration to pyproject.toml  https://review.opendev.org/c/openstack/os-vif/+/97635916:23
opendevreviewStephen Finucane proposed openstack/os-vif master: tests: Remove errant parameters  https://review.opendev.org/c/openstack/os-vif/+/97636016:23
opendevreviewStephen Finucane proposed openstack/os-vif master: trivial: Use bit shifting  https://review.opendev.org/c/openstack/os-vif/+/97636116:23
opendevreviewStephen Finucane proposed openstack/os-vif master: typing: Add initial type annotations  https://review.opendev.org/c/openstack/os-vif/+/74516216:23
opendevreviewStephen Finucane proposed openstack/os-vif master: typing: Add typing to vif_plug_noop module  https://review.opendev.org/c/openstack/os-vif/+/97636216:23
opendevreviewStephen Finucane proposed openstack/os-vif master: WIP: typing: Add typing to vif_plug_ovs module  https://review.opendev.org/c/openstack/os-vif/+/97636316:23
opendevreviewDan Smith proposed openstack/python-novaclient master: Remove deprecation warning workaround  https://review.opendev.org/c/openstack/python-novaclient/+/98599318:21
sean-k-mooneymelwitt: so picking up our conversation form yestereday  https://zuul.opendev.org/t/openstack/build/a5324089fe45434ca8ed15a8de9e12a3 the failure in that job is weried.19:25
melwittyeah I have seen that before and I never understand how the thing doesn't have the 'openstack' command19:26
sean-k-mooneyif i look at the local.conf the SERVICE_HOST didnt make it to the local.conf https://zuul.opendev.org/t/openstack/build/a5324089fe45434ca8ed15a8de9e12a3/log/controller/logs/local_conf.txt19:26
melwittthe last revision of the ipv6 devstack stuff I think messed up the expected format for an apache config for the listen address so I'm trying to figure that out19:27
melwitthm19:27
sean-k-mooneythat form world dbump if we fail really early i.e. beofre devstack has run and installed it19:27
sean-k-mooneyack so looking at the logs we dont have a stack log19:28
sean-k-mooneyso we failed super early in the job19:28
melwittoh maybe it's expected it wouldn't be in local.conf bc it's under devstack_localrc not devstack_local_conf19:28
melwittit's been comical how difficult getting the f'ing ipv6 url format right for the different parts has been19:29
sean-k-mooneyits also not in https://zuul.opendev.org/t/openstack/build/a5324089fe45434ca8ed15a8de9e12a3/log/controller/logs/_.localrc_auto.txt19:29
melwittsome need the brackets or it blows up, some have to not have the brackets or it blows up. I'm really not liking whoever invented brackets19:29
sean-k-mooneylol you might be cursed.19:30
sean-k-mooneyhttps://zuul.opendev.org/t/openstack/build/a5324089fe45434ca8ed15a8de9e12a3/log/zuul-info/zuul-info.controller.txt19:30
melwittprobably19:31
sean-k-mooneyi dont know if that will break it but that host does nto have a publicly reoutable ipv6 adresss19:31
sean-k-mooneyi guess link local ipv6 could work19:31
sean-k-mooneybut i dont know if that will show up in teh default ipv6 varibale19:31
sean-k-mooneySERVICE_HOST: '[{{hostvars[''controller''][''ansible_default_ipv6''][''address'']}}]'19:32
melwittyeah ... I wonder is this a futile endeavor? how can we make sure our zuul nodes have routable ipv6 addresses?19:32
sean-k-mooneyalthough HOST_IP: '{{ hostvars[inventory_hostname][''nodepool''][''private_ipv4'']19:32
sean-k-mooney          }}'19:32
sean-k-mooneyis the host ip var so the double '' i guess is expected19:32
melwittansible_default_ipv6: {}19:33
melwittin https://zuul.opendev.org/t/openstack/build/a5324089fe45434ca8ed15a8de9e12a3/log/zuul-info/host-info.controller.yaml19:33
sean-k-mooneyhttps://zuul.opendev.org/t/openstack/build/a5324089fe45434ca8ed15a8de9e12a3/log/zuul-info/host-info.controller.yaml#21419:34
melwittI was just looking at that19:34
sean-k-mooneyso we could get the first ipv6 adress if we knw the interface name19:34
melwittand that shows up in ansible_all_ipv6_addresses19:35
sean-k-mooneyoh it does19:35
sean-k-mooneyand ansible_all_ipv6_addresses does not inlcude loopback19:35
sean-k-mooneyok so maybe we shoudl use that instead jsut take the first one19:35
* melwitt nods19:36
sean-k-mooney"{{ ansible_all_ipv6_addresses | sort | first }}"19:41
sean-k-mooneywell that need the hostvars bit too19:41
sean-k-mooneygemini gave me a rmore corerct slolution that is robust 19:41
sean-k-mooneybut link lofal adresses start with fe which is after public unicase adreses that start with 2000 prefxi19:42
melwittI did not know that19:43
sean-k-mooneyin knew fe80 was link local19:43
sean-k-mooneybut not the other prefix ranges19:43
sean-k-mooney2000::/3 Global Unicast19:43
sean-k-mooneyfe80::/10 Link-Local19:43
sean-k-mooneyfc00::/7 Unique Local (ULA)19:44
sean-k-mooney::1/128 Loopback19:44
sean-k-mooneythere are probaly other for mutli cast and such but sorting will put the public adress first if there is one19:44
sean-k-mooneythe correct way to do this is apprently 19:45
sean-k-mooney{{ 19:45
sean-k-mooney        ansible_all_ipv6_addresses | select('ansible.utils.in_network', '2000::/3') | first 19:45
sean-k-mooney        | default(ansible_all_ipv6_addresses | reject('match', '^::1') | first) 19:45
sean-k-mooney      }}19:45
sean-k-mooneywhich makes sense sort of but it way mor ecomplciate then it need to be19:45
melwittO_o19:46
sean-k-mooneySERVICE_HOST: "[{{ (hostvars['controller']['ansible_all_ipv6_addresses'] | sort | first) }}]"19:48
sean-k-mooneymight work19:49
sean-k-mooneyim tempted to just put clause in a vm with ipv6 and tell it to figure out how to make it work with ipv619:50
sean-k-mooney*claude19:50
melwittyou could try haha. I've been using it to make the WIP devstack changes and it has gone on tangents changing things that didn't need to, so ymmv19:51
sean-k-mooneythe actual failure in devstack was that it promted for the db password by the way19:52
sean-k-mooneyand it did that because it didnt proplry create the local.conf19:53
melwitthah19:53
sean-k-mooneypresumabel becasue it was unhappy with the way we set the ip19:53
sean-k-mooneyok im going to go make dinner but i might spin up  1 or 2 vms locally and try this out but maybe next week19:58
melwittok, thanks for the help o/19:59
opendevreviewDan Smith proposed openstack/python-novaclient master: Remove deprecation warning workaround  https://review.opendev.org/c/openstack/python-novaclient/+/98599320:13
opendevreviewmelanie witt proposed openstack/nova master: DNM test ceph ipv6  https://review.opendev.org/c/openstack/nova/+/98462322:24
gmaanUggla: just realized that 14 UTC tomorrow is board sessions for agentic AI which overalap with service-to-service communication, can we keep that at 13 UTC in original time ?22:50
gmaanUggla: or if we are keeping the nova-reviewers talks at 13 UTC then right after that 13.30 UTC also works fine22:53

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!