opendevreview | Wenping Song proposed openstack/nova-specs master: Usage of new trait of OWNER_NOVA https://review.opendev.org/c/openstack/nova-specs/+/836583 | 01:04 |
---|---|---|
whoami-rajat | bauzas, hey, we've a topic by sean-k-mooney regarding os-brick rootwrap config , is there a desired time when nova-cinder could meet up? our PTG is conducted 1300-1700 UTC | 05:45 |
*** amoralej|off is now known as amoralej | 07:08 | |
bauzas | whoami-rajat: we can discuss after 14:30UTC I guess, we have a nova-neutron x-p discussion for only one topic at 2pm | 07:10 |
gibi | morning | 07:45 |
bauzas | opening the nova-cyborg room | 07:58 |
bauzas | songwenping__: are you around ? | 08:02 |
songwenping__ | bauzas: yes | 08:03 |
bauzas | can you go on https://www.openstack.org/ptg/rooms/newton ? | 08:04 |
bauzas | songwenping__: ^ | 08:04 |
songwenping__ | let me try. | 08:05 |
songwenping__ | bauzas: it's still joining on the website, :( | 08:09 |
bauzas | songwenping__: ok, are you able to join with the webapp, maybe ? | 08:10 |
bauzas | you won't then need to install any zoom app | 08:10 |
songwenping__ | i haven't install the app | 08:10 |
opendevreview | Takashi Natsume proposed openstack/python-novaclient master: Update pre-commit hook, hacking versions https://review.opendev.org/c/openstack/python-novaclient/+/836768 | 08:53 |
Uggla | bauzas, only Neutron/nova discussion and the virtiofs point remaining for today ? | 09:03 |
bauzas | Uggla: and maybe a Cinder discussion too | 09:04 |
bauzas | I got asked by whoami-rajat | 09:04 |
Uggla | bauzas, yep I have just seen ^ | 09:04 |
whoami-rajat | bauzas, I've planned the session for 1430 L#520 https://etherpad.opendev.org/p/zed-ptg-cinder | 09:11 |
bauzas | whoami-rajat: a bit of warning tho, the session could be a bit late if we can't wrap the neutron discussion after 30 mins | 09:12 |
whoami-rajat | bauzas, hmm, how about 1500 then? | 09:13 |
bauzas | whoami-rajat: let's do it | 09:13 |
bauzas | whoami-rajat: if we're quickier than 1 hour, I have one topic to discuss from Uggla that's left | 09:13 |
whoami-rajat | bauzas, ack, 1500 makes sense then | 09:14 |
whoami-rajat | sean-k-mooney, ^^ | 09:14 |
whoami-rajat | bauzas, I assume the nova team will join the cinder meeting? | 09:15 |
bauzas | yup | 09:16 |
bauzas | whoami-rajat: this makes sense, not all nova contributors could be interested in this topic and that's more a cinder stuff | 09:16 |
whoami-rajat | yeah, that's my understanding as well | 09:18 |
whoami-rajat | here's the meeting link: meeting link https://bluejeans.com/556681290 | 09:18 |
whoami-rajat | but i will share again during the discussed time | 09:18 |
bauzas | whoami-rajat: already setup | 09:19 |
bauzas | I updated our etherpad | 09:19 |
whoami-rajat | great | 09:19 |
sean-k-mooney | bauzas: i have one slight followup form arbiterd | 09:56 |
sean-k-mooney | bauzas: basically i wanted to ask about hosting it in opendev and or making it a deliverable fo the compute project | 09:57 |
sean-k-mooney | i.e. if i move it shoudl it be openstack/arbiterd or x/arbiterd the delta woudl be is it under offical governace and a deliverable of the compute team or not. or i could keep it hosted on my github for now where i already have some ci in place for the time being until we have somethign working end to end with openstack and then move it after | 09:59 |
sean-k-mooney | bauzas: we dont need to make a desission on that today | 10:00 |
sean-k-mooney | just wanted to know how other feel as there is a non zero amount of wokr i would have to do to make it pti complient amoung other things for move it to openstack/arbiterd even though that is where i eventually would like to see it be in a year or so | 10:02 |
kashyap | stephenfin: sean-k-mooney: gibi: Don't look at the newer CPU APIs patch yet, I've accidentally undid some changes during rebase. (I'll also splitting the patches once I post the monolithic patch with rebase fixed.) | 10:04 |
bauzas | sean-k-mooney: ack, we'll see then | 10:05 |
sean-k-mooney | ack | 10:05 |
* bauzas goes afk for gym | 10:05 | |
sean-k-mooney | i likely wont look until monday but ill try to review it then | 10:05 |
sean-k-mooney | if there is a section in partcalar you want me to look at before then jsut ping me | 10:06 |
kashyap | Yep | 10:06 |
gibi | kashyap: ack | 10:12 |
*** bhagyashris_ is now known as bhagyashris | 10:19 | |
*** dasm|off is now known as dasm | 11:16 | |
opendevreview | Merged openstack/python-novaclient master: Update pre-commit hook, hacking versions https://review.opendev.org/c/openstack/python-novaclient/+/836768 | 12:14 |
*** amoralej is now known as amoralej|lunch | 12:41 | |
Uggla | Quick question: anyway to get and see libvirt driver capabilities on a running devstack ? | 13:05 |
sean-k-mooney | am they get converted to traits in placment | 13:08 |
sean-k-mooney | i think you could also see them in the db | 13:09 |
sean-k-mooney | likely on the compute node record in the cell1 db in your case | 13:09 |
Uggla | sean-k-mooney, I'm gonna have a look | 13:11 |
Uggla | sean-k-mooney, I find my traits in placement. | 13:22 |
Uggla | sean-k-mooney, but not in cell1, I may not look at the right place | 13:23 |
*** amoralej|lunch is now known as amoralej | 13:24 | |
sean-k-mooney | it might just be in memory i belive the resouce tracker has this info in the host state object | 13:24 |
sean-k-mooney | which its usign for schduling | 13:25 |
*** haleyb_ is now known as haleyb | 13:25 | |
sean-k-mooney | its not in compute nodes and i dont see any other table that would hold it | 13:27 |
sean-k-mooney | no | 13:28 |
sean-k-mooney | it looks like its just in memory in the driver | 13:28 |
sean-k-mooney | and we are usign the traits for schduleing | 13:29 |
sean-k-mooney | oh | 13:30 |
sean-k-mooney | https://github.com/openstack/nova/blob/master/nova/scheduler/filters/compute_capabilities_filter.py#L34-L66= | 13:30 |
sean-k-mooney | so its using the stats field | 13:30 |
*** artom__ is now known as artom | 13:31 | |
Uggla | sean-k-mooney, https://paste.openstack.org/show/bj273VR3I4dh787TCSV6/ is it the capabilities my compute have currently ? | 13:32 |
artom | Uggla, we do `self.driver.capabilities.get('blah', False) in a bunch of places, is that what you mean? | 13:32 |
sean-k-mooney | Uggla: not just capablities | 13:33 |
sean-k-mooney | teh traits that are reported are a combanition of a few different things | 13:33 |
Uggla | artom, not really I just would like to know if there is a easy way to display the capabilities. Just to make sure that COMPUTE_STORAGE_VIRTIO_FS is available on my compute. | 13:34 |
sean-k-mooney | no there is not | 13:35 |
sean-k-mooney | they are only stored in memmory i think | 13:35 |
artom | Uggla, as a user (vs Python code)? Then yeah, you want to look at the traits of the corresponding resource provider | 13:35 |
artom | sean-k-mooney, wait, what? | 13:35 |
artom | Oh, is the capability not converted into a trait (yet?) | 13:35 |
sean-k-mooney | in this case likely not sicne we have not merged the standard trait | 13:36 |
sean-k-mooney | so placement shoudl reject it unless Uggla has used their modified version | 13:36 |
Uggla | yep I use my modified version | 13:36 |
Uggla | placement seems happy with it. | 13:37 |
sean-k-mooney | in which case you shoudl see it in the placment output | 13:37 |
sean-k-mooney | then ya that is the best you can do | 13:37 |
Uggla | placement output is log ? | 13:37 |
sean-k-mooney | the capabliteis are only sotred in memory from what i can see in the dirver | 13:37 |
sean-k-mooney | last line of https://paste.openstack.org/show/bj273VR3I4dh787TCSV6/ | 13:37 |
sean-k-mooney | but you can also get that from the api | 13:38 |
sean-k-mooney | using openstack client if you need too with the osc-placement plugin | 13:38 |
sean-k-mooney | anyway your code is correctly reporting to placment that the host ia capable of supportign virtio fs | 13:39 |
Uggla | https://paste.openstack.org/show/b9aK9NbdS1ja4hxwmP8E/ | 13:43 |
Uggla | from osc | 13:43 |
artom | Yeah, looks good, congrats :) | 13:47 |
Uggla | I'm gonna try to activate fileback memory | 13:48 |
Uggla | to see if it is reported... | 13:48 |
Uggla | sean-k-mooney, is it normal to have COMPUTE_ACCELERATORS reported ? | 13:51 |
sean-k-mooney | yes | 13:54 |
sean-k-mooney | the traits are what nova can do not what the host can do | 13:54 |
sean-k-mooney | COMPUTE traits are not about hardware they are about nova/libvirts capablities | 13:54 |
artom | ... which is sometimes about hardware ;) | 13:56 |
elodilles | heads up from me: i can't attend the nova session today (at least the 1st hour or so) as i need to be at the release mgmt session o:) | 14:00 |
bauzas | nova-neutron meeting | 14:01 |
bauzas | by now | 14:01 |
bauzas | Nova PTG sessions start now https://www.openstack.org/ptg/rooms/newton | 14:02 |
bauzas | sean-k-mooney: joining ? | 14:02 |
sean-k-mooney | oh yes one sec | 14:04 |
whoami-rajat | bauzas, hey, is the nova team ready for the nova-cinder topic? | 14:58 |
bauzas | whoami-rajat: we're a bit late and ending nova-neutron call | 14:59 |
bauzas | whoami-rajat: are you ok if we arrive in 5 mins ? | 14:59 |
whoami-rajat | bauzas, sure, we're on a break as well | 15:00 |
bauzas | whoami-rajat: cool, I'll ping you once we're done | 15:00 |
whoami-rajat | bauzas, ack, thanks | 15:00 |
bauzas | whoami-rajat: in the session :) | 15:06 |
sean-k-mooney | bauzas: can ye hear me ok by the way | 15:24 |
bauzas | 5 secs ago yeah | 15:25 |
sean-k-mooney | ok im not speaking | 15:25 |
stephenfin | sean-k-mooney: Sorry, should have piped up sooner with the patch. I didn't realise we were meeting again today (or that you were planning to talk about this /o\) | 15:39 |
sean-k-mooney | no worreis | 15:43 |
bauzas | folks, hope you still have energy for one last round of nova discussions | 15:46 |
gibi | make it quick :) | 15:47 |
bauzas | gibi: we share the same timezone, hence sharing the same concern | 15:53 |
bauzas | Friday 6pm discussions aren't productive | 15:53 |
artom | bauzas, so, talking to Uggla earlier 1:1, I did have a question around scheduling virtio-fs-enabled instances, and figured it would be a nice PTG discussion | 15:54 |
artom | But I'm not forcing anyone :) | 15:54 |
opendevreview | Stephen Finucane proposed openstack/nova master: db: Narrow down deprecation warning filter https://review.opendev.org/c/openstack/nova/+/837154 | 15:54 |
opendevreview | Stephen Finucane proposed openstack/nova master: db: Trivial rewrapping of warning filters https://review.opendev.org/c/openstack/nova/+/837155 | 15:54 |
opendevreview | Stephen Finucane proposed openstack/nova master: db: Replace use of strings in join, defer operations https://review.opendev.org/c/openstack/nova/+/837156 | 15:54 |
opendevreview | Stephen Finucane proposed openstack/nova master: db: Remove use of empty 'and_()' https://review.opendev.org/c/openstack/nova/+/837157 | 15:54 |
opendevreview | Stephen Finucane proposed openstack/nova master: db: Replace use of Connection.connect() method https://review.opendev.org/c/openstack/nova/+/837158 | 15:54 |
opendevreview | Stephen Finucane proposed openstack/nova master: db: Remove inplicit coercion of SELECTs https://review.opendev.org/c/openstack/nova/+/837159 | 15:54 |
opendevreview | Stephen Finucane proposed openstack/nova master: db: Replace use of Column.copy() method https://review.opendev.org/c/openstack/nova/+/837160 | 15:54 |
opendevreview | Stephen Finucane proposed openstack/nova master: db: Don't rely on autocommit behavior https://review.opendev.org/c/openstack/nova/+/837161 | 15:54 |
opendevreview | Stephen Finucane proposed openstack/nova master: db: Resolve additional SAWarning warnings https://review.opendev.org/c/openstack/nova/+/837162 | 15:54 |
bauzas | 5-min break folks | 15:55 |
bauzas | Uggla: we'll discuss your topic next | 15:56 |
stephenfin | bauzas: That's another one probably deserving of review-priority. That will address the final SQLAlchemy 2.0 deprecation warnings, meaning we're basically done worrying about that ^ | 16:00 |
sean-k-mooney | stephenfin: by the way not that you will like need this much now but you might be interested in https://github.com/SeanMooney/ansible_role_devstack | 16:36 |
sean-k-mooney | docs are still pending | 16:37 |
sean-k-mooney | im currently working on ceph support | 16:37 |
sean-k-mooney | https://github.com/SeanMooney/ansible_role_devstack/tree/master/molecule | 16:37 |
sean-k-mooney | but it can now deploy master train and wllaby multinode | 16:37 |
Uggla | stephenfin, thx for making me realised that I did not use the sdk. | 16:38 |
sean-k-mooney | i might and manilla support to it too | 16:38 |
sean-k-mooney | once i figure out the best way to do that | 16:39 |
stephenfin | sean-k-mooney: neat! :) | 16:40 |
sean-k-mooney | stephenfin: im conidering doing the same with openshift evenutally but its a pain in the ass to install | 16:43 |
*** amoralej is now known as amoralej|off | 16:43 | |
clarkb | sean-k-mooney: if you sort out openshift I'd be interested for nodepool's integration testing wiht it | 16:43 |
clarkb | currently we only test openshift 3 becaues 4 is basically impossible to install without a rack of computers | 16:43 |
sean-k-mooney | clarkb: i have been playing with building my own fedora core os image for vagrant | 16:44 |
sean-k-mooney | and then passing an ignition file | 16:44 |
sean-k-mooney | im hoping ot automate generating the files need to do the okd install | 16:44 |
bauzas | gibi: you ok if I put you co-speaker of the ops feedback forum session I propose ? | 16:44 |
gibi | sure | 16:45 |
bauzas | thanks | 16:45 |
sean-k-mooney | clarkb: so unfortully you woudl need to generate tehm before running molecule but then the rest woudl be automated | 16:45 |
gibi | bauzas: we can brainstorm on a small agenda together later | 16:45 |
sean-k-mooney | clarkb: but its a lot of work to avoid buying a rack | 16:45 |
sean-k-mooney | clarkb: by the way i know the devstack/zuul roles are not ment to be used the way im using them really but it works surprisingly well | 16:46 |
clarkb | well I think the idea was that they could function as normal roles so good to hear (whether or not we ahcived that was neve rreally tested) | 16:47 |
sean-k-mooney | clarkb: i have to do some hack like this https://github.com/SeanMooney/ansible_role_devstack/blob/master/ansible/deploy_multinode_devstack.yaml#L90-L101= | 16:48 |
sean-k-mooney | but a lot of them seam to be usable | 16:48 |
clarkb | I'm not sure that is a huge hack. Basically those are expected inputs to the role (but maybe not very explicitly stated) | 16:49 |
sean-k-mooney | right its not | 16:49 |
sean-k-mooney | i just need to meet the envionment expections of the roles | 16:49 |
sean-k-mooney | they expect zuul and nodepool varibels to exist | 16:49 |
sean-k-mooney | in somecase and thats fine | 16:49 |
sean-k-mooney | the same is ture for groups defiend in the node set | 16:50 |
sean-k-mooney | https://github.com/SeanMooney/ansible_role_devstack/blob/master/molecule/multinode-stable-wallaby/molecule.yml#L45-L48= | 16:50 |
sean-k-mooney | its pretty easy to translate nodeset to the platforms section | 16:51 |
sean-k-mooney | of molecuel | 16:51 |
sean-k-mooney | clarkb: im hoping ot use this to play with creating zuul jobs when i neeed to upstream too | 16:51 |
sean-k-mooney | clarkb: anyway im going to finish up for the weekend soon if i get openshift/okd working ill let you know o/ | 16:52 |
clarkb | thanks! enjoy the weekend | 16:53 |
mloza | hello, is there a policy in nova that will a regular user to view the OS-EXT-SRV-ATTR:hypervisor_hostname and OS-EXT-SRV-ATTR:host of an instance? | 18:39 |
gmann | mloza: yes, you can modify this policy to allow regular user to show these attribute but this is admin only by default - "os_compute_api:os-extended-server-attributes" | 18:46 |
mloza | gmann: copy | 18:52 |
mloza | thx so much | 18:52 |
*** mfo is now known as Guest1503 | 20:49 | |
*** mfo_ is now known as mfo | 20:49 | |
*** dasm is now known as dasm|off | 22:44 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!