Thursday, 2022-08-04

« Wednesday, 2022-08-03 Index Friday, 2022-08-05 »
*** ysandeep|out is now known as ysandeep06:04
*** ysandeep is now known as ysandeep|afk07:27
jrosser_morning07:40
noonedeadpunkhey \o/08:08
*** ysandeep|afk is now known as ysandeep08:22
damiandabrowskihi!08:23
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-haproxy_server master: Allow haproxy to bind on the interface  https://review.opendev.org/c/openstack/openstack-ansible-haproxy_server/+/85203908:40
* jrosser_ ^ interesting08:44
jrosser_this is SO_BINDTODEVICE stuff i guess08:48
noonedeadpunkyup, it is08:50
jrosser_why is it needed to bind to 0.0.0.0 in that case?08:51
noonedeadpunkwell, I define *, but in facts it's still binding to 0.0.0.008:52
jrosser_maybe i just miss some reason that this ip isn't known08:52
noonedeadpunkor well, it can bind to [::]08:53
noonedeadpunkso why I'm trying to do that, is to implement DNS round-robin. With that, 3 keepalived groups are configured, each with it's own master on different controllers. So if define each VIP individually, you would get quite a lot bindings which is not that efficient and config gets really big, and you can't use DNS as only first resolved IP would be used08:56
noonedeadpunkSo IPs are known, there're just too much of them at the end. Like 3 public and 3 private VIPs08:57
noonedeadpunkthat would mean 9 frontends defined per service08:58
noonedeadpunks/9/6/08:58
mrfMorning :)09:10
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Pass br-bmaas or internal container interface IP to os_ironic role  https://review.opendev.org/c/openstack/openstack-ansible/+/85212009:18
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Install only the required dhcp config files for inspector  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85212109:27
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Bind http and tftp services to the bmaas network  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85212209:28
noonedeadpunkjrosser_: regarding https://review.opendev.org/c/openstack/openstack-ansible/+/852120 - are you sure about that? As looking at how os_octavia handling that https://opendev.org/openstack/openstack-ansible-os_octavia/src/branch/master/defaults/main.yml#L35109:39
noonedeadpunknot really saying it's better, but just pointing to the fact octavia does this differently :)09:40
jrosser_i was looking at octavia too to see how it is done09:40
noonedeadpunkbut it's playbook not calling for  dynamic-address-fact.yml ?09:41
jrosser_no it's not09:41
noonedeadpunkso my question is more - should we patch also octavia to match ? or there're different needs for ironic?09:41
jrosser_looking at the octavia defaults what it seems to be wanting is the interface name, for iptables09:42
jrosser_and then a list of all the HM in octavia_hm_hosts09:43
noonedeadpunkwell, we are getting address there with `hostvars[host]['container_networks'][octavia_container_network_name]['address']`09:43
jrosser_so yeah, i was looking at dynamic-address-fact and seeing how it handles metal/not metal quite specifically09:43
jrosser_then being confused about why we do things quite differently in the octaiva role09:43
noonedeadpunkbut I don't really have an opinion on what's better approach09:44
noonedeadpunkdynamic-address-fact was quite confusing from what I can recall when debugging it09:44
jrosser_maybe it needs another look09:46
jrosser_because of this https://github.com/openstack/openstack-ansible/blob/master/playbooks/common-tasks/dynamic-address-fact.yml#L30-L3209:46
jrosser_i think you might be right, that dynamic-address-fact is not needed in this case09:50
noonedeadpunkSo I feel that octavia might be more flexible because you can override that all if our assumptions are not correct for some reason09:51
jrosser_container_networks['bmaas_address']['address'] gives the right thing09:51
noonedeadpunkAt the same time I guess we get all information we need from basically same place (facts)09:51
noonedeadpunkor well, inventory09:52
jrosser_i'll make it like the octavia one09:52
opendevreviewMerged openstack/openstack-ansible stable/xena: Bump OpenStack-Ansible Xena  https://review.opendev.org/c/openstack/openstack-ansible/+/85197409:59
noonedeadpunkok, so hopefully we can merge Yoga stuff now...10:06
noonedeadpunkgive rocky tests are fixed10:07
noonedeadpunk*given10:07
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible stable/yoga: rabbitmq: default to using TLS for management user interface  https://review.opendev.org/c/openstack/openstack-ansible/+/85138010:15
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-lxc_hosts stable/yoga: Take account of lxc_apt_mirror in new debootstrap command  https://review.opendev.org/c/openstack/openstack-ansible-lxc_hosts/+/85138710:17
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-haproxy_server master: Allow haproxy to bind on the interface  https://review.opendev.org/c/openstack/openstack-ansible-haproxy_server/+/85203910:38
noonedeadpunkdamn, I tested new scenario pretty good but clean forgot to check if it's not breaking old one. We defenitally need to come up with more precise individual roles testing.10:41
noonedeadpunk*definitely10:41
opendevreviewMerged openstack/openstack-ansible stable/yoga: Correct series names for documentation.  https://review.opendev.org/c/openstack/openstack-ansible/+/85188210:41
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-rabbitmq_server stable/ussuri: Use cloudsmith repo for rabbit and erlang  https://review.opendev.org/c/openstack/openstack-ansible-rabbitmq_server/+/85035011:17
*** dviroel|afk is now known as dviroel11:34
*** ysandeep is now known as ysandeep|afk12:03
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Bind http and tftp services to the bmaas network  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85212212:05
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Bind http and tftp services to the bmaas network  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85212213:17
jrosser_jamesdenton: thanks for looking at the ironic patches13:17
jrosser_there seems to be a lot of cleanup to do13:17
jamesdentonno problem! and yes, there is a bit of cleanup needed13:17
jamesdentonmy test setup has been metal, as you pointed out. i have some thoughts on what this change might affect, but will have to see13:18
jrosser_i expect it's pretty much working in metal13:18
jamesdentonyep.13:18
mrfjrosser i can suggest that we osa modify the etc/hosts file save a .original file and then if lxc-containers-destroy rollback to .original hostfile :P13:18
jrosser_jamesdenton: i'd certainly be interested in your thoughts on the network / port binding changes13:19
jamesdentonin fact, jrosser_, i do not have a bmaas network of any kind13:19
jrosser_right, i was thinking it must be something like that as i just don't understand how some of this ever worked13:20
jrosser_unless it was some sort of metal / everything-routes-to-everything setup13:20
jrosser_mrf: but what would you roll it back to? :)13:21
jrosser_you can delete any / some / all containers13:21
jamesdentonwell, i am using the 'neutron' interface, too, which relies on a provider network to exist that is used for PXE. it's routed and can reach the mgmt network of the hosts, where port 8051 is listening13:21
jrosser_i think we might be doing that too13:21
jrosser_but on the controllers hook it to br-bmaas and then that into the ironic LXC13:22
jamesdentongotcha.13:22
jamesdentoni have avoided br-bmaas and br-lbaas in my deployment13:22
jrosser_so we need to make sure that both of these approachescan be made to work13:22
jamesdentonagreed13:23
jamesdentoni have some free cycles next week and will try and document the metal scenario w/ diagrams and overrides and whatnot13:23
jrosser_it's likey that if you changed my new var `ironic_container_network_name: "bmaas_address"` to be container_address then it would bind to the mgmt address13:24
mrfjrosser rollback to the original file :P13:25
mrfas you done a backup before touch :O13:25
jamesdentoni think you're right13:25
jrosser_not sure i'm totally happy with that var name tbh13:25
jrosser_i was looking at the octavia role and omg the naming is confusing there13:26
jamesdentoni tried to simplify my life there. too many bridges and right turns13:26
jamesdentonoctavia_hm_hosts: "{% for host in groups[octavia_hm_group] %}{{ hostvars[host]['container_address'] }}{% if not loop.last %},{% endif %}{% endfor %}"13:27
jamesdentonand now i don't need the bridge, it's just routed13:27
jrosser_the other thing that you have easier will be the provisioning callback address13:27
jamesdentonin this environment, container network == management network. iirc you did something similar13:27
jamesdentonyes13:27
jrosser_as the provisioning network can route to the internal VIP13:28
jamesdentonyeah, lxc makes that trickier, i admit13:28
jrosser_with LXC/br-bmaas thats tricker13:28
jrosser_currently we've overridenn the callback address to be the api endpoint in the ironic containers (different in each one) rather than use the VIP13:28
jrosser_as thats reachable from the br-bmaas13:29
NeilHanlonrocky 9 images should be available to build via DIB now. just put in https://review.opendev.org/c/openstack/project-config/+/852167 to get it onto nodepool13:29
jrosser_though the HA for ironic is something that i don't fully understand atm13:29
jrosser_as it seems each baremetal node is assigned to a specific conductor for provisioning13:30
jamesdentonyes, each node is assigned to a conductor13:30
jamesdentoni can't recall how a new conductor is selected13:30
jrosser_so it's not actually a great loss that each conductor gives out the IP of it's adjacent ironic-api as the callback location13:30
jamesdentonseems pretty reasonable, until all of your nodes end up on the same conductor13:31
jamesdentoni have a 4:1:0 distribution at the moment13:31
jrosser_NeilHanlon: thats great - hopefully it's not so bad to get OSA working on those as we already have some centos-913:37
NeilHanlonyep, i've got a machine booting up now to test and try and get ahead13:39
jrosser_excellent13:41
mgariepyhttps://www.theregister.com/2022/08/04/gitlab_data_retention_policy/13:42
mgariepywow.13:42
*** ysandeep|afk is now known as ysandeep13:45
NeilHanloni.. uhh.13:45
NeilHanlonwow. yep. that's the word mgariepy13:45
mgariepywtf would also be correct on this one...13:46
NeilHanlonmy first thought was 'wtaf'13:46
jrosser_it would just take one script to automatically post an issue to each repo and they'd no longer be dormant13:50
mgariepydormant project would be more like project that don't get cloned for a number of years imo.13:50
jrosser_`A single comment, commit, or new issue posted to a project during a 12-month period will be sufficient to keep the project alive.`13:51
mgariepyrun that script from a running in gitlab itself ! :P13:51
mgariepyrunner**13:51
NeilHanlon😂13:52
mgariepyhttps://www.vice.com/en/article/ywen8x/13000-projects-ditched-github-for-gitlab-monday-morning13:53
mgariepygood old days..13:53
* NeilHanlon grumbles about github and gitlab and how everyone should just run their own gitea server13:54
mgariepyunless gitea code is hosted in gitlab :P which might get it deleted haha13:54
NeilHanlonlol 😂14:04
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Ensure ironic inspector dhcp server listen address is defined  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85217314:12
*** frenzyfriday|rover is now known as frenzyfriday|rover|lunch14:16
NeilHanlonhrm.. jrosser_ - i feel like we've run into this before https://paste.opendev.org/show/bHs06pPUkLUi29wXKZdH/14:22
jrosser_NeilHanlon: lxc-templates i think?14:23
jrosser_there was a packaging wierdness that meant they were missing so we did have something similar14:24
jrosser_and i think you addressed that in your PPA <- insert correct term here14:24
NeilHanlonah, right14:24
NeilHanlon(COPR)14:24
mrfhmmm im getting 14:25
jrosser_i think you can hack around it by grabbing the template from somewhere else and dropping it in the right /usr/share/somewhere/...14:26
NeilHanloni remember now.. brain is a hot mess after the last few months lol14:29
NeilHanlonI had been speaking to the individual who maintains the epel lxc package about it, and he seemed open to putting it into either the lxc package or perhaps a new package altogether--so i'm going to track that down. will help in c9s and r9.14:29
NeilHanlonthough I guess those templates are 'legacy' so maybe we (OSA) should investigate if there's a better way to accomplish the tasks that OSA performs?14:30
mrfjrosser from distro to master distro the haproxy required for us haproxy_keepalived_authentication_password 14:32
mrfas mandatory param14:32
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Attach bmaas network to ironic_api containers  https://review.opendev.org/c/openstack/openstack-ansible/+/85217414:32
mrfok im retard :P14:32
mrfi deleted the secrets file too...14:33
jrosser_oops14:34
jrosser_mrf: having /etc/openstack-deploy as a git repo is a good plan14:35
mgariepyand do not forget to commit your changes :D14:35
NeilHanlonwrap popd to git commit whenever you leave the directory /s14:41
*** ysandeep is now known as ysandeep|out14:52
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Bind http and tftp services to the bmaas network  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85212215:11
opendevreviewNeil Hanlon proposed openstack/openstack-ansible master: Use the centos logic as Rocky 8 is not supported  https://review.opendev.org/c/openstack/openstack-ansible/+/85218115:26
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Remove ironic_server from env.d  https://review.opendev.org/c/openstack/openstack-ansible/+/85218215:29
opendevreviewNeil Hanlon proposed openstack/openstack-ansible-plugins master: Bring redhat vars in line with centos overrides  https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/85218315:29
jrosser_NeilHanlon: if you want to backport Rocky-9 to Yoga you might want to do things in two steps15:32
NeilHanlonw.r.t. the bootstrap script?15:33
jrosser_well just generally - i was going to mention it earlier15:33
jrosser_as theres no doubt tons of opportunity to clean up 8 vs 9 variables and stuff15:33
NeilHanlongotcha gotcha15:33
noonedeadpunkI believe we even hold up a bit with cleaning out r8 just for ease of backport15:34
jrosser_i think oversimplifying hugely we can add rocky9 and backport it as needed, then go and clean out 8 properly and refactor stuff on master to make it tidy15:36
* jrosser_ wtf at os_ironic role targeting ironic_compute group15:37
jrosser_mess mess mess15:37
NeilHanlonokay, that makes sense. my strategy right now is just manually walking through an AIO install and trying to fix what I find along the way15:37
NeilHanlone.g. the ansible-plugins change is due to it trying to install the non-existent powertools on rocky 915:39
jrosser_NeilHanlon: you can make those vars files more OS specific15:44
jrosser_https://github.com/openstack/openstack-ansible-plugins/blob/master/roles/glusterfs/tasks/main.yml#L16-L3015:45
jrosser_if you need a rocky-9 specific one then i think you can make rocky-9.yml, or redhat-9.yml, and perhaps rename the existing one to -8 to make that apply only to the -8 releases15:46
jrosser_with it called 'redhat.yml' at the moment it applies to * in that OS family15:46
jrosser_i expect quite a bit of adding rocky-9 will be messing with vars file naming like this15:46
*** dviroel is now known as dviroel|lunch15:47
NeilHanlonUnderstood. that makes sense15:50
*** ysandeep|out is now known as ysandeep|PTO15:51
NeilHanloni'm hearing "don't be afraid to make new files for rocky even if it means they have duplicate content for a while"15:51
jrosser_theres a judgement to be made, if the file is small and really only contains the things that vary then thats fine15:51
jrosser_but if its large and you can do the same by adding a ternary(this, that) to handle 8 and 9, then use that15:52
NeilHanlonack15:52
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Remove ironic_server from env.d  https://review.opendev.org/c/openstack/openstack-ansible/+/85218216:03
jrosser_jamesdenton: this is puzzling https://paste.opendev.org/show/bQa0Q4GG7BIE7VbksMqT/16:24
jrosser_like how does the ironic_compute thing end up in ironic_all?16:24
jamesdentonhmm16:27
jrosser_it tries to template out all sorts of ironic config files into there16:28
jrosser_which is just wrong16:28
jamesdentonyeah it's really just a vehicle for nova compute service, right>16:28
jrosser_thats my understandling16:28
jrosser_lets also note that it deploys a copy of nova-scheduler too16:28
jrosser_which looks also pretty bogus16:28
jamesdentonlol, i'd say so16:29
jrosser_something wierd is happening there, env.d and the inventory json don't appear to put the compute host in ironic_all16:30
jamesdentonagreed16:31
jamesdentonis that master?16:32
jamesdentonjrosser_ less nefarious than we thought... https://github.com/openstack/openstack-ansible/blob/master/inventory/inventory.ini#L12916:50
*** dviroel|lunch is now known as dviroel16:52
opendevreviewJames Denton proposed openstack/openstack-ansible master: Remove ironic_compute container from ironic_all  https://review.opendev.org/c/openstack/openstack-ansible/+/85219716:55
*** frenzyfriday|rover|lunch is now known as frenzyfriday|rover16:55
jrosser_jamesdenton: 8-O16:58
jrosser_omg i had no idea that file did anything16:58
jrosser_it even says [WARNING]: Unable to parse /etc/openstack_deploy/inventory.ini as an inventory source16:59
jrosser_oh well, thats not it is it /o\16:59
jrosser_noonedeadpunk: have you seen this?17:00
jamesdentonhttps://paste.opendev.org/show/bU9rh1VL2WUEFQ8GuH8A/17:01
jrosser_yeah i've got the same here17:01
jamesdentonbut it's complaining about /etc, not /opt17:01
jrosser_i'd always thought that the inventory/inventory.ini was there as an example, if you wanted to make a static inventory17:02
jrosser_not that it was merged in with the dynamically generated one17:02
jamesdenton./scripts/openstack-ansible.rc:19:export ANSIBLE_INVENTORY="${ANSIBLE_INVENTORY:-OSA_INVENTORY_PATH/dynamic_inventory.py,OSA_INVENTORY_PATH/inventory.ini,${OSA_CONFIG_DIR}/inventory.ini}"17:02
jamesdentoninteresting.17:03
jrosser_well there you go17:03
jrosser_i wonder if thats actually what we want17:03
jamesdentonguess it's a convenient way to create new groups. or maybe that's just the brains behind the python, i dunno17:04
jamesdentonwell, TIL i guess17:05
jrosser_snap :)17:05
jamesdentonmoving on then :D17:06
jrosser_indeed17:06
jrosser_i have a whole bunch of patches for ironic now17:06
jamesdentonsweet! i'm actually OOO tomorrow thru Wednesday, but will happily test upon my return17:07
jrosser_cool that would be great17:07
opendevreviewJames Denton proposed openstack/openstack-ansible-os_ironic master: Remove ironic_server from inventory  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85219917:13
jamesdentonjrosser_ do you use the MNAIO?17:17
jrosser_not for a long time17:17
jrosser_though maybe when we have ironic deployed that will be less hassle17:18
jrosser_as it needs a pretty huge machine iirc17:18
jamesdentonit does need some beef, yes17:18
jrosser_we really don't test os_ironic at all in CI17:18
jamesdentoni thought about an MNAIO on AIO type thing, but the AIO deployment is fairly lengthy by itself, and the trouble of Ironic w/ OVS wasn't something i wanted to tackle.17:19
NeilHanlonoh.. MNAIO sounds fun :D17:24
jamesdentonNeilHanlon https://github.com/openstack/openstack-ansible-ops/tree/master/multi-node-aio - feel free to poke17:24
jamesdentonit's been handy for multi-node upgrade testing and quick rollback17:25
NeilHanlonTY :) it sounds excellent, assuming I can repurpose a box to run it lol17:26
jrosser_fast disks == win for that17:26
jamesdentonfor sure. NVMe if you've got it. 17:27
mgariepyor https://xkcd.com/303/ 17:28
jamesdentonguilty17:28
* NeilHanlon swipes the credit card and calls the accountant to make sure I can expense it17:31
jrosser_i think i need to make a metal ironic AIO17:32
jrosser_not sure what is happening with the networking with that17:32
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Bind http and tftp services to the bmaas network  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85212218:07
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-rabbitmq_server stable/ussuri: Use cloudsmith repo for rabbit and erlang  https://review.opendev.org/c/openstack/openstack-ansible-rabbitmq_server/+/85035018:13
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Attach bmaas network to ironic_api containers  https://review.opendev.org/c/openstack/openstack-ansible/+/85217418:28
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Ensure ironic inspector dhcp server listen address is defined  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85217318:28
spatelnoonedeadpunk around18:50
spatelor jrosser_ or you :)18:50
spatelDoes nova support sound and mic on RDP desktop of windows? 18:51
spatelmy customer has very strange requirement :) 18:51
mgariepyi guess it doesn't really matter does it ?18:51
mgariepyi mean do you need a sound card to play sound via network socket ?18:52
spatelRequirement is VDI design where on windows RDP people can do zoom conference and listen music + they can use mic  etc..18:52
spatelThey tested it on VMware and its working.. so i gave them openstack VM to test out and audio/mic doesn't working on kvm 18:53
spatellook like sound adaptor doesn't support in KVM or may be openstack doesn't load driver 18:54
spatelI didn't find any stanza in kvm dumpxml file related sound adaptor 18:54
jrosser_are you sure it's a feature of the virtualisation and not that some sort of audio redirection needs to be enabled in windows?19:07
jamesdentonsome vintage mailing list threads mention that while KVM can do it, it's not exposed via Nova/libvirt. The recommendation was to implement libvirt hooks to modify the xml on the fly19:12
jamesdentonvery vintage though19:12
*** dviroel is now known as dviroel|afk20:59
opendevreviewMerged openstack/openstack-ansible-os_rally stable/yoga: Control rally-openstack installed version  https://review.opendev.org/c/openstack/openstack-ansible-os_rally/+/84922721:54
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Bind http and tftp services to the bmaas network  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85212221:55
opendevreviewJonathan Rosser proposed openstack/openstack-ansible-os_ironic master: Ensure ironic inspector dhcp server listen address is defined  https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/85217321:55
*** dviroel|afk is now known as dviroel23:30
*** dviroel is now known as dviroel|out23:36
« Wednesday, 2022-08-03 Index Friday, 2022-08-05 »

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!