October 2024 - openstack-discuss - lists.openstack.org

[ptg][publiccloud-sig] PTG meetings for tthe publiccloud-sig
by Felix Kronlage-Dammers 19 Oct '24

19 Oct '24

Ahoi, as all other teams the publiccloud-sig is getting ready for the upcoming PTG as well :) We’re very much looking forward to the next week. The publiccloud-sig will meet for the PTG on Tuesday at 1600 UTC and on Wednesday at 0700 UTC. The regular etherpad[1] contains a list of items that we collected and that we will likely address. The tuesday slot should also allow for participants not in one of the EU and NZ timezones to join - we would love to get a nice group together. The PTG also serves as a good starting point to become more active in the publiccloud-sig :) See ya next week! felix [1] https://etherpad.opendev.org/p/publiccloud-sig

1 0

[watcher]Interest in a session at upcoming PTG?
by Ronelle Landy 18 Oct '24

18 Oct '24

Hello, There has been some renewed interest in the Watcher project. We wanted to reach out to see if the project PTL (or anyone else) would be willing to run/attend a session on Watcher at the upcoming PTG. Would it make sense to hold a dedicated session or discuss plans at one of the TC sessions? Thank you!

3 3

[qa][ptg] Virtual 2025.1 vPTG Planning
by Martin Kopec 18 Oct '24

18 Oct '24

Hello everyone, here is [1] our etherpad for the Epoxy/2025.1 virtual PTG. Please, add your topics there if there is anything you would like to discuss / propose for discussion ... You can also vote for time slots of the sessions so that they fit your schedule [2]. The number of slots will depend on the number of topics proposed in [1]. [1] https://etherpad.opendev.org/p/oct2024-ptg-os-qa [2] https://framadate.org/juHJSryIYHMRwdHA Thanks, -- Martin Kopec Principal Software Quality Engineer Red Hat EMEA IM: kopecmartin

1 1

[ptg][sdk][cli] PTG meeting of the SDK team
by Artem Goncharov 18 Oct '24

18 Oct '24

I have booked a slot for Monday 14:00-15:59 UTC for the SDK team (Monday is the day with least conflicts). Etherpad: [1] Topics: * state of the OpenAPI * codegenerator: state and next steps [1]: https://etherpad.opendev.org/p/oct2024-ptg-os-sdkcli See you there, Artem

1 0

[kolla-ansible] Problems with my lab . Many bugs with python
by Franck VEDEL 18 Oct '24

18 Oct '24

hello. I'm going crazy. Last year I set up a lab for students. The platform: ubuntu 22.04, kolla-ansible, antelope. Everything works perfectly by following the link https://docs.openstack.org/kolla-ansible/2023.1/user/quickstart.html We start with all-in-one in a virtual machine, then 20 hours later, a cluster of 3 nodes on DELL servers, with TLS, Ceph. I was super happy. I'm coming back to the subject a year later. I'm thinking about doing an update and so I start, still on ubuntu 22.04, by following the link https://docs.openstack.org/kolla-ansible/2024.1/user/quickstart.html well, it doesn't work. after hours of research, I resume an installation with Antelope. It doesn't work either. New research... and I come across a bug. Someone explains that you have to downgrade MarKupSafe to version 2.1.5. Okay, I'll do it. I start again with Caracal... and now, in the Horizon interface, we no longer see multiple domains. It doesn't work. New bug. I come back with Antelope and MarkupSafe == 2.1.5... it works, and my two domains are OK in Horizon. So I continue with Antelope ..... and now, when I try to activate TLS, it's impossible to do the "deploy". New bug but on rabbitmq. I'm going crazy. New bug but no solutions... The only solution I found is to restart a virtual machine that worked last year, do a pip freeze and in this year's installation, use last year's pip freeze. I can't use Caracal because the second domain is not visible. I can (maybe I'm not done yet) use Antelope but with last year's requirements.txt. How do I get out of this situation? How would you do it in my place knowing that I want to keep Ubuntu 22.04? Bobcat? I've already spent days on the subject. I don't have time anymore. That's why I'm asking for your help. Thansk a lot Franck

1 0

[neutron] Canceling weekly meetings tomorrow and next week
by Brian Haley 17 Oct '24

17 Oct '24

Hello Neutrinos: I'm going to cancel tomorrow's drivers meeting as well as our meetings next week as we have the PTG sessions. Please add any items there if you wish to discuss them [0]. Have a nice weekend! -Brian [0] https://etherpad.opendev.org/p/oct2024-ptg-neutron

1 0

[kolla][neutron][ovn] High ovs-dbserver CPU Usage with OVN in Kolla-ansible Large-Scale Deployment
by 韩光宇 17 Oct '24

17 Oct '24

Hi, I have a Wallaby cluster with 5 controller and 130 compute node. Host OS: Rocky 8.7 Deploy tool: Kolla-ansbile kolla_base_distro(container images): centos Network Mode：**OVN** After completing the deployment, the functionality for creating a single instance works as expected, and instances are created successfully. However, the process is extremely slow. When attempting to create instances in bulk, the operation fails on a large scale. During this time, I observed that the ovsdb-server is consuming close to 100% CPU usage. Even without any tasks being processed, the CPU usage remains consistently around 90%, just from the cluster's regular operations. I redeployed the cluster and switched the network mode to Open vSwitch (OVS), and everything works as expected with normal performance. Could anyone offer guidance on what might be causing this issue? I’m aiming to use the OVN mode with Kolla in a large-scale OpenStack cluster (100+ nodes). What configuration or adjustments would I need to make? Thank you in advance for your help! Han Guangyu

2 2

[all][qa] Moving grenade to install things in venv: Enable GLOBAL_VENV
by Ghanshyam Mann 17 Oct '24

17 Oct '24

Hello Everyone, Since 2023.2 devstack by default, installs things in virtual env and due to the old node was not ready in the grenade job, we disabled that with the flag GLOBAL_VENV. Now it is time to enable it in greande also. This also fixes the failure we are getting in the grenade setup for the current master (upgrade testing from stable/2024.2 -> master) which is setuptool error for projects that have pyproject.toml[1]. I have proposed the changes in granade, but at least Octavia, Ironic, and heat jobs are failing. I have proposed some fixes for these projects. - https://review.opendev.org/q/topic:%22grenade-venv%22 I would like to request the project to test their grenade jobs with https://review.opendev.org/c/openstack/grenade/+/930507 and fix the failure if any. I would like to merge the grenade change as soon as its gate passes so that we can set up the grenade testing correctly for the current master. If any project requires a large change/more time to make their grenade jobs work in venv then you can disable the GLOBAL_VENV there for the time being and enable it once things work. [1] https://zuul.opendev.org/t/openstack/build/517b21fc57d6480f902273e58f9032b5 -gmann

1 1

[Glance] Cancelling weekly meetings Today and Next week
by Pranali Deore 17 Oct '24

17 Oct '24

Hello Team !! I won't be able to chair today's meeting, I will be out for some personal reason and also there is nothing much on agenda for this week, So cancelling the meeting for today. Sorry for the last minute update ! Next week's meeting is also cancelled because of the PTG session. So you all next week!! Thanks & Regards, Pranali Deore

1 0

[keystone] What does a multi-region service catalog look like?
by Stephen Finucane 16 Oct '24

16 Oct '24

o/ The question in $subject came up in a discussion earlier today and I neither knew the answer nor was able to discover it by looking at a service catalogs for a few production deployments I have access to (also, the multi-region feature is currently broken in DevStack and likely has been for some time). Crucially, in a "standard" multi-region deployment where there is one Keystone instances and N Nova/Cinder/Glance/Neutron/... deployments, would I have something like the below, where there is only one identity endpoint but multiple e.g. compute endpoints, or would I expect there always to be as many identity endpoints as there are e.g. compute endpoints? { "token": { ... "catalog": [ { "endpoints": [ { "id": "fb757b386a23430d8f4d636b8b45fa89", "interface": "public", "region_id": "RegionOne", "url": "http://10.0.110.125/identity", "region": "RegionOne" } ], "id": "50c4feff9d6149b898fce20ce23904a3", "type": "identity", "name": "keystone" }, { "endpoints": [ { "id": "2ed0cb59adb0415c948f8351b6546d3e", "interface": "public", "region_id": "RegionOne", "url": "http://10.0.110.125/compute/v2.1", "region": "RegionOne" }, { "id": "2ed0cb59adb0415c948f8351b6546d3e", "interface": "public", "region_id": "RegionTwo", "url": "http://10.0.112.232/compute/v2.1", "region": "RegionOne" } ], "id": "87322637c5ab4e7b90bea7a8c1eb9525", "type": "compute", "name": "nova" }, ... ] } } I'm guessing the latter, but I'd like to confirm this. Stephen

3 4