| jgilaber | Hi sean-k-mooney, I was wrong the last time we talked about the mtty whitebox vpgu tests. I thought the pre playbook was needed to configure the ssh key to the instances | 08:51 |
|---|---|---|
| jgilaber | but the tests actually create a ssh keypair and pass it when creating the instance | 08:51 |
| jgilaber | after some more digging and asking claude I saw some warnings in the qemu logs for the instance https://storage.bhs.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_583/openstack/58333f7196aa482c9d2b7c737045b906/controller/logs/libvirt/libvirt/qemu/instance-000000f6_log.txt | 08:52 |
| jgilaber | do you think those could be preventing the instance os from booting? From the tempest logs the networking ssh keys and security groups seems correct | 08:53 |
| sean-k-mooney | hum | 08:55 |
| sean-k-mooney | so maybe | 08:55 |
| sean-k-mooney | have you treid doing this locally in devstack yet? | 08:55 |
| sean-k-mooney | we may need to twak the kernel parmaters slightly like im doing in my pci sim job | 08:56 |
| jgilaber | no, I tried before without nova | 08:56 |
| jgilaber | I can update my env and try quickly | 08:56 |
| sean-k-mooney | so pci_sim_sysfs_write Y /sys/module/vfio_iommu_type1/parameters/allow_unsafe_interrupts | 08:57 |
| sean-k-mooney | that just echo Y | sudo tee /sys/module/vfio_iommu_type1/parameters/allow_unsafe_interrupts | 08:57 |
| sean-k-mooney | failing that we can try a newer qemu/kernel by movign the job to debian 13 if its currently on ubutnu 24.04 | 08:59 |
| sean-k-mooney | i will hopefully have 26.04 working in the ci in the next week or two as well | 08:59 |
| sean-k-mooney | tehre are some kernel parmater related to eventfd that we might also ned to tweak | 09:00 |
| sean-k-mooney | sepcificly vm.unprivileged_userfaultfd should be set to 1 | 09:01 |
| sean-k-mooney | asking gemini it say the issue si cosmetic | 09:04 |
| sean-k-mooney | jgilaber: for what its worth i did see the vm boot probely in at least one of the test runs | 09:05 |
| sean-k-mooney | it just didnt get networking properly | 09:05 |
| jgilaber | ack thanks, I'll try in my devstack and see what happens | 09:05 |
| sean-k-mooney | oh | 09:06 |
| sean-k-mooney | gemina has a good point | 09:06 |
| sean-k-mooney | so we used to use 64mb cirros images in the past | 09:06 |
| sean-k-mooney | but we stoped becasuse newer cirros iamge need a little more ram to boot properly | 09:06 |
| sean-k-mooney | we now use 128 or 192 mb flavors | 09:06 |
| sean-k-mooney | but this vm was using a 64mb flavor | 09:07 |
| sean-k-mooney | with 64mb it was prone ot ooming or very slow boots | 09:07 |
| sean-k-mooney | that is proably worth looking ito and or fixing to match the new defautls we use in devstack | 09:08 |
| jgilaber | hmm that's worth a shot while I test things locally | 09:08 |
| jgilaber | I need to update the patch to revert some changes anyway | 09:08 |
| *** jgilaber_ is now known as jgilaber | 12:53 | |
| jgilaber | sean-k-mooney, it's taken me a while but I got a vm with a mtty attached in my devstack | 13:41 |
| jgilaber | lscpi -nn -k gives: "00:05.0 Class 0700: 4348:3253 serial" | 13:42 |
| jgilaber | and in the ci job now it seems to ssh into the vm just fine | 13:43 |
| jgilaber | https://fad10fc5f2c070b32033-beb494a5db5cda67da23a1381224245f.ssl.cf5.rackcdn.com/openstack/5f8a4405ba0f46c78746e4583a18d6e7/testr_results.html | 13:43 |
| sean-k-mooney | cool | 13:44 |
| sean-k-mooney | ok so now its the pre playbook | 13:44 |
| sean-k-mooney | FileNotFoundError: [Errno 2] No such file or directory: '/home/zuul/compute_nodes.yaml' | 13:44 |
| jgilaber | yes, I'm thinking about creating a new base job in the whitebox plugin | 13:45 |
| sean-k-mooney | so your running thest that are tryign to ssh into the vms or connect to the db so it need that file | 13:45 |
| sean-k-mooney | you could | 13:45 |
| sean-k-mooney | but we dont wnat nova-next to intherit form whitebox | 13:45 |
| sean-k-mooney | we could do this testignin the vtpm job instead | 13:46 |
| sean-k-mooney | that inherits form whitebox alrady | 13:46 |
| sean-k-mooney | that might be a better base | 13:46 |
| jgilaber | ack, if there no particular reason to use nova-next i can switch | 13:47 |
| jgilaber | from nova pov what would be better, add the tests to the vptm job or create a new job just for this? | 13:47 |
| sean-k-mooney | we jus tchose it because it test nonstandard things | 13:47 |
| sean-k-mooney | btu nova-next is getting quite big | 13:47 |
| sean-k-mooney | we shoudl dicss this in nova channel with melwitt | 13:47 |
| jgilaber | ack, let me ask there, thanks! | 13:48 |
| sean-k-mooney | but i think we coudl rename nova-vtpm to nova-whitebox and just add it there | 13:48 |
| sean-k-mooney | jgilaber: for context the vtpm job didnt exist when those intial patches were written | 13:48 |
| opendevreview | sean mooney proposed openstack/cyborg master: devstack: add pci-sim fake SR-IOV PCI passthrough test fixture https://review.opendev.org/c/openstack/cyborg/+/989470 | 15:20 |
| opendevreview | sean mooney proposed openstack/cyborg master: Add pci-sim developer guide https://review.opendev.org/c/openstack/cyborg/+/991177 | 15:20 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!