| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-cyborg - 2026-03-31 | |||
| 14:05:13 | chandankumar | Do we want to drop the devstack_setup.rst in favor of devstack_nvme.rst? | |
| 14:05:20 | sean-k-mooney | in my opion no | |
| 14:05:32 | sean-k-mooney | i just started reviewing https://review.opendev.org/c/openstack/cyborg/+/982711/2/doc/source/contributor/devstack_nvme.rst | |
| 14:05:36 | sean-k-mooney | before the meeting | |
| 14:05:48 | sean-k-mooney | and there are a number of things i woudl perfer to change before proceedign with it | |
| 14:05:48 | chandankumar | https://review.opendev.org/c/openstack/cyborg/+/982711 provides https://review.opendev.org/c/openstack/cyborg/+/982711/2/devstack/local-conf | |
| 14:06:20 | sean-k-mooney | yep im ok with providing sample local.confs | |
| 14:06:43 | chandankumar | In that case, I will wait for the review comments. | |
| 14:06:47 | sean-k-mooney | the main devstack document shoud ljust descibe the basics fo deploying cyborg | |
| 14:07:17 | sean-k-mooney | then we shoudl have driver specific docs that explain how to use emulated or real hardware with devstack for each dirver | |
| 14:07:35 | sean-k-mooney | so to me they fufill diffent usecase | |
| 14:08:02 | sean-k-mooney | what i dont want to have to do is try and figureou out what parts to remove if im just working on the nvidia driver or pci driver | |
| 14:08:14 | sean-k-mooney | form the devstack_nvme doc | |
| 14:08:39 | sean-k-mooney | does that make sense | |
| 14:08:54 | chandankumar | yes, sounds good. | |
| 14:09:08 | sean-k-mooney | the other thing i woudl personally change is remove the use of virt-install and instead provide an example domain exml and use virsh define | |
| 14:09:28 | sean-k-mooney | the current mix of virsh install + virsh edit is a littel confusing and messy | |
| 14:09:42 | sean-k-mooney | as in something are done on the cli then we have to add the iommu eleemtns | |
| 14:10:18 | sean-k-mooney | idealy i woudl have a seperate doc for creating a development vm tahat we can cross refence and have the nvme guide focus on the compliation and cybrog configureation | |
| 14:11:14 | sean-k-mooney | so baiscly 3 docs 1 for optional vm creation, 1 for the geneeric devstack usage to deploy cyborg and a final one that docuemtn how to configure the nvme driver with real or fake nvme devices | |
| 14:12:11 | chandankumar | ok, these are good suggestions, it will make the docs experience cleaner. | |
| 14:12:16 | sean-k-mooney | https://review.opendev.org/c/openstack/cyborg/+/982711/2/doc/source/contributor/devstack_nvme.rst is currently covering 4 related but difernet topics | |
| 14:12:53 | sean-k-mooney | host config/vm creation, emulated deivce creation, devstack install. and finally using the nvme driver | |
| 14:12:57 | chandankumar | I will break the patch into 2 patches one for devstack with fresh instructions and another for nvme or fake driver with vs creation | |
| 14:12:59 | sean-k-mooney | well or pci driver currently | |
| 14:13:11 | sean-k-mooney | the lattter part probaly shoudl be in the user docs for the driver | |
| 14:13:49 | chandankumar | ok | |
| 14:14:02 | sean-k-mooney | ack hte tempest plugin and verifcaion part including the creation fo the device profile ectra proably shoudl be its own doc too | |
| 14:14:17 | sean-k-mooney | since that will be pretty commen for all drivers | |
| 14:14:28 | sean-k-mooney | and we shoudl avoid duplciating it | |
| 14:14:58 | sean-k-mooney | i can leave some of these comemnts in the review later but that the majority fo my feed back | |
| 14:15:18 | chandankumar | Ok, I will work on breaking it into different docs | |
| 14:15:29 | sean-k-mooney | content looks reasonable over all but the doc is covering to many topic today | |
| 14:15:41 | sean-k-mooney | cool ill re reivew when you revise it | |
| 14:15:51 | chandankumar | sure, thank you! | |
| 14:16:16 | chandankumar | Now moving to reviews since there is no other topic | |
| 14:16:20 | chandankumar | #topic reviews | |
| 14:16:31 | chandankumar | I was working on cleaning up python-cyborgclient | |
| 14:16:44 | chandankumar | proposed these two simple reviews | |
| 14:16:47 | chandankumar | #link 982246: Drop dead code common/http.py | https://review.opendev.org/c/openstack/python-cyborgclient/+/982246 | |
| 14:17:15 | chandankumar | common/http.py is dead code, not used anywhere in cyborgclient. | |
| 14:17:44 | chandankumar | and second on cleaning up unused deps | |
| 14:17:47 | sean-k-mooney | i was going to ask have you cofnirmed its not used in horizon but we dont have a plugin | |
| 14:17:47 | chandankumar | # 982244: Cleanup requirements.txt unused packages | https://review.opendev.org/c/openstack/python-cyborgclient/+/982244 | |
| 14:18:43 | chandankumar | https://github.com/openstack/python-cyborgclient/blob/master/cyborgclient/common/httpclient.py is used in the cyborgclient not http.py | |
| 14:19:22 | sean-k-mooney | https://minio-api.teim.app/zuul-logs/8bf/main/8bf05a369b744132ab643a1fdf5d17c3/code-review/review-report.html yep the ai review agrees | |
| 14:19:39 | sean-k-mooney | i think the removal shoudl be safe but ill verify that later | |
| 14:19:45 | sean-k-mooney | i think directionaly this is good | |
| 14:20:00 | sean-k-mooney | we will need a release note | |
| 14:20:29 | sean-k-mooney | otherwise i think the patch is fine. the cyborg client may be used by thirdparty external tools | |
| 14:20:48 | chandankumar | ah ok, will add a release note, good point about external tools | |
| 14:20:50 | sean-k-mooney | so we will need to do a major version bump when we release it | |
| 14:21:00 | sean-k-mooney | but i think that is fine | |
| 14:21:27 | sean-k-mooney | as for the requireme clean up i agree that also looks reasonabe. | |
| 14:21:45 | chandankumar | https://review.opendev.org/c/openstack/python-cyborgclient/+/982244/2/requirements.txt it drops cryptography | |
| 14:21:51 | sean-k-mooney | oslo.log we may revisit but in general the client "logs" via print statmetnt | |
| 14:22:20 | chandankumar | as we were discussing earlier, other deps might bring down cryptography, Do we still want to keep it? | |
| 14:22:34 | sean-k-mooney | if we are currently using python loging in teh cleint the correct fix woudl be to keep oslo.log and port the code to use it | |
| 14:22:46 | sean-k-mooney | chandankumar: no we only ist direct deps | |
| 14:23:03 | sean-k-mooney | so if our indirect deps pull it in (openstacksdk likely will) | |
| 14:23:09 | sean-k-mooney | then thats ok | |
| 14:23:28 | chandankumar | for oslo.log, I will keep it and port the code to use it in a seperate patch. | |
| 14:24:26 | chandankumar | that's it on review | |
| 14:24:30 | chandankumar | Moving to bugs now | |
| 14:24:39 | chandankumar | #topic bugs | |
| 14:24:52 | chandankumar | #link Remove deprecated HTTPClient and VerifiedHTTPSConnection: https://bugs.launchpad.net/python-cyborgclient/+bug/2146351 | |
| 14:25:35 | chandankumar | VerifiedHTTPSConnection is used only at https://github.com/openstack/python-cyborgclient/blob/dcab1011666bcb9fc0e0a6062db3b6badb843ce9/cyborgclient/common/httpclient.py#L91 | |
| 14:25:41 | sean-k-mooney | we should be using the httpclint form the sdk exlusively | |
| 14:25:52 | sean-k-mooney | pror to that we woudl have used the keystoneauth one directly | |
| 14:26:38 | chandankumar | Do you have the link to httpclient from the sdk? | |
| 14:26:58 | sean-k-mooney | we shoudl not be using it directly in most case we shoudl be using the sdk functionaltiy | |
| 14:27:08 | sean-k-mooney | but its part of the connection object | |
| 14:27:41 | sean-k-mooney | when you create an openstack sdk connection it proves 3 levels of api access | |
| 14:27:52 | sean-k-mooney | a direct authencated http client for raw requests | |
| 14:28:11 | sean-k-mooney | a service/proxy api and a clout api wiht some cleanups | |
| 14:28:47 | sean-k-mooney | we shoudl be using that middel layer but if that is missing functionatliy fall back to the sdk's client since that handels the keystone auth for us | |
| 14:29:27 | sean-k-mooney | https://github.com/openstack/openstacksdk/tree/master/openstack/accelerator | |
| 14:29:40 | sean-k-mooney | cyborg has sdk supprot so we shoudl not need to use the raw client at all | |
| 14:31:06 | sean-k-mooney | we shoudl likely remvoe https://opendev.org/openstack/python-cyborgclient/src/branch/master/cyborgclient/v1 entirly by the way | |
| 14:31:18 | sean-k-mooney | can we file a bug to remvoe the remvoe v1 api | |
| 14:31:23 | sean-k-mooney | that was remvoed in 2018 | |
| 14:31:42 | chandankumar | ok | |
| 14:31:51 | sean-k-mooney | or perhaps 2019 but it well past the point where all non eol cyborg version have delete it | |
| 14:32:11 | sean-k-mooney | same for https://opendev.org/openstack/python-cyborgclient/src/branch/master/cyborgclient/osc/v1 | |
| 14:32:14 | chandankumar | I am trying to understand we need use https://github.com/openstack/openstacksdk/tree/master/openstack/accelerator in python cyborgclient that will gget us everything | |
| 14:32:23 | chandankumar | I will file a bug for v1 removal | |
| 14:32:48 | sean-k-mooney | correct the sdk has full supprot for the v2 api | |
| 14:33:03 | sean-k-mooney | so i think if we remvoe the v1 api we can remove the use of the raw http client entirly | |
| 14:33:14 | sean-k-mooney | i belvie https://opendev.org/openstack/python-cyborgclient/src/branch/master/cyborgclient/osc/v2 | |
| 14:33:21 | sean-k-mooney | only uses the sdk today | |
| 14:34:32 | chandankumar | Ok I will modify this bug then drop http raw client of sdk support | |
| 14:34:39 | chandankumar | *to drop | |
| 14:35:00 | sean-k-mooney | lets use it to track fully removing all supprot for the v1 api | |
| 14:36:04 | sean-k-mooney | that will remove all useage of https://opendev.org/openstack/python-cyborgclient/src/branch/master/cyborgclient/common/httpclient.py | |
| 14:36:08 | sean-k-mooney | and perhaps a few other files | |
| 14:37:24 | sean-k-mooney | if you want to split this up into multiple parts we coudl do that | |
| 14:38:11 | chandankumar | I have updated this bug to drop v1 api support | |
| 14:38:13 | sean-k-mooney | we shodul also be removing the native shell (when not used via osc) but that needs deprecation before we can do that | |
| 14:38:54 | sean-k-mooney | python-cyborgclient should eventuly only provdie the osc plugin | |
| 14:39:04 | sean-k-mooney | but we can do that in phases | |