New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
HDDMODEL is insufficiently flexible, cannot add buses #399
Comments
How about keeping it more generic and adding a variable like "HDDOPTIONS" which are added after HDDMODEL e.g. in 8eb1912#diff-ce37f278fec25235d9f905442ecd8e5aR429
|
AFAICS the choices are basically either:
I might be missing something (I'm hardly a qemu expert), but at least to my current understanding, what we need to add is a bus, and buses are just I might spend a bit of time today poking around the man pages and trying different commands and see what I come up with, I guess. |
typing off the top of my head (ie. I consider the opinion I am about to share as open to persuasion) I am nervous about an "add any qemu parameter you like" variable as it would be too easy to create very easily broken combinations I like the idea of adding specific A very abstracted 'add a specific bus' variable might be nice, but I dont think it'll be flexible enough |
I guess I missed an option 4, which might be viable - I'll look into it and see if I can send a patch. It's simply:
i.e. we could just have os-autoinst know all the reasonable values for |
I tend to prefer 4th. option. Make a list of supported bus values which are allowed as Do we use qemu device name as Later I would also like the availability to specify different buses and parameters for different drives. So to move from generic |
@aaannz I also considered the 'different settings for different drives' case, but figured since neither we nor you have any uses for it ATM, there's no point getting fancy. :) My aim, if possible, is to remain backwards-compatible - so all values for |
On the topic of multiple disks, btw, we already have |
Just to throw in my oppinion: I know the downsides, but I would favor option 1. I consider this command line creation through variables rather crude. |
Translated my last comment into https://progress.opensuse.org/issues/18466 |
In Fedora, we were trying to use the
HDDMODEL
variable for a SATA test. Long story short, it turns out that doesn't work. In os-autoinst 4.2, there was a bit of code that sneakily turned the SATA disk we were trying to add into a PATA disk. In os-autoinst 4.3, that's gone, but there's simply no way you can cause os-autoinst to construct a qemu invocation which will successfully include a SATA disk, because you can't add a SATA bus. The way HDDMODEL is set up, it's only capable of attaching a disk to a bus that the VM would have otherwise; you can't use HDDMODEL or any other variable I can find to add a bus to the VM, and nothing else causes os-autoinst VMs to have a SATA bus, so this just doesn't seem possible at present.There's more details on this at https://phab.qadevel.cloud.fedoraproject.org/T691 . It would be good if we could fix this somehow, I don't know how much abstraction we want to put in; the low-abstraction way would just be to have numbered variables (DEVICE1, DEVICE2...) for constructing
-device
arguments, a high level way might be to have variables likeSATA
orSCSI
which cause os-autoinst to add an appropriate-device
argument to add the specified bus.The text was updated successfully, but these errors were encountered: