Custom configuration possible?

[jmharris]

Hi, thanks for your work on Proxmox. There are a couple of things I'd like to be able to do and I'm not sure if they're already possible (compared to launching a KVM process manually).

* Attaching block devices (and iSCSI volumes) directly to a VM. I understand this is in the roadmap for 2.0 (and should be somewhat available in 1.4beta due soon). Is there any way of manually specifying this in the meanwhile? Editing config files directly is fine.

* In addition to the above I need to supply kernel and initrd params to the KVM process. I have not seen this on the roadmap- can I accomplish this now or is it a planned feature?

 

[tom]

Hi, thanks for your work on Proxmox. There are a couple of things I'd like to be able to do and I'm not sure if they're already possible (compared to launching a KVM process manually).

* Attaching block devices (and iSCSI volumes) directly to a VM. I understand this is in the roadmap for 2.0 (and should be somewhat available in 1.4beta due soon). Is there any way of manually specifying this in the meanwhile? Editing config files directly is fine.

the plan is to release the beta next week.

* In addition to the above I need to supply kernel and initrd params to the KVM process. I have not seen this on the roadmap- can I accomplish this now or is it a planned feature?

can you describe this a bit more detailed - what do you want to achieve with this?

 

[jmharris]

the plan is to release the beta next week.

Great! Look forward to trying it.

can you describe this a bit more detailed - what do you want to achieve with this?

We have many VMs running under Xen where the root fs storage is an LVM volume. The kernel and initrd are stored seperately and booted by xen upon domain creation, supplying the block device as the root fs to the kernel params.

The KVM equivalent seems to work well, allowing us to migrate with no VM changes. For example I can use the following command (where /dev/sdb is the block device containing the root fs):

kvm -vnc :0 -m 800 -k en-gb -kernel vmlinuz-2.6.26-1-xen-686 -append root=/dev/sdb ro -drive if=ide,boot=on,file=/dev/sdb -initd initrd.img-2.6.26-1-xen-686

The problem with this being that there is no management of the KVM process- a problem I'm hoping Proxmox can solve.

 

[dietmar]

The problem with this being that there is no management of the KVM process- a problem I'm hoping Proxmox can solve.

We added an 'args' options to the kvm configuration file to pass arbitrary arguments directly to kvm.

 

[jmharris]

Thanks for supporting this use case. Is this something you have added recently and will be available in the next release, or is that something already present in 1.3?

I presume you mean that to support this I will need to create the VM via the web UI and then edit the cfg file on disk (e.g. /etc/qemu-server/101.conf). That is fine, but are there any plans to add support for managing the kernel and initrd images via the web ui or is this not a recommended way of structuring VMs?

 

[dietmar]

That is fine, but are there any plans to add support for managing the kernel and initrd images via the web ui

no

or is this not a recommended way of structuring VMs?

This options uses local files, so you will have problems in a cluster when you migrate the VM.

 

[jmharris]

This options uses local files, so you will have problems in a cluster when you migrate the VM.

This is a good point, thanks. We will look at how we can migrate our older Xen images so they contain a bootloader.

I can see now that the args option is already supported from the wiki FAQ page, thanks.