How to commit paravirtprovider bugfix for multi vcpu scenarios with coreos-vagrant in virtualbox?

Oliver Kuntze

2018-05-03 05:51:32 UTC

Hi there,

while setting up a local dev environment using
https://github.com/coreos/coreos-vagrant I ran into the problem that
container linux only booted successfully when setting vm_cpu = 1.

With vm_cpu > 1 boot process hangs during vagrant up at "SSH auth method:
private key".
Starting the vm directly from within Oracle's virtualbox admin tool shows
that booting hangs at "x86: Booting SMP configuration:".

Reason is that coreos-vagrant doesn't set the vm's paravirtprovider
property.
Because of this the vm is created in virtualbox with paravirtualization
(under "system->acceleration) set to "legacy".
This iin turn causes the vm not to boot.

I added paravirtprovider paravirtprovider paravirtprovider

*vb.customize ["modifyvm", :id, "--paravirtprovider", "#{$vb_paravirtprovider}"]*
and set
*$vb_paravirtprovider = "default"*
in the config section.

That did the job. My vm is booting fine with multiple vcpus now.

Now my question is: What's the way to provide a fix like this so that it can be merged back into the coreos-vagrant repo?

I also made the update channel configurable in the vagrant config since it was hardcoded to use the alpha channel.

So probably my work is of use for other coreos-vagrant users under windows os. I am happy to share.

Cheers
Oliver