I needed a non-libvirt qemu driver, so this is it.
It worked in M1 Mac already but with user networking mode only.
But I needed to create seperated network topology for guest instances.
This is edited driver that support vmnet-*. (It requires sudo privileges)
from @SvenDowideit
Its initial use is going to be for running the Rancher OS tests, but maybe you'll find a use for it too.
from @fventuri
Create machines locally using QEMU. This driver requires QEMU to be installed on your host.
$ docker-machine create --driver=qemu qemu-test
Options:
--qemu-boot2docker-url: The URL of the boot2docker image. Defaults to the latest available version.--qemu-disk-size: Size of disk for the host in MB. Default:20000--qemu-memory: Size of memory for the host in MB. Default:1024--qemu-cpu-count: Number of CPUs. Default:1--qemu-program: Name of the qemu program to run. Default:qemu-system-x86_64--qemu-display: Show the graphical display output to the user. Default: false--qemu-display-type: Select type of display to use (sdl/vnc=localhost:0/etc)--qemu-nographic: Use -nographic instead of -display none. Default: false--qemu-virtio-drives: Use virtio for drives (cdrom and disk). Default: false--qemu-network: Networking to be used: vmnet-shared, vmnet-host or vmnet-bridged. Default:vmnet-shared--qemu-network-bridge: Name of the network bridge to be used for networking. (for vmnet-bridged) Default:en0--qemu-network-vmnet-id: The vmnet id to use for networking. Default:net0
The --qemu-boot2docker-url flag takes a few different forms. By
default, if no value is specified for this flag, Machine will check locally for
a boot2docker ISO. If one is found, that will be used as the ISO for the
created machine. If one is not found, the latest ISO release available on
boot2docker/boot2docker will be
downloaded and stored locally for future use. Note that this means you must run
docker-machine upgrade deliberately on a machine if you wish to update the "cached"
boot2docker ISO.
This is the default behavior (when --qemu-boot2docker-url=""), but the
option also supports specifying ISOs by the http:// and file:// protocols.
file:// will look at the path specified locally to locate the ISO: for
instance, you could specify --qemu-boot2docker-url file://$HOME/Downloads/rc.iso to test out a release candidate ISO that you have
downloaded already. You could also just get an ISO straight from the Internet
using the http:// form.
Note that when using virtio the drives will be mounted as /dev/vda and /dev/vdb,
instead of the usual /dev/cdrom and /dev/sda, since they are using paravirtualization.
Environment variables:
Here comes the list of the supported variables with the corresponding options. If both environment variable and CLI option are provided the CLI option takes the precedence.
| Environment variable | CLI option |
|---|---|
QEMU_BOOT2DOCKER_URL |
--qemu-boot2docker-url |
QEMU_VIRTIO_DRIVES |
--qemu-virtio-drives |
It strongly depends on arp -a command to resolve IP from Mac address.
Not tested on other systems, it assumes output of arp -a will be list of ? <IP> <MAC>.
UPDATED: It is updated to read and parse DHCP leases file but still not tested on various systems.
Planning to use QMP to solve it.