runq is a hypervisor-based Docker runtime based on runc to run regular Docker images as a lightweight KVM/Qemu virtual machine. The focus is on solving real problems, not on number of features.
Key differences to other hypervisor-based runtimes:
- minimalistic design, small code base
- no modification to existing Docker tools (dockerd, containerd, runc...)
- coexistence of runq containers and regular runc containers
- no extra state outside of Docker (no libvirt, no changes to /var/run/...)
- small init program, no systemd
- no custom guest kernel or custom qemu needed
- runs on x86_64 and s390x (>= z13)
runc container runq container
+-------------------------+ +-------------------------+
| | | VM |
| | | +---------------------+ |
| | | | | |
| | | | | |
| | | | | |
| application | | | application | |
| | | | | |
| | | | | |
| | | +---------------------+ |
| | | | guest kernel | |
| | | +---------------------+ |
| | | qemu |
+-------------------------+ +-------------------------+
----------------------------------------------------------------------
host kernel
runq requires a host kernel >= 4.8 with KVM and VHOST_VSOCK support enabled. The easiest way to build runq and to put all dependencies together is using Docker. For fast development cycles a regular build environment might be more efficient. For this refer to section Developing runq.
# get the runq and runc source code
git clone --recurse-submodules https://github.com/gotoz/runq.git
# compile and create a release tar file in a Docker container
cd runq
make release
# install runq to `/var/lib/runq`
make release-installRegister runq as Docker runtime with appropriate defaults. See daemon.json for more options.
/etc/docker/daemon.json
{
"runtimes": {
"runq": {
"path": "/var/lib/runq/runq",
"runtimeArgs": [
"--cpu", "1",
"--mem", "256",
"--dns", "8.8.8.8,8.8.4.4",
"--tmpfs", "/tmp"
]
}
}
}reload Docker config
systemctl reload docker.servicerunq-exec creates a secure connection between host and VM guests. Users of runq-exec are authenticated via a client certificate. Access to the client certificate must be limited to Docker users only.
The CA and server certificates must be installed in /var/lib/runq/qemu/certs.
Access must be limited to the root user only.
Examples of server and client TLS certificates can be created with the script:
/var/lib/runq/qemu/mkcerts.shNote: On x86 and s390x < z14 the host must provide sufficient entropy to the VM guests via virtio-rng. If there is not enough entropy available on the host booting of guests can fail with a timeout error. The entropy that's currently available can be checked with:
cat /proc/sys/kernel/random/entropy_availThe number returned should always be greater than 1000. On s390x >=z14 random data is provided by the hardware driven trng device (kernel module s390-trng).
The kernel module vhost_vsock must be loaded on the host. This can be achieved by creating
a config file for the systemd-modules-load service: /etc/modules-load.d/vhost-vsock.conf:
# Load vhost_vsock for runq
vhost_vsockthe simplest example
docker run --runtime runq -ti busybox shcustom VM with 512MiB memory and 2 CPUs
docker run --runtime runq -e RUNQ_MEM=512 -e RUNQ_CPU=2 -ti busybox shallow loading of extra kernel modules by adding the SYS_MODULE capability
docker run --runtime runq --cap-add sys_module -ti busybox sh -c "modprobe brd && lsmod"
full example PostgreSQL with custom storage
dd if=/dev/zero of=data.img bs=1M count=200
mkfs.ext4 -F data.img
docker run \
--runtime runq \
--name pgserver \
-e RUNQ_CPU=2 \
-e RUNQ_MEM=512 \
-e POSTGRES_PASSWORD=mysecret \
-v $PWD/data.img:/dev/runq/0001/none/ext4/var/lib/postgresql \
-d postgres:alpine
sleep 10
docker run \
--runtime runq \
--link pgserver:postgres \
--rm \
-e PGPASSWORD=mysecret \
postgres:alpine psql -h postgres -U postgres -c "select 42 as answer;"
# answer
# --------
# 42
# (1 row)
Containers running Systemd should have the environment variable RUNQ_SYSTEMD set to 1.
With RUNQ_SYSTEMD=1 the behaviour changes as follows:
-
The entry-point return code (Systemd exit code) must be treated differently to ensure that
poweroffandrebootexecuted inside the container work as expected especially to support Docker containers created with '--restart on-failure:1'. Therefore- on poweroff and halt the Systemd exit code is 2 but the return code as seen by Docker is 0.
- In all other cases the return code as seen by Docker is 1.
-
When stopping or restarting a container the signal sent to the docker-entry point is SIGTERM. Systemd usually ignores SIGTERM. In order to trigger a clean Systemd shutdown the signal is therefore modified from SIGTERM to SIGRTMIN+4. When Systemd receives SIGRTMIN+4 it starts the poweroff.target unit.
-
With
RUNQ_SYSTEMD=1Linux cgroups are not mounted because this will be done later by Systemd.
See test/examples/Dockerfile.systemd and test/examples/systemd.sh for an example.
Runq can write the container environment variables in a file named /.runqenv placed in
the root directory of the container. This might be useful for containers running Systemd
as entry point. This feature can be enabled globally by configuring --runqenv in
/etc/docker/daemon.json or for a single container by setting
the environment variable RUNQ_RUNQENV to a true value.
The default 9p cache mode is 'mmap' but can be configured by setting the global
runtime parameter --9pcache in /etc/docker/daemon.json
or for each container individually by setting the container environment variable RUNQ_9PCACHE.
Valid cache modes are none, loose, fscache and mmap. For details see 9prst.txt.
The default Qemu CPU model is 'host' with no flags (KVM processor with all supported host features) but can be
configured by setting the global runtime parameter --cpuargs in /etc/docker/daemon.json
or for each container individually by setting the container environment variable RUNQ_CPUARGS.
E.g. --env RUNQ_CPUARGS=host,rtm=off.
See qemu-system-x86_64 -cpu help for a list of available CPU models and CPUID flags.
docker cli
dockerd engine
docker-containerd-shim
runq container
+--------------------------------------------------------+
| |
docker0 | VM |
`veth <------> veth +--------------------------+ |
| `<--- macvtap ---|-> eth0 | |
| proxy <-----------------> init | |
runq-exec <-----------tls----------------> `vsockd | |
| |+-------------namespace--+| |
overlayfs <-----9pfs-------------------||-> / || |
| || || |
block dev <-----virtio-blk-------------||-> /dev/vdx || |
| || || |
| || || |
| || || |
| || application || |
| || || |
| |+------------------------+| |
| | guest kernel | |
| +--------------------------+ |
| qemu |
+--------------------------------------------------------+
--------------------------------------------------------------------------
host kernel
-
cmd/runq
- new docker runtime
-
cmd/proxy
- new Docker entry point
- first process in container (PID 1)
- configures and starts Qemu (network, disks, ...)
- forwards signals to VM init
- receives application exit code
-
cmd/init
- first process in VM (PID 1)
- initializes the VM guest (network, disks, ...)
- starts entry-point in PID and Mount namespace
- sends signals to target application
- forwards application exit code back to proxy
-
cmd/runq-exec
- command line utility similar to docker exec
-
cmd/nsenter
- enters the namespaces of entry-point for runq-exec
-
qemu
- creates
/var/lib/runq/qemu - read-only volume attached to every container
- contains qemu rootfs (proxy, qemu, kernel and initrd)
- creates
-
initrd
- prepares the initrd to boot the VM
-
pkg
- helper packages
runq-exec (/var/lib/runq/runq-exec) is a command line utility similar to docker exec. It allows running
additional commands in existing runq containers executed from the host. It uses
VirtioVsock for the communication
between host and VMs. TLS is used for encryption and client authorization. Support for
runq-exec can be disabled by setting the container environment variable RUNQ_NOEXEC
to a true value or by --noexec in /etc/docker/daemon.json.
Usage:
runq-exec [options] <container> command args
Run a command in a running runq container
Options:
-c, --tlscert string TLS certificate file (default "/var/lib/runq/cert.pem")
-k, --tlskey string TLS private key file (default "/var/lib/runq/key.pem")
-e, --env stringArray Set environment variables for command
-h, --help Print this help
-i, --interactive Keep STDIN open even if not attached
-t, --tty Allocate a pseudo-TTY
-v, --version Print version
Environment Variable:
DOCKER_HOST specifies the Docker daemon socket.
Example:
runq-exec -ti a6c3b7c bash
runq runs Qemu and Linux Kernel from the /var/lib/runq/qemu directory
on the host. This directory is populated by make -C qemu. For simplicity
Qemu and the Linux kernel are taken from the Ubuntu 20.04 LTS Docker base image.
See qemu/x86_64/Dockerfile for details.
This makes runq independent of the Linux distribution on the host.
Qemu does not need to be installed on the host.
The kernel modules directory (/var/lib/runq/qemu/lib/modules)
is bind-mounted into every container to /lib/modules.
This allows the loading of extra kernel modules in any container if needed.
For this SYS_MODULES capability is required (--cap-add sys_modules).
runq uses Macvtap devices to connect Qemu VirtIO interfaces to Docker bridges. By default a single Ethernet interface is created. Multiple networks can be used by connecting a container to the networks before start. See test/integration/net.sh as an example.
runq container can also be connected to one or more Docker networks of type Macvlan. This allows a direct connection between the VM and the physical host network without bridge and without NAT. See https://docs.docker.com/network/macvlan/ for details.
For custom networks the docker daemon implements an embedded DNS server which provides built-in service discovery for any container created with a valid container name. This Docker DNS server (listen address 127.0.0.11:53) is reachable only by runc containers and not by runq containers. A work-around is to run one or more DNS proxy container in the custom network with runc and use the proxy IP address for DNS of runq containers. See test/examples/dnsproxy.sh for details on how to setup a DNS proxy.
DNS configuration without proxy can be done globally via runtime options specified in
'/etc/docker/daemon.json' (see example above) or via environment variables for each
container at container start.
The environment variables are RUNQ_DNS, RUNQ_DNS_OPT and RUNQ_DNS_SEARCH.
Environment variables have priority over global options.
Setting the environment variable RUNQ_DNS_PRESERVE to "1" completely disables generation of /etc/resolv.conf by runq.
Extra storage can be added in the form of Qcow2 images, raw file images or
regular block devices. Storage devices will be mounted automatically if
a filesystem and a mount point has been specified.
Supported filesystems are ext2, ext3, ext4, xfs and btrfs.
Cache type must be writeback, writethrough, none or unsafe.
Cache type "none" is recommended for filesystems that support O_DIRECT.
See man qemu(1) for details about different cache types.
Syntax:
--volume <image name>:/dev/runq/<id>/<cache type>[/<filesystem type><mount point>]
--device <device name>:/dev/runq/<id>/<cache type>[/<filesystem type><mount point>]<id> is used to create symbolic links inside the VM guest that point to the Qemu Virtio device
files. The id can be any character string that matches the regex pattern "^[a-zA-Z0-9-_]{1,36}$"
but it must be unique within a container.
/dev/disk/by-runq-id/0001 -> ../../vdaMount the existing Qcow image /data.qcow2 with xfs filesystem to /mnt/data:
docker run -v /data.qcow2:/dev/runq/0001/none/xfs/mnt/data ...Attach the host device /dev/sdb1 formatted with ext4 to /mnt/data2:
docker run --device /dev/sdb1:/dev/runq/0002/writethrough/ext4/mnt/data2 ...Attach the host device /dev/sdb2 without mounting:
docker run --device /dev/sdb2:/dev/runq/0003/writethrough ...A block device or a raw file with an EXT2 or EXT4 filesystem can be used as rootdisk
of the VM. On first boot of the container the content of the Docker image is copied into the rootdisk.
The block device or raw file will then be used as root filesystem via virtio-blk instead of 9pfs. But be aware that changes to the root filesystem will not be reflected in the source docker container filesystem. (docker cp will no longer work as expected)
# existing block device with empty ext4 filesystem
docker run --runtime runq --device /dev/sdb1:/dev/runq/0001/none/ext4 -e RUNQ_ROOTDISK=0001 -ti alpine sh
# new raw file
fallocate -l 1G disk.raw
mkfs.ext4 disk.raw
docker run --runtime runq --volume $PWD/disk.raw:/dev/runq/0001/none/ext4 -e RUNQ_ROOTDISK=0001 -ti alpine shDirectories can be excluded from being copied with the RUNQ_ROOTDISK_EXCLUDE environment
variable. E.g. -e RUNQ_ROOTDISK_EXCLUDE="/foo,/bar"
See Dockerfile.rootdisk and rootdisk.sh as a further example.
By default runq drops all capabilities except those needed (same as regular Docker does). The white list of the remaining capabilities is provided by the Docker engine.
AUDIT_WRITE CHOWN DAC_OVERRIDE FOWNER FSETID KILL MKNOD NET_BIND_SERVICE NET_RAW SETFCAP SETGID SETPCAP SETUID SYS_CHROOT
See man capabilities for a list of all available capabilities.
Additional Capabilities can be added to the white list at container start:
docker run --cap-add SYS_TIME --cap-add SYS_MODULE ...`runq supports the default Docker seccomp profile as well as custom profiles.
docker run --security-opt seccomp=<profile-file> ...The default profile is defined by the Docker daemon and gets applied automatically. Note: Only the runq init binary is statically linked against libseccomp. Therefore libseccomp is needed only at compile time.
If the host operating system where runq is being built does not provide static libseccomp libraries one can also simply build and install libseccomp from the sources.
Seccomp can be disabled at container start:
docker run --security-opt seccomp=unconfined ...Note: Some Docker daemon don't support custom Seccomp profiles. Run docker info to verify
that Seccomp is supported by your daemon. If it is supported the output of docker info looks like this:
Security Options:
seccomp
Profile: default
AP devices provide cryptographic functions to all CPUs assigned to a Linux system running in
an IBM Z system LPAR. AP devices can be made available to a runq container by passing a VFIO mediated
device from the host through Qemu into the runq VM guest. VFIO mediated devices are enabled by the
vfio_ap kernel module and allow for partitioning of AP devices and domains. The environment variable RUNQ_APUUID specifies the VFIO mediated device UUID. runq automatically loads the required zcrypt kernel modules inside the VM. E.g.:
docker run --runtime runq -e RUNQ_APUUID=b34543ee-496b-4769-8312-83707033e1de ...For details on how to setup mediated devices on the host see https://www.kernel.org/doc/html/latest/s390/vfio-ap.html
Most docker commands and options work as expected. However, due to the fact that the target application runs inside a Qemu VM which itself runs inside a Docker container and because of the minimalistic design principle of runq some docker commands and options don't work. E.g:
- adding / removing networks and storage dynamically
- docker exec (see runq-exec)
- docker swarm
- privileged mode
- apparmor, selinux, ambient
- docker HEALTHCHECK
The following common options of docker run are supported:
--attach --name
--cap-add --network
--cap-drop --publish
--cpus --restart
--cpuset-cpus --rm
--detach --runtime
--entrypoint --sysctl
--env --security-opt seccomp=unconfined
--env-file --security-opt no-new-privileges
--expose --security-opt seccomp=<filter-file>
--group-add --tmpfs
--help --tty
--hostname --ulimit
--init --user
--interactive --volume
--ip --volumes-from
--link --workdir
--mount
A nested VM is a virtual machine that runs inside of a virtual machine. In plain KVM this feature is
considered working but not meant for production use. Running KVM guests inside guests of other
hypervisors such as VMware might not work as expected or might not work at all.
However to try out runq in a VM guest the (experimental) runq runtime configuration parameter
--nestedvm can be used. It modifies the parameters of the Qemu process.
For fast development cycles runq can be build on the host as follows:
-
Prerequisites:
- Docker >= 19.03.x-ce
- Go >= 1.20
/var/lib/runqmust be writable by the current user- Libseccomp static library.
E.g.
libseccomp-devfor Ubuntu orlibseccomp-staticfor Fedora
-
Download runq and runc source code
git clone --recurse-submodules https://github.com/gotoz/runq.git
-
Install Qemu and guest kernel to
/var/lib/runq/qemu. All files are taken from the Ubuntu 20.04 LTS Docker base image. (/var/lib/runqmust be writeable by the current user.)cd runq make -C qemu all install -
Compile and install runq components to
/var/lib/runqmake install
-
Create TLS certificates
/var/lib/runq/qemu/mkcerts.sh
-
Adjust file and directory permissions
sudo chown -R root:root /var/lib/runq
-
Register runq as Docker runtime with appropriate defaults as shown in section Installation above.
See CONTRIBUTING for details.
The code is licensed under the Apache License 2.0.
See LICENSE for further details.