Awesome
libvirt
Package go-libvirt
provides a pure Go interface for interacting with libvirt.
Rather than using libvirt's C bindings, this package makes use of libvirt's RPC interface, as documented here. Connections to the libvirt server may be local, or remote. RPC packets are encoded using the XDR standard as defined by RFC 4506.
libvirt's RPC interface is quite extensive, and changes from one version to the next, so this project uses a pair of code generators to build the go bindings. The code generators should be run whenever you want to build go-libvirt for a new version of libvirt. See the next section for directions on re-generating go-libvirt.
Feel free to join us in #go-libvirt
on libera chat
if you'd like to discuss the project.
Running the Code Generators
The code generator doesn't run automatically when you build go-libvirt. It's meant to be run manually any time you change the version of libvirt you're using. When you download go-libvirt it will come with generated files corresponding to a particular version of libvirt. You can use the library as-is, but the generated code may be missing libvirt functions, if you're using a newer version of libvirt, or it may have extra functions that will return 'unimplemented' errors if you try to call them. If this is a problem, you should re-run the code generator. To do this, follow these steps:
- First, download a copy of the libvirt sources corresponding to the version you want to use.
- Change directories into where you've unpacked your distribution of libvirt.
- The second step depends on the version of libvirt you'd like to build against.
It's not necessary to actually build libvirt, but it is necessary to run libvirt's
"configure" step because it generates required files.
- For libvirt < v6.7.0:
$ mkdir build; cd build
$ ../autogen.sh
- For libvirt >= v6.7.0:
$ meson setup build
- For libvirt < v6.7.0:
- Finally, set the environment variable
LIBVIRT_SOURCE
to the directory you put libvirt into, and rungo generate ./...
from the go-libvirt directory. This runs both of the go-libvirt's code generators.
How to Use This Library
Once you've vendored go-libvirt into your project, you'll probably want to call some libvirt functions. There's some example code below showing how to connect to libvirt and make one such call, but once you get past the introduction you'll next want to call some other libvirt functions. How do you find them?
Start with the libvirt API reference.
Let's say you want to gracefully shutdown a VM, and after reading through the
libvirt docs you determine that virDomainShutdown() is the function you want to
call to do that. Where's that function in go-libvirt? We transform the names
slightly when building the go bindings. There's no need for a global prefix like
"vir" in Go, since all our functions are inside the package namespace, so we
drop it. That means the Go function for virDomainShutdown()
is just DomainShutdown()
,
and sure enough, you can find the Go function DomainShutdown()
in libvirt.gen.go,
with parameters and return values equivalent to those documented in the API
reference.
Suppose you then decide you need more control over your shutdown, so you switch
over to virDomainShutdownFlags()
. As its name suggests, this function takes a
flag parameter which has possible values specified in an enum called
virDomainShutdownFlagValues
. Flag types like this are a little tricky for the
code generator, because the C functions just take an integer type - only the
libvirt documentation actually ties the flags to the enum types. In most cases
though we're able to generate a wrapper function with a distinct flag type,
making it easier for Go tooling to suggest possible flag values while you're
working. Checking the documentation for this function:
godoc github.com/digitalocean/go-libvirt DomainShutdownFlags
returns this:
func (l *Libvirt) DomainShutdownFlags(Dom Domain, Flags DomainShutdownFlagValues) (err error)
If you want to see the possible flag values, godoc
can help again:
$ godoc github.com/digitalocean/go-libvirt DomainShutdownFlagValues
type DomainShutdownFlagValues int32
DomainShutdownFlagValues as declared in libvirt/libvirt-domain.h:1121
const (
DomainShutdownDefault DomainShutdownFlagValues = iota
DomainShutdownAcpiPowerBtn DomainShutdownFlagValues = 1
DomainShutdownGuestAgent DomainShutdownFlagValues = 2
DomainShutdownInitctl DomainShutdownFlagValues = 4
DomainShutdownSignal DomainShutdownFlagValues = 8
DomainShutdownParavirt DomainShutdownFlagValues = 16
)
DomainShutdownFlagValues enumeration from libvirt/libvirt-domain.h:1121
One other suggestion: most of the code in go-libvirt is now generated, but a few hand-written routines still exist in libvirt.go, and wrap calls to the generated code with slightly different parameters or return values. We suggest avoiding these hand-written routines and calling the generated routines in libvirt.gen.go instead. Over time these handwritten routines will be removed from go-libvirt.
Warning
While these package are reasonably well-tested and have seen some use inside of DigitalOcean, there may be subtle bugs which could cause the packages to act in unexpected ways. Use at your own risk!
In addition, the API is not considered stable at this time. If you would like
to include package libvirt
in a project, we highly recommend vendoring it into
your project.
Example
package main
import (
"fmt"
"log"
"net/url"
"github.com/digitalocean/go-libvirt"
)
func main() {
uri, _ := url.Parse(string(libvirt.QEMUSystem))
l, err := libvirt.ConnectToURI(uri)
if err != nil {
log.Fatalf("failed to connect: %v", err)
}
v, err := l.ConnectGetLibVersion()
if err != nil {
log.Fatalf("failed to retrieve libvirt version: %v", err)
}
fmt.Println("Version:", v)
flags := libvirt.ConnectListDomainsActive | libvirt.ConnectListDomainsInactive
domains, _, err := l.ConnectListAllDomains(1, flags)
if err != nil {
log.Fatalf("failed to retrieve domains: %v", err)
}
fmt.Println("ID\tName\t\tUUID")
fmt.Printf("--------------------------------------------------------\n")
for _, d := range domains {
fmt.Printf("%d\t%s\t%x\n", d.ID, d.Name, d.UUID)
}
if err = l.Disconnect(); err != nil {
log.Fatalf("failed to disconnect: %v", err)
}
}
Version: 1.3.4
ID Name UUID
--------------------------------------------------------
1 Test-1 dc329f87d4de47198cfd2e21c6105b01
2 Test-2 dc229f87d4de47198cfd2e21c6105b01
Example (Connect to libvirt via TLS over TCP)
package main
import (
"crypto/tls"
"crypto/x509"
"fmt"
"io/ioutil"
"log"
"github.com/digitalocean/go-libvirt"
"github.com/digitalocean/go-libvirt/socket/dialers"
)
func main() {
// This dials libvirt on the local machine
// It connects to libvirt via TLS over TCP
// To connect to a remote machine, you need to have the ca/cert/key of it.
// The private key is at ~/.pki/libvirt/clientkey.pem
// or /etc/pki/libvirt/private/clientkey.pem
// The Client Cert is at ~/.pki/libvirt/clientcert.pem
// or /etc/pki/libvirt/clientcert.pem
// The CA Cert is at ~/.pki/libvirt/cacert.pem
// or /etc/pki/CA/cacert.pem
// Use host name or IP which is valid in certificate
addr := "10.10.10.10"
l := libvirt.NewWithDialer(dialers.NewTLS(addr))
if err := l.Connect(); err != nil {
log.Fatalf("failed to connect: %v", err)
}
v, err := l.Version()
if err != nil {
log.Fatalf("failed to retrieve libvirt version: %v", err)
}
fmt.Println("Version:", v)
// Return both running and stopped VMs
flags := libvirt.ConnectListDomainsActive | libvirt.ConnectListDomainsInactive
domains, _, err := l.ConnectListAllDomains(1, flags)
if err != nil {
log.Fatalf("failed to retrieve domains: %v", err)
}
fmt.Println("ID\tName\t\tUUID")
fmt.Println("--------------------------------------------------------")
for _, d := range domains {
fmt.Printf("%d\t%s\t%x\n", d.ID, d.Name, d.UUID)
}
if err := l.Disconnect(); err != nil {
log.Fatalf("failed to disconnect: %v", err)
}
}
Running the Integration Tests
GitHub actions workflows are defined in .github/workflows and can be triggered manually in the GitHub UI after pushing a branch. There are not currently convenient scripts for setting up and running integration tests locally, but installing libvirt and defining only the artifacts described by the files in testdata should be sufficient to be able to run the integration test file against.