containerd supports extending its functionality using most of its defined interfaces. This includes using a customized runtime, snapshotter, content store, and even adding gRPC interfaces.
containerd has a smart client architecture, meaning any functionality which is not required by the daemon is done by the client. This includes most high level interactions such as creating a container‘s specification, interacting with an image registry, or loading an image from tar. containerd’s Go client gives a user access to many points of extensions from creating their own options on container creation to resolving image registry names.
See containerd's Go documentation
External plugins allow extending containerd's functionality using an officially released version of containerd without needing to recompile the daemon to add a plugin.
containerd allows extensions through two methods:
containerd supports multiple container runtimes. Each container can be invoked with a different runtime.
When using the Container Runtime Interface (CRI) plugin, named runtimes can be defined in the containerd configuration file. When a container is run without specifying a runtime, the configured default runtime is used. Alternatively, a different named runtime can be specified explicitly when creating a container via CRI gRPC by selecting the runtime handler to be used.
When a client such as ctr or nerdctl creates a container, it can optionally specify a runtime and options to use. If a runtime is not specified, containerd will use its default runtime.
containerd invokes v2 runtimes as binaries on the system, which are used to start the shim process for containerd. This, in turn, allows containerd to start and manage those containers using the runtime shim api returned by the binary.
For more details on runtimes and shims, including how to invoke and configure them, see the runtime v2 documentation
A proxy plugin is configured using containerd‘s config file and will be loaded alongside the internal plugins when containerd is started. These plugins are connected to containerd using a local socket serving one of containerd’s gRPC API services. Each plugin is configured with a type and name just as internal plugins are.
Update the containerd config file, which by default is at /etc/containerd/config.toml. Add a [proxy_plugins] section along with a section for your given plugin [proxy_plugins.myplugin]. The address must refer to a local socket file which the containerd process has access to. The currently supported types are snapshot, content, and diff.
version = 2 [proxy_plugins] [proxy_plugins.customsnapshot] type = "snapshot" address = "/var/run/mysnapshotter.sock"
Implementing a proxy plugin is as easy as implementing the gRPC API for a service. For implementing a proxy plugin in Go, look at the go doc for content store service, snapshotter service, and diff service.
The following example creates a snapshot plugin binary which can be used with any implementation of containerd's Snapshotter interface
package main import ( "fmt" "net" "os" "google.golang.org/grpc" snapshotsapi "github.com/containerd/containerd/api/services/snapshots/v1" "github.com/containerd/containerd/v2/contrib/snapshotservice" "github.com/containerd/containerd/v2/plugins/snapshots/native" ) func main() { // Provide a unix address to listen to, this will be the `address` // in the `proxy_plugin` configuration. // The root will be used to store the snapshots. if len(os.Args) < 3 { fmt.Printf("invalid args: usage: %s <unix addr> <root>\n", os.Args[0]) os.Exit(1) } // Create a gRPC server rpc := grpc.NewServer() // Configure your custom snapshotter, this example uses the native // snapshotter and a root directory. Your custom snapshotter will be // much more useful than using a snapshotter which is already included. // https://godoc.org/github.com/containerd/containerd/snapshots#Snapshotter sn, err := native.NewSnapshotter(os.Args[2]) if err != nil { fmt.Printf("error: %v\n", err) os.Exit(1) } // Convert the snapshotter to a gRPC service, // example in github.com/containerd/containerd/contrib/snapshotservice service := snapshotservice.FromSnapshotter(sn) // Register the service with the gRPC server snapshotsapi.RegisterSnapshotsServer(rpc, service) // Listen and serve l, err := net.Listen("unix", os.Args[1]) if err != nil { fmt.Printf("error: %v\n", err) os.Exit(1) } if err := rpc.Serve(l); err != nil { fmt.Printf("error: %v\n", err) os.Exit(1) } }
Using the previous configuration and example, you could run a snapshot plugin with
# Start plugin in one terminal
$ go run ./main.go /var/run/mysnapshotter.sock /tmp/snapshots
# Use ctr in another
$ CONTAINERD_SNAPSHOTTER=customsnapshot ctr images pull docker.io/library/alpine:latest
$ tree -L 3 /tmp/snapshots
/tmp/snapshots
|-- metadata.db
`-- snapshots
`-- 1
|-- bin
|-- dev
|-- etc
|-- home
|-- lib
|-- media
|-- mnt
|-- proc
|-- root
|-- run
|-- sbin
|-- srv
|-- sys
|-- tmp
|-- usr
`-- var
18 directories, 1 file
containerd uses plugins internally to ensure that internal implementations are decoupled, stable, and treated equally with external plugins. To see all the plugins containerd has, use ctr plugins ls
$ ctr plugins ls TYPE ID PLATFORMS STATUS io.containerd.content.v1 content - ok io.containerd.snapshotter.v1 btrfs linux/amd64 ok io.containerd.snapshotter.v1 aufs linux/amd64 error io.containerd.snapshotter.v1 native linux/amd64 ok io.containerd.snapshotter.v1 overlayfs linux/amd64 ok io.containerd.snapshotter.v1 zfs linux/amd64 error io.containerd.metadata.v1 bolt - ok io.containerd.differ.v1 walking linux/amd64 ok io.containerd.gc.v1 scheduler - ok io.containerd.service.v1 containers-service - ok io.containerd.service.v1 content-service - ok io.containerd.service.v1 diff-service - ok io.containerd.service.v1 images-service - ok io.containerd.service.v1 leases-service - ok io.containerd.service.v1 namespaces-service - ok io.containerd.service.v1 snapshots-service - ok io.containerd.runtime.v1 linux linux/amd64 ok io.containerd.runtime.v2 task linux/amd64 ok io.containerd.monitor.v1 cgroups linux/amd64 ok io.containerd.service.v1 tasks-service - ok io.containerd.internal.v1 restart - ok io.containerd.grpc.v1 containers - ok io.containerd.grpc.v1 content - ok io.containerd.grpc.v1 diff - ok io.containerd.grpc.v1 events - ok io.containerd.grpc.v1 healthcheck - ok io.containerd.grpc.v1 images - ok io.containerd.grpc.v1 leases - ok io.containerd.grpc.v1 namespaces - ok io.containerd.grpc.v1 snapshots - ok io.containerd.grpc.v1 tasks - ok io.containerd.grpc.v1 version - ok io.containerd.grpc.v1 cri linux/amd64 ok
From the output all the plugins can be seen as well those which did not successfully load. In this case aufs and zfs are expected not to load since they are not support on the machine. The logs will show why it failed, but you can also get more details using the -d option.
$ ctr plugins ls -d id==aufs id==zfs
Type: io.containerd.snapshotter.v1
ID: aufs
Platforms: linux/amd64
Exports:
root /var/lib/containerd/io.containerd.snapshotter.v1.aufs
Error:
Code: Unknown
Message: modprobe aufs failed: "modprobe: FATAL: Module aufs not found in directory /lib/modules/4.17.2-1-ARCH\n": exit status 1
Type: io.containerd.snapshotter.v1
ID: zfs
Platforms: linux/amd64
Exports:
root /var/lib/containerd/io.containerd.snapshotter.v1.zfs
Error:
Code: Unknown
Message: path /var/lib/containerd/io.containerd.snapshotter.v1.zfs must be a zfs filesystem to be used with the zfs snapshotter
The error message which the plugin returned explains why the plugin was unable to load.
Plugins are configured using the [plugins] section of containerd's config. Every plugin can have its own section using the pattern [plugins."<plugin type>.<plugin id>"].
example configuration
version = 2 [plugins] [plugins."io.containerd.monitor.v1.cgroups"] no_prometheus = false
To see full configuration example run containerd config default. If you want to get the configuration combined with your configuration, run containerd config dump.
containerd has several configuration versions:
A configuration for Version 2 or 3 must specify the version version = 2 or version = 3 in the header, and must have fully qualified plugin IDs in the [plugins] section:
version = 3 [plugins] [plugins.'io.containerd.monitor.task.v1.cgroups'] no_prometheus = false
version = 2 [plugins] [plugins."io.containerd.monitor.v1.cgroups"] no_prometheus = false
A configuration with Version 1 may not have version header, and does not need fully qualified plugin IDs.
[plugins] [plugins.cgroups] no_prometheus = false