buildr/internal/containers/orchestrator.go

package containers

import (
	"context"
	"errors"
	"fmt"
	"io"
	"log/slog"
	"os"
	"path"
	"path/filepath"
	"runtime"
	"time"

	"github.com/docker/docker/api/types"
	"github.com/docker/docker/api/types/container"
	"github.com/docker/docker/api/types/network"
	"github.com/docker/docker/client"
	"github.com/docker/go-connections/nat"

	"code.icb4dc0.de/buildr/buildr/internal/archive"
	"code.icb4dc0.de/buildr/buildr/internal/ignore"
)

const defaultGRPCDialTimeout = 10 * time.Second

var (
	_                   Shutdowner = (*Orchestrator)(nil)
	architectureMapping            = map[string]string{
		"amd64":   "amd64",
		"x86_64":  "amd64",
		"aarch64": "arm64",
	}
)

func NewOrchestrator(ctx context.Context, cli *client.Client, ignorer *ignore.Ignorer) (*Orchestrator, error) {
	hostInfo, err := cli.Info(ctx)
	if err != nil {
		return nil, err
	}

	if runtime.GOOS != hostInfo.OSType {
		return nil, fmt.Errorf("buildr is running on %s, but Docker is running on %s", runtime.GOOS, hostInfo.OSType)
	}

	if runtime.GOARCH != architectureMapping[hostInfo.Architecture] {
		return nil, fmt.Errorf("buildr is compiled for %s, but Docker is running on %s", runtime.GOARCH, hostInfo.Architecture)
	}

	o := Orchestrator{
		client:           cli,
		ignorer:          ignorer,
		dockerHostOSType: hostInfo.OSType,
		dockerHostArch:   architectureMapping[hostInfo.Architecture],
	}
	return &o, nil
}

type Orchestrator struct {
	client  *client.Client
	ignorer *ignore.Ignorer

	dockerHostOSType string
	dockerHostArch   string
}

func (o *Orchestrator) BuildRContainer(ctx context.Context, spec *BuildRContainerSpec) (con Container, baseUrl string, err error) {
	conRef := &containerRef{
		shutdowner:   o,
		client:       o.client,
		resourceName: fmt.Sprintf("buildr-%s-%s", spec.ModuleName, spec.ID),
	}

	containerSetupCtx, cancel := context.WithCancel(ctx)
	defer cancel()

	if err := o.pullImage(containerSetupCtx, spec.Image); err != nil {
		return nil, "", err
	}

	if err := o.createContainerNetwork(containerSetupCtx, conRef); err != nil {
		return nil, "", err
	}

	var buildrExecutablePath string
	if e, err := os.Executable(); err != nil {
		return nil, "", fmt.Errorf("failed to determine buildr executable: %w", err)
	} else {
		buildrExecutablePath = e
	}

	_, buildrExecutableName := filepath.Split(buildrExecutablePath)

	containerRepoRoot, conSpec := spec.containerSpec(buildrExecutableName)

	if err := o.createContainer(containerSetupCtx, conSpec, conRef); err != nil {
		return nil, "", err
	}

	for i := range spec.ExtraBinaries {
		extraBin := spec.ExtraBinaries[i]
		spec.Content[path.Join(spec.BinariesDir, extraBin)] = path.Join("/", "opt", "buildr", "bin", extraBin)
	}

	spec.Content[buildrExecutablePath] = path.Join("/", "opt", "buildr", "bin", buildrExecutableName)

	if err := o.copyFilesToContainer(containerSetupCtx, spec.RepoRoot, conRef, spec.Content, containerRepoRoot); err != nil {
		return nil, "", err
	}

	if err := o.client.ContainerStart(containerSetupCtx, conRef.containerID, types.ContainerStartOptions{}); err != nil {
		return nil, "", err
	}

	if status, err := o.client.ContainerInspect(containerSetupCtx, conRef.containerID); err != nil {
		return nil, "", err
	} else if !status.State.Running {
		return nil, "", fmt.Errorf("container %s is not running: %d - %s", conRef.containerID, status.State.ExitCode, status.State.Error)
	}

	_, port, err := conRef.MappedPort(containerSetupCtx, "3000/tcp")
	if err != nil {
		return nil, "", err
	}

	baseUrl = fmt.Sprintf("http://localhost:%s", port)

	prober := NewContainerProbe(
		slog.Default(),
		conRef.containerID,
		NewGRPCHealthProbe(baseUrl),
		o.client,
	)

	if err = prober.WaitReady(containerSetupCtx); err != nil {
		return nil, "", fmt.Errorf("error occurred while waiting for container to be ready: %w", err)
	}

	return conRef, baseUrl, nil
}

func (o *Orchestrator) ShutdownContainer(ctx context.Context, con Container) error {
	ignoreNotFound := func(err error) error {
		if client.IsErrNotFound(err) {
			return nil
		}
		return err
	}

	return errors.Join(
		ignoreNotFound(o.client.ContainerStop(ctx, con.ID(), container.StopOptions{})),
		ignoreNotFound(o.client.ContainerRemove(ctx, con.ID(), types.ContainerRemoveOptions{})),
		ignoreNotFound(o.client.NetworkRemove(ctx, con.NetworkID())),
	)
}

func (o *Orchestrator) pullImage(ctx context.Context, image string) error {
	_, _, err := o.client.ImageInspectWithRaw(ctx, image)
	var shouldPullImage bool
	if err != nil {
		if client.IsErrNotFound(err) {
			shouldPullImage = true
		} else {
			return err
		}
	}

	if !shouldPullImage {
		return nil
	}

	slog.Info("Pulling image", slog.String("image_name", image))
	rc, err := o.client.ImagePull(ctx, image, types.ImagePullOptions{})
	if err != nil {
		return err
	}

	_, err = io.Copy(io.Discard, rc)

	return errors.Join(err, rc.Close())
}

func (o *Orchestrator) createContainerNetwork(ctx context.Context, conRef *containerRef) (err error) {
	var shouldCreateNetwork bool

	res, err := o.client.NetworkInspect(ctx, conRef.resourceName, types.NetworkInspectOptions{})
	if err != nil {
		if client.IsErrNotFound(err) {
			shouldCreateNetwork = true
		} else {
			return err
		}
	}

	if !shouldCreateNetwork {
		conRef.networkID = res.ID
		return nil
	}

	createOpts := types.NetworkCreate{
		CheckDuplicate: true,
	}

	if resp, err := o.client.NetworkCreate(ctx, conRef.resourceName, createOpts); err != nil {
		return err
	} else {
		conRef.networkID = resp.ID
		return nil
	}
}

func (o *Orchestrator) createContainer(ctx context.Context, spec *ContainerSpec, conRef *containerRef) error {
	ports, binding, err := nat.ParsePortSpecs(spec.ExposedPorts)
	if err != nil {
		return err
	}

	containerConfig := container.Config{
		Image:        spec.Image,
		WorkingDir:   spec.WorkDir(),
		ExposedPorts: ports,
		Entrypoint:   spec.Entrypoint,
		Cmd:          spec.Cmd,
	}

	if spec.User != "" {
		containerConfig.User = spec.User
	}

	for k, v := range spec.Env {
		containerConfig.Env = append(containerConfig.Env, k+"="+v)
	}

	hostConfig := container.HostConfig{
		AutoRemove:   spec.AutoRemove,
		Privileged:   spec.Privileged,
		CapAdd:       spec.Capabilities.Add,
		CapDrop:      spec.Capabilities.Drop,
		PortBindings: binding,
		Mounts:       spec.Mounts,
	}

	networkConfig := network.NetworkingConfig{
		EndpointsConfig: map[string]*network.EndpointSettings{
			conRef.resourceName: {
				NetworkID: conRef.networkID,
			},
		},
	}

	resp, err := o.client.ContainerCreate(ctx, &containerConfig, &hostConfig, &networkConfig, nil, conRef.resourceName)
	if err != nil {
		return err
	}

	conRef.containerID = resp.ID
	return nil
}

func (o *Orchestrator) copyFilesToContainer(
	ctx context.Context,
	repoRoot string,
	conRef *containerRef,
	content map[string]string,
	targetDir string,
) error {
	copyOpts := types.CopyToContainerOptions{}

	tarTmpFile, err := os.CreateTemp(os.TempDir(), "buildr-container-*")
	if err != nil {
		return fmt.Errorf("failed to create temp directory: %w", err)
	}

	defer func(tmpFileName string) {
		err = errors.Join(err, os.Remove(tmpFileName))
	}(tarTmpFile.Name())

	contentArchive := archive.Tar{Ignorer: o.ignorer}
	for src, dest := range content {
		if !path.IsAbs(src) {
			src = path.Join(repoRoot, src)
		}
		if !path.IsAbs(dest) {
			dest = path.Join(targetDir, dest)
		}

		if err := contentArchive.Add(src, dest); err != nil {
			return err
		}
	}

	if err := contentArchive.Write(tarTmpFile); err != nil {
		return err
	}

	if _, err := tarTmpFile.Seek(0, 0); err != nil {
		return fmt.Errorf("failed to reset file offset: %w", err)
	}

	if err := o.client.CopyToContainer(ctx, conRef.containerID, "/", tarTmpFile, copyOpts); err != nil {
		return fmt.Errorf("failed to copy files to container: %w", err)
	}

	return nil
}