Merge pull request #7 from LINUXexpert-org/docker-stage-image

Stage a container image the way a binary is staged
2026-08-28 17:23:41 -07:00
parent f34cd95454 0a6b3ad173
commit d3d596b3c2
3 changed files with 380 additions and 0 deletions
@@ -63,6 +63,20 @@ func cmp(a, b int) int {
 // automates - see ARCHITECTURE.md §1/§4.1.
 var minSupportedSource = semver{0, 15, 0}

+// VersionFromOutput extracts a semver from whatever a Stalwart build
+// printed when asked for its version. Exported because the same question
+// gets asked of a container image (internal/stage), where the command is
+// `docker run <image> --version` rather than the binary directly - and the
+// answer has to be parsed identically or the two paths could disagree
+// about what they staged.
+func VersionFromOutput(out string) (string, error) {
+	v, err := parseSemver(out)
+	if err != nil {
+		return "", err
+	}
+	return v.String(), nil
+}
+
 // DetectVersion runs the installed binary's --version flag and extracts a
 // semver from its output.
 func DetectVersion(ctx context.Context, binaryPath string) (string, error) {
@@ -0,0 +1,166 @@
+// SPDX-FileCopyrightText: 2026 LINUXexpert-org
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+package stage
+
+import (
+	"bytes"
+	"context"
+	"fmt"
+	"os/exec"
+	"strings"
+
+	"github.com/LINUXexpert-org/stalwart-migrator/internal/checkpoint"
+	"github.com/LINUXexpert-org/stalwart-migrator/internal/preflight"
+)
+
+// DefaultDockerBinary is what shells out unless a caller names another.
+const DefaultDockerBinary = "docker"
+
+// ImageOptions configures staging a container image - the container
+// deployment's answer to downloading a binary. See issue #3.
+type ImageOptions struct {
+	// Image is the target image, named in full by the operator:
+	// "stalwartlabs/stalwart:v0.16.14". It is deliberately never derived
+	// from the running container's image by swapping the tag. That
+	// derivation is wrong for a digest-pinned image, wrong for a mirror,
+	// and wrong for a fork - and being wrong here means pulling the wrong
+	// software into a mail server, which is not a mistake worth a
+	// convenience.
+	Image string
+
+	// TargetVersion is what the image must report, e.g. "0.16.14".
+	TargetVersion string
+
+	// SkipPull uses an image already present locally. For an air-gapped
+	// host that loaded it from a tarball, where a pull cannot work and its
+	// failure would say nothing useful.
+	SkipPull bool
+
+	DockerBinary string
+}
+
+func (o ImageOptions) docker() string {
+	if o.DockerBinary == "" {
+		return DefaultDockerBinary
+	}
+	return o.DockerBinary
+}
+
+// StagedImage is a target image that is present locally and has been asked
+// what it is.
+type StagedImage struct {
+	// Ref is what to actually run: the image's own ID, not the tag it
+	// arrived under. A tag can move between staging and cutover - that is
+	// the whole reason `latest` is a hazard - and running the tag later
+	// would run something other than what was verified here.
+	Ref string
+
+	// Tag is what the operator named, kept for reports.
+	Tag string
+
+	// Version is what the image reported, which is the only thing that
+	// settles what was pulled.
+	Version string
+}
+
+// RunImage pulls the target image and confirms it is the version that was
+// asked for, checkpointed as the stage phase's "stage-image" step.
+//
+// The version check is the point of the phase, exactly as it is for a
+// binary: the tag, the registry and the repository name are all assumptions
+// about someone else's publishing process, and the image's own answer is
+// the only thing that settles what arrived.
+//
+// One caveat, recorded rather than hidden. Asking an image its version
+// means running it with --version, which assumes its entrypoint is the
+// server and passes flags through. That holds for an image built the
+// obvious way and has not been confirmed against a published Stalwart
+// image - there was none to hand when this was written. If it turns out
+// not to hold, this fails loudly with the image's own output rather than
+// staging something unverified, and the fix is a way to name the command
+// rather than a reason to skip the check.
+func RunImage(ctx context.Context, store *checkpoint.Store, rs *checkpoint.RunState, opts ImageOptions) (StagedImage, error) {
+	if opts.Image == "" {
+		return StagedImage{}, fmt.Errorf("stage: no target image named - pass the image to migrate to; it is never guessed from the running container")
+	}
+	if opts.TargetVersion == "" {
+		return StagedImage{}, fmt.Errorf("stage: no target version to check %s against", opts.Image)
+	}
+
+	var staged StagedImage
+	_, err := store.RunStep(rs, checkpoint.PhaseStage, "stage-image", func() (checkpoint.StepOutcome, error) {
+		if !opts.SkipPull {
+			if out, err := run(ctx, opts.docker(), "pull", opts.Image); err != nil {
+				return checkpoint.StepOutcome{}, fmt.Errorf("stage: pull %s: %w (%s)", opts.Image, err, out)
+			}
+		}
+
+		id, err := run(ctx, opts.docker(), "image", "inspect", "-f", "{{.Id}}", opts.Image)
+		if err != nil {
+			return checkpoint.StepOutcome{}, fmt.Errorf("stage: %s is not present locally after pull: %w (%s)", opts.Image, err, id)
+		}
+		id = strings.TrimSpace(id)
+		if id == "" {
+			return checkpoint.StepOutcome{}, fmt.Errorf("stage: docker reported no image ID for %s", opts.Image)
+		}
+
+		// The image's own entrypoint first, which is what an image built
+		// the obvious way answers to. Only if that fails is the entrypoint
+		// overridden to call the binary by name - a second guess, tried
+		// because failing the whole phase over an image that merely wraps
+		// its server in a shim would be worse than one extra attempt.
+		out, err := run(ctx, opts.docker(), "run", "--rm", id, "--version")
+		if err != nil {
+			out, err = run(ctx, opts.docker(), "run", "--rm", "--entrypoint", "stalwart", id, "--version")
+		}
+		if err != nil {
+			return checkpoint.StepOutcome{}, fmt.Errorf(
+				"stage: %s would not report its version: %w (output: %s). This tool will not stage an image it cannot identify",
+				opts.Image, err, strings.TrimSpace(out))
+		}
+
+		got, err := preflight.VersionFromOutput(out)
+		if err != nil {
+			return checkpoint.StepOutcome{}, fmt.Errorf("stage: parse version from %s (%q): %w", opts.Image, strings.TrimSpace(out), err)
+		}
+		wanted := strings.TrimPrefix(opts.TargetVersion, "v")
+		if got != wanted {
+			return checkpoint.StepOutcome{}, fmt.Errorf(
+				"stage: image %s reports %s but %s was asked for - the tag does not contain what it claims",
+				opts.Image, got, wanted)
+		}
+
+		staged = StagedImage{Ref: id, Tag: opts.Image, Version: got}
+		return checkpoint.StepOutcome{
+			Detail: fmt.Sprintf("staged %s as %s, which reports %s", opts.Image, shortRef(id), got),
+			Extra:  id,
+		}, nil
+	})
+	if err != nil {
+		return StagedImage{}, err
+	}
+	if staged.Ref == "" {
+		// A resumed run skipped the step, so the values above were never
+		// assigned. The recorded ID is what that run verified.
+		staged = StagedImage{Ref: rs.Outcome(checkpoint.PhaseStage, "stage-image").Extra, Tag: opts.Image, Version: opts.TargetVersion}
+	}
+	return staged, nil
+}
+
+func run(ctx context.Context, name string, args ...string) (string, error) {
+	cmd := exec.CommandContext(ctx, name, args...)
+	var out bytes.Buffer
+	cmd.Stdout = &out
+	cmd.Stderr = &out
+	err := cmd.Run()
+	return out.String(), err
+}
+
+func shortRef(id string) string {
+	id = strings.TrimPrefix(id, "sha256:")
+	if len(id) > 12 {
+		return id[:12]
+	}
+	return id
+}
@@ -0,0 +1,200 @@
+// SPDX-FileCopyrightText: 2026 LINUXexpert-org
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+package stage
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+// fakeDockerImage installs a `docker` that records its arguments and
+// answers each subcommand as scripted. versionBody is what `docker run`
+// prints for the version probe.
+func fakeDockerImage(t *testing.T, pullExit, imageID, versionBody string) (log string) {
+	t.Helper()
+	dir := t.TempDir()
+	log = filepath.Join(dir, "docker-args.log")
+	script := fmt.Sprintf(`#!/bin/sh
+echo "$@" >> %q
+case "$1" in
+  pull) exit %s ;;
+  image) echo %q ;;
+  run) %s ;;
+esac
+`, log, pullExit, imageID, versionBody)
+	if err := os.WriteFile(filepath.Join(dir, "docker"), []byte(script), 0o755); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("PATH", dir+string(os.PathListSeparator)+os.Getenv("PATH"))
+	return log
+}
+
+const testImageID = "sha256:abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789"
+
+func TestRunImageStagesAndVerifies(t *testing.T) {
+	log := fakeDockerImage(t, "0", testImageID, "echo 'stalwart 0.16.14'")
+	store, rs := newRun(t)
+
+	staged, err := RunImage(context.Background(), store, rs, ImageOptions{
+		Image: "stalwartlabs/stalwart:v0.16.14", TargetVersion: "0.16.14",
+	})
+	if err != nil {
+		t.Fatalf("RunImage: %v", err)
+	}
+	if staged.Version != "0.16.14" {
+		t.Errorf("Version = %q", staged.Version)
+	}
+	// The ID, not the tag: a tag can move between staging and cutover, and
+	// running the tag later would run something else.
+	if staged.Ref != testImageID {
+		t.Errorf("Ref = %q, want the image ID %q", staged.Ref, testImageID)
+	}
+	if got := readFile(t, log); !strings.Contains(got, "pull stalwartlabs/stalwart:v0.16.14") {
+		t.Errorf("image was never pulled:\n%s", got)
+	}
+}
+
+// The version check is the point of the phase. An image whose tag claims
+// one version and whose binary reports another must not be staged.
+func TestRunImageRefusesAVersionMismatch(t *testing.T) {
+	fakeDockerImage(t, "0", testImageID, "echo 'stalwart 0.16.9'")
+	store, rs := newRun(t)
+
+	_, err := RunImage(context.Background(), store, rs, ImageOptions{
+		Image: "stalwartlabs/stalwart:v0.16.14", TargetVersion: "0.16.14",
+	})
+	if err == nil {
+		t.Fatal("expected a refusal when the image reports a different version")
+	}
+	if !strings.Contains(err.Error(), "does not contain what it claims") {
+		t.Errorf("error should name the mismatch, got: %v", err)
+	}
+}
+
+// An image that will not say what it is cannot be staged - the alternative
+// is migrating a mail server with software nothing identified.
+func TestRunImageRefusesAnImageThatWontReportItsVersion(t *testing.T) {
+	fakeDockerImage(t, "0", testImageID, "echo 'no such flag' >&2; exit 1")
+	store, rs := newRun(t)
+
+	_, err := RunImage(context.Background(), store, rs, ImageOptions{
+		Image: "img:1", TargetVersion: "0.16.14",
+	})
+	if err == nil {
+		t.Fatal("expected a refusal when the image will not report a version")
+	}
+	if !strings.Contains(err.Error(), "will not stage an image it cannot identify") {
+		t.Errorf("error should explain the refusal, got: %v", err)
+	}
+}
+
+// If the plain entrypoint refuses the flag, the binary is tried by name
+// before giving up.
+func TestRunImageFallsBackToNamingTheBinary(t *testing.T) {
+	dir := t.TempDir()
+	log := filepath.Join(dir, "docker-args.log")
+	// Fails without --entrypoint, succeeds with it.
+	script := fmt.Sprintf(`#!/bin/sh
+echo "$@" >> %q
+case "$1" in
+  pull) exit 0 ;;
+  image) echo %q ;;
+  run) for a in "$@"; do if [ "$a" = "--entrypoint" ]; then echo 'stalwart 0.16.14'; exit 0; fi; done; exit 1 ;;
+esac
+`, log, testImageID)
+	if err := os.WriteFile(filepath.Join(dir, "docker"), []byte(script), 0o755); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("PATH", dir+string(os.PathListSeparator)+os.Getenv("PATH"))
+
+	store, rs := newRun(t)
+	staged, err := RunImage(context.Background(), store, rs, ImageOptions{
+		Image: "img:1", TargetVersion: "0.16.14",
+	})
+	if err != nil {
+		t.Fatalf("RunImage should have fallen back to --entrypoint: %v", err)
+	}
+	if staged.Version != "0.16.14" {
+		t.Errorf("Version = %q", staged.Version)
+	}
+}
+
+func TestRunImageRefusesAFailedPull(t *testing.T) {
+	fakeDockerImage(t, "1", testImageID, "echo 'stalwart 0.16.14'")
+	store, rs := newRun(t)
+
+	if _, err := RunImage(context.Background(), store, rs, ImageOptions{
+		Image: "img:1", TargetVersion: "0.16.14",
+	}); err == nil {
+		t.Fatal("expected a refusal when the pull failed")
+	}
+}
+
+// SkipPull is for a host that loaded the image from a tarball, where a
+// pull cannot work and its failure would say nothing useful.
+func TestSkipPullUsesALocalImage(t *testing.T) {
+	log := fakeDockerImage(t, "1", testImageID, "echo 'stalwart 0.16.14'")
+	store, rs := newRun(t)
+
+	if _, err := RunImage(context.Background(), store, rs, ImageOptions{
+		Image: "img:1", TargetVersion: "0.16.14", SkipPull: true,
+	}); err != nil {
+		t.Fatalf("SkipPull should not have pulled: %v", err)
+	}
+	if got := readFile(t, log); strings.Contains(got, "pull ") {
+		t.Errorf("SkipPull still pulled:\n%s", got)
+	}
+}
+
+// The image is never guessed from what is running: a derived tag is wrong
+// for a digest-pinned image, a mirror or a fork.
+func TestRunImageRefusesWithoutAnImage(t *testing.T) {
+	store, rs := newRun(t)
+	_, err := RunImage(context.Background(), store, rs, ImageOptions{TargetVersion: "0.16.14"})
+	if err == nil {
+		t.Fatal("expected a refusal when no image was named")
+	}
+	if !strings.Contains(err.Error(), "never guessed") {
+		t.Errorf("error should say the image is not inferred, got: %v", err)
+	}
+}
+
+// A resumed run skips the step, and must still report the image the
+// earlier run verified rather than an empty reference.
+func TestRunImageOnResumeReturnsTheVerifiedImage(t *testing.T) {
+	fakeDockerImage(t, "0", testImageID, "echo 'stalwart 0.16.14'")
+	store, rs := newRun(t)
+	opts := ImageOptions{Image: "img:1", TargetVersion: "0.16.14"}
+
+	if _, err := RunImage(context.Background(), store, rs, opts); err != nil {
+		t.Fatal(err)
+	}
+	resumed, err := store.Load(rs.RunID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	staged, err := RunImage(context.Background(), store, resumed, opts)
+	if err != nil {
+		t.Fatalf("resume: %v", err)
+	}
+	if staged.Ref != testImageID {
+		t.Errorf("resumed Ref = %q, want %q", staged.Ref, testImageID)
+	}
+}
+
+func readFile(t *testing.T, path string) string {
+	t.Helper()
+	b, err := os.ReadFile(path)
+	if os.IsNotExist(err) {
+		return ""
+	}
+	if err != nil {
+		t.Fatal(err)
+	}
+	return string(b)
+}