Files
orca/internal/cli/upgrade.go
T
Jon Chery 9a28dc907b feat(P14a): v0.8→v1.0 data migration (REQ-066, C-07) + orca upgrade (REQ-115, C-25, C-27)
internal/migration/migrate.go: Migratev08tov11 (flat→multi-ns, schema
migration, CA import to step-ca, config.hcl preserve). internal/cli/
upgrade.go: orca upgrade --to (thin wrapper, R-017 binding cutover
with C-25 post-verify+rollback, C-27 orca user creation, --import-ca,
--dry-run). Tests: detect/migrate/dry-run/idempotent, cutover verify/
rollback, user creation.

---ci---
project: orca
phase: 14a
milestone: v0.11
status: execute
---/ci---
2026-08-07 08:00:10 +00:00

399 lines
13 KiB
Go

// Package cli: upgrade.go implements the `orca upgrade --to <version>`
// subcommand (REQ-115, gates C-25 and C-27). It is a thin wrapper
// around scripts/install.sh + orca restore that also handles the
// R-017 binding cutover (Traefik :443 → 127.0.0.1:8443 + nftables) for
// existing v0.9/v0.10 clusters, creates the `orca` system user on
// existing peers (C-27, needed before P10b drift detection works),
// and optionally imports the v0.8 internal CA into step-ca.
package cli
import (
"context"
"fmt"
"log/slog"
"net/http"
"os"
"os/exec"
"path/filepath"
"strings"
"time"
"github.com/spf13/cobra"
"git.cloudinit.dev/coreci/orca/internal/migration"
"git.cloudinit.dev/coreci/orca/internal/paths"
)
var (
upgradeTo string
upgradeImportCA bool
upgradeForce bool
upgradeDryRun bool
)
// commandRunner is the seam for running shell commands (install.sh,
// nft, useradd). Tests inject a mock; production uses execRunner.
type commandRunner interface {
Run(ctx context.Context, name string, args ...string) ([]byte, error)
}
// execRunner runs commands via os/exec.
type execRunner struct{}
func (execRunner) Run(ctx context.Context, name string, args ...string) ([]byte, error) {
cmd := exec.CommandContext(ctx, name, args...)
return cmd.CombinedOutput()
}
// upgradeRunnerOverride is the package-level test seam for the
// command runner.
var upgradeRunnerOverride commandRunner
// httpClientOverride is the package-level test seam for the cutover
// verification HTTP check (C-25). Tests inject a mock.
var httpClientOverride func(url string) (int, error)
// upgradeTransport is the SSH surface the upgrade command needs for
// C-27 orca user creation on peers. Mirrors peerSetupTransport.
type upgradeTransport interface {
Exec(ctx context.Context, peer string, cmd string) ([]byte, error)
}
// upgradeTransportOverride is the package-level test seam for the
// SSH transport.
var upgradeTransportOverride upgradeTransport
// peersListerOverride is the package-level test seam for listing
// peers to create the orca user on. Returns a list of peer addresses.
var peersListerOverride func() ([]string, error)
var upgradeCmd = &cobra.Command{
Use: "upgrade",
Short: "Upgrade orca to a new version (REQ-115, R-017 cutover)",
Long: `orca upgrade --to <version> is a thin wrapper around
install.sh + orca restore. It handles:
- R-017 binding cutover: Traefik :443 → 127.0.0.1:8443 + nftables
for existing v0.9/v0.10 clusters (with C-25 post-cutover
verification and rollback on failure)
- C-27: creates the 'orca' system user on existing peers
(useradd -r orca, idempotent) — needed before P10b drift detection
- Optional --import-ca: imports the v0.8 internal CA into step-ca
- If a v0.8 layout is detected, runs Migratev08tov11 first
- Idempotent: if already at the target version, no-op`,
Args: cobra.NoArgs,
RunE: func(cmd *cobra.Command, args []string) error {
return runUpgrade(cmd, cmd.OutOrStdout())
},
}
func init() {
upgradeCmd.Flags().StringVar(&upgradeTo, "to", "", "target version (e.g. v0.11.0) (required)")
upgradeCmd.Flags().BoolVar(&upgradeImportCA, "import-ca", false, "import the v0.8 internal CA into step-ca during upgrade")
upgradeCmd.Flags().BoolVar(&upgradeForce, "force", false, "skip cutover verification (use with caution)")
upgradeCmd.Flags().BoolVar(&upgradeDryRun, "dry-run", false, "report what would be done without making changes")
rootCmd.AddCommand(upgradeCmd)
}
// UpgradeResult is the JSON-serializable summary of an upgrade run.
type UpgradeResult struct {
TargetVersion string `json:"target_version"`
CurrentVersion string `json:"current_version"`
DryRun bool `json:"dry_run"`
MigratedV08 bool `json:"migrated_v08"`
CutoverNeeded bool `json:"cutover_needed"`
CutoverOK bool `json:"cutover_ok,omitempty"`
CutoverRolled bool `json:"cutover_rolled_back,omitempty"`
UsersCreated []string `json:"users_created,omitempty"`
CAImported bool `json:"ca_imported,omitempty"`
BinaryUpdated bool `json:"binary_updated"`
}
func runUpgrade(cmd *cobra.Command, out interface{ Write([]byte) (int, error) }) error {
if upgradeTo == "" {
return fmt.Errorf("--to is required (e.g. --to v0.11.0)")
}
res := UpgradeResult{
TargetVersion: upgradeTo,
CurrentVersion: version,
DryRun: upgradeDryRun,
}
if !jsonOutput {
fmt.Fprintf(out, "orca upgrade --to %s (current: %s)\n", upgradeTo, version)
}
if version == strings.TrimPrefix(upgradeTo, "v") && !upgradeDryRun {
if !jsonOutput {
fmt.Fprintf(out, "✓ Already at target version %s; no-op\n", upgradeTo)
}
res.BinaryUpdated = false
if jsonOutput {
return printJSON(res)
}
return nil
}
home := paths.Root()
if migration.Detectv08(home) {
if !jsonOutput {
fmt.Fprintf(out, "• v0.8 layout detected; running data migration first\n")
}
if upgradeDryRun {
fmt.Fprintf(out, " [dry-run] would run Migratev08tov11(source=%s)\n", home)
} else {
if err := migration.Migratev08tov11(migration.MigrateOptions{
SourceDir: home,
TargetDir: home,
ImportCA: upgradeImportCA,
DryRun: false,
}); err != nil {
return fmt.Errorf("v0.8 migration: %w", err)
}
}
res.MigratedV08 = true
}
runner := upgradeRunnerOverride
if runner == nil {
runner = execRunner{}
}
cutoverNeeded := detectOldTraefikBinding()
res.CutoverNeeded = cutoverNeeded
if cutoverNeeded {
if !jsonOutput {
fmt.Fprintf(out, "• R-017 cutover: Traefik on :443 detected; cutting over to 127.0.0.1:8443 + nftables\n")
}
if upgradeDryRun {
fmt.Fprintf(out, " [dry-run] would reconfigure Traefik to 127.0.0.1:8443 + add nftables redirect\n")
} else {
ok, err := performCutover(cmd.Context(), runner, out, upgradeForce)
if err != nil {
return err
}
res.CutoverOK = ok
if !ok {
res.CutoverRolled = true
if jsonOutput {
return printJSON(res)
}
return fmt.Errorf("cutover verification failed; rolled back to :443")
}
}
}
peers, err := listPeers()
if err != nil {
slog.Warn("upgrade: list peers failed", "err", err)
}
if len(peers) > 0 {
if !jsonOutput {
fmt.Fprintf(out, "• C-27: creating orca system user on %d peer(s)\n", len(peers))
}
created, err := createOrcaUserOnPeers(cmd.Context(), peers)
if err != nil {
slog.Warn("upgrade: orca user creation on peers failed", "err", err)
}
res.UsersCreated = created
}
if upgradeImportCA && !res.MigratedV08 {
if !jsonOutput {
fmt.Fprintf(out, "• Importing v0.8 internal CA into step-ca\n")
}
if upgradeDryRun {
fmt.Fprintf(out, " [dry-run] would import ca.crt/ca.key into step-ca\n")
} else {
if err := importCA(cmd.Context()); err != nil {
slog.Warn("upgrade: CA import failed", "err", err)
} else {
res.CAImported = true
}
}
}
if upgradeDryRun {
fmt.Fprintf(out, " [dry-run] would download and install orca %s via install.sh\n", upgradeTo)
} else {
if !jsonOutput {
fmt.Fprintf(out, "• Downloading and installing orca %s via install.sh\n", upgradeTo)
}
installSh := findInstallScript()
if _, err := runner.Run(cmd.Context(), "bash", installSh, "--version", upgradeTo); err != nil {
return fmt.Errorf("install.sh: %w", err)
}
res.BinaryUpdated = true
}
if jsonOutput {
return printJSON(res)
}
fmt.Fprintf(out, "\n✓ upgrade to %s complete\n", upgradeTo)
if cutoverNeeded && !upgradeDryRun {
fmt.Fprintf(out, "\nRollback procedure (if cutover failed):\n")
fmt.Fprintf(out, " 1. Restore Traefik entrypoint to :443\n")
fmt.Fprintf(out, " 2. Remove nftables rules: nft delete table inet orca_redirect\n")
fmt.Fprintf(out, " 3. Restart Traefik: systemctl restart traefik\n")
fmt.Fprintf(out, " 4. Verify: curl -k https://localhost:443/\n")
}
return nil
}
// detectOldTraefikBinding returns true if Traefik is listening on :443
// (the v0.9/v0.10 default that R-017 cuts over from). Best-effort: if
// the check fails, returns false (no cutover needed).
func detectOldTraefikBinding() bool {
runner := upgradeRunnerOverride
if runner == nil {
runner = execRunner{}
}
ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
defer cancel()
out, err := runner.Run(ctx, "ss", "-tlnp")
if err != nil {
return false
}
return strings.Contains(string(out), ":443")
}
// performCutover reconfigures Traefik to listen on 127.0.0.1:8443 and
// adds nftables DNAT redirect from :443 → 127.0.0.1:8443. Then runs
// C-25 post-cutover verification: curl -k https://localhost:443/ must
// return 200. On failure, rolls back (restores :443, removes nft rules)
// and returns (false, nil). On success returns (true, nil). With
// force=true, verification is skipped.
func performCutover(ctx context.Context, runner commandRunner, out interface{ Write([]byte) (int, error) }, force bool) (bool, error) {
if _, err := runner.Run(ctx, "sed", "-i", "s/:443/127.0.0.1:8443/g", "/etc/traefik/traefik.yml"); err != nil {
return false, fmt.Errorf("cutover: edit traefik.yml: %w", err)
}
if _, err := runner.Run(ctx, "systemctl", "restart", "traefik"); err != nil {
return false, fmt.Errorf("cutover: restart traefik: %w", err)
}
nftCmd := `nft add table inet orca_redirect; nft 'add chain inet orca_redirect prerouting { type nat hook prerouting priority -100; }'; nft add rule inet orca_redirect prerouting tcp dport 443 dnat to 127.0.0.1:8443`
if _, err := runner.Run(ctx, "bash", "-c", nftCmd); err != nil {
slog.Warn("cutover: nftables rule add failed (non-fatal if already present)", "err", err)
}
if force {
fmt.Fprintf(out, " --force: skipping cutover verification\n")
return true, nil
}
if err := verifyCutover(out); err != nil {
fmt.Fprintf(out, " ✗ C-25 cutover verification failed: %v\n", err)
fmt.Fprintf(out, " Rolling back to :443...\n")
if rbErr := rollbackCutover(ctx, runner); rbErr != nil {
slog.Error("cutover: rollback failed", "err", rbErr)
}
return false, nil
}
fmt.Fprintf(out, " ✓ C-25 cutover verification passed (200 from Traefik)\n")
return true, nil
}
// verifyCutover runs the C-25 post-cutover check: curl -k
// https://localhost:443/ must return HTTP 200.
func verifyCutover(out interface{ Write([]byte) (int, error) }) error {
if httpClientOverride != nil {
code, err := httpClientOverride("https://localhost:443/")
if err != nil {
return err
}
if code != 200 {
return fmt.Errorf("HTTP %d (want 200)", code)
}
return nil
}
client := &http.Client{Timeout: 10 * time.Second}
resp, err := client.Get("https://localhost:443/")
if err != nil {
return fmt.Errorf("curl: %w", err)
}
defer resp.Body.Close()
if resp.StatusCode != 200 {
return fmt.Errorf("HTTP %d (want 200)", resp.StatusCode)
}
return nil
}
// rollbackCutover restores Traefik to :443 and removes nftables rules.
func rollbackCutover(ctx context.Context, runner commandRunner) error {
if _, err := runner.Run(ctx, "sed", "-i", "s/127.0.0.1:8443/:443/g", "/etc/traefik/traefik.yml"); err != nil {
return fmt.Errorf("rollback: edit traefik.yml: %w", err)
}
if _, err := runner.Run(ctx, "systemctl", "restart", "traefik"); err != nil {
return fmt.Errorf("rollback: restart traefik: %w", err)
}
if _, err := runner.Run(ctx, "nft", "delete", "table", "inet", "orca_redirect"); err != nil {
slog.Warn("rollback: nft delete table failed (non-fatal if not present)", "err", err)
}
return nil
}
// createOrcaUserOnPeers runs the idempotent `useradd -r orca` on each
// peer via SSH (C-27). Returns the list of peers where the user was
// created (or already existed).
func createOrcaUserOnPeers(ctx context.Context, peers []string) ([]string, error) {
transport := upgradeTransportOverride
if transport == nil {
return nil, fmt.Errorf("no SSH transport available for peer user creation")
}
var created []string
for _, peer := range peers {
cmd := "useradd -r orca -s /usr/sbin/nologin 2>/dev/null || true"
if _, err := transport.Exec(ctx, peer, cmd); err != nil {
slog.Warn("upgrade: useradd on peer failed", "peer", peer, "err", err)
continue
}
created = append(created, peer)
}
return created, nil
}
// listPeers returns the list of peer addresses for C-27 orca user
// creation. Uses the peersListerOverride test seam if set; otherwise
// returns an empty list (no peers configured).
func listPeers() ([]string, error) {
if peersListerOverride != nil {
return peersListerOverride()
}
return nil, nil
}
// importCA imports the v0.8 internal CA into step-ca. This is the
// C-07 gate. Production wires a real step-ca client; tests use the
// migration.caImporterOverride seam.
func importCA(ctx context.Context) error {
caCrt := filepath.Join(paths.Root(), "ca.crt")
caKey := filepath.Join(paths.Root(), "ca.key")
if _, err := os.Stat(caCrt); err != nil {
return fmt.Errorf("import CA: ca.crt not found: %w", err)
}
if _, err := os.Stat(caKey); err != nil {
return fmt.Errorf("import CA: ca.key not found: %w", err)
}
importer := migration.GetCAImporter()
if importer == nil {
return fmt.Errorf("import CA: no CA importer available")
}
return importer.ImportCA(ctx, caCrt, caKey)
}
// findInstallScript returns the path to scripts/install.sh. Checks
// the repo-local path first; falls back to downloading via curl
// (handled by the caller).
func findInstallScript() string {
candidates := []string{
"scripts/install.sh",
"/usr/local/share/orca/scripts/install.sh",
}
for _, c := range candidates {
if _, err := os.Stat(c); err == nil {
return c
}
}
return "scripts/install.sh"
}