Track active/passive boot for upgrade workflow (#133)

* Track active/passive boot for upgrade workflow

Track what are we booting from during upgrade in order to not backup
active into passive if we are booting from passive.

If we are booting from active or recovery the workflow keeps as it was,
backup active into passive, and then move the transition image into
active.

Signed-off-by: Itxaka <itxaka@kairos.io>

go.mod

@@ -12,7 +12,7 @@ require (
 	github.com/hashicorp/go-multierror v1.1.1
 	github.com/jaypipes/ghw v0.12.0
 	github.com/joho/godotenv v1.5.1
-	github.com/kairos-io/kairos-sdk v0.0.13
+	github.com/kairos-io/kairos-sdk v0.0.14
 	github.com/labstack/echo/v4 v4.11.1
 	github.com/mitchellh/mapstructure v1.5.0
 	github.com/mudler/go-nodepair v0.0.0-20221223092639-ba399a66fdfb

go.sum

@@ -358,6 +358,8 @@ github.com/jtolds/gls v4.20.0+incompatible/go.mod h1:QJZ7F/aHp+rZTRtaJ1ow/lLfFfV
 github.com/julienschmidt/httprouter v1.2.0/go.mod h1:SYymIcj16QtmaHHD7aYtjjsJG7VTCxuUUipMqKk8s4w=
 github.com/kairos-io/kairos-sdk v0.0.13 h1:YI+ETmlk86+mHTQYBNPmVTiO2Ux63HmkRzxhAJRr9JI=
 github.com/kairos-io/kairos-sdk v0.0.13/go.mod h1:+fhVoCEA2hD7OFTmhDmy4pP3uC9q84OYhDa7irkEVWQ=
+github.com/kairos-io/kairos-sdk v0.0.14 h1:yVKaay70cjmAUY6m4uLoV9UguBv3Dx3JxAh2faRlUZo=
+github.com/kairos-io/kairos-sdk v0.0.14/go.mod h1:+fhVoCEA2hD7OFTmhDmy4pP3uC9q84OYhDa7irkEVWQ=
 github.com/kbinani/screenshot v0.0.0-20210720154843-7d3a670d8329 h1:qq2nCpSrXrmvDGRxW0ruW9BVEV1CN2a9YDOExdt+U0o=
 github.com/kbinani/screenshot v0.0.0-20210720154843-7d3a670d8329/go.mod h1:2VPVQDR4wO7KXHwP+DAypEy67rXf+okUx2zjgpCxZw4=
 github.com/kendru/darwin/go/depgraph v0.0.0-20221105232959-877d6a81060c h1:eKb4PqwAMhlqwXw0W3atpKaYaPGlXE/Fwh+xpCEYaPk=

pkg/action/upgrade.go

@@ -18,15 +18,16 @@ package action
 
 import (
 	"fmt"
-	agentConfig "github.com/kairos-io/kairos-agent/v2/pkg/config"
-	"github.com/kairos-io/kairos-agent/v2/pkg/utils/fs"
 	"path/filepath"
 	"time"
 
+	agentConfig "github.com/kairos-io/kairos-agent/v2/pkg/config"
 	"github.com/kairos-io/kairos-agent/v2/pkg/constants"
 	"github.com/kairos-io/kairos-agent/v2/pkg/elemental"
 	v1 "github.com/kairos-io/kairos-agent/v2/pkg/types/v1"
 	"github.com/kairos-io/kairos-agent/v2/pkg/utils"
+	"github.com/kairos-io/kairos-agent/v2/pkg/utils/fs"
+	"github.com/kairos-io/kairos-sdk/state"
 )
 
 // UpgradeAction represents the struct that will run the upgrade from start to finish
@@ -121,6 +122,16 @@ func (u *UpgradeAction) Run() (err error) {
 	var upgradeImg v1.Image
 	var finalImageFile string
 
+	// Track where we booted from to have a different workflow
+	// Booted from active: backup active into passive, upgrade active
+	// Booted from passive: Upgrade active, leave passive as is
+	var bootedFrom state.Boot
+
+	bootedFrom, err = state.DetectBootWithVFS(u.config.Fs)
+	if err != nil {
+		u.config.Logger.Warnf("error detecting boot: %s", err)
+	}
+
 	cleanup := utils.NewCleanStack()
 	defer func() { err = cleanup.Cleanup(err) }()
 
@@ -230,9 +241,10 @@ func (u *UpgradeAction) Run() (err error) {
 		return err
 	}
 
-	// If not upgrading recovery, backup active into passive
+	// If not upgrading recovery and booting from non passive, backup active into passive
-	if !u.spec.RecoveryUpgrade {
+	// We dont want to overwrite passive if we are booting from passive as it could mean that active is broken and we would
-		//TODO this step could be part of elemental package
+	// be overriding a working passive with a broken/unknown  active
+	if u.spec.RecoveryUpgrade == false && bootedFrom != state.Passive {
 		// backup current active.img to passive.img before overwriting the active.img
 		u.Info("Backing up current active image")
 		source := filepath.Join(u.spec.Partitions.State.MountPoint, "cOS", constants.ActiveImgFile)

pkg/action/upgrade_test.go

@@ -149,6 +149,13 @@ var _ = Describe("Runtime Actions", func() {
 				err = fsutils.MkdirAll(config.Fs, filepath.Join(spec.Active.MountPoint, "etc"), constants.DirPerm)
 				Expect(err).ShouldNot(HaveOccurred())
 
+				err = fsutils.MkdirAll(config.Fs, "/proc", constants.DirPerm)
+				Expect(err).ShouldNot(HaveOccurred())
+
+				// Write proc/cmdline so we can detect what we booted from
+				err = fs.WriteFile("/proc/cmdline", []byte(constants.ActiveLabel), constants.FilePerm)
+				Expect(err).ShouldNot(HaveOccurred())
+
 				err = fs.WriteFile(
 					filepath.Join(spec.Active.MountPoint, "etc", "os-release"),
 					[]byte("GRUB_ENTRY_NAME=TESTOS"),
@@ -161,9 +168,6 @@ var _ = Describe("Runtime Actions", func() {
 				spec.Recovery.Size = 10
 
 				runner.SideEffect = func(command string, args ...string) ([]byte, error) {
-					if command == "cat" && args[0] == "/proc/cmdline" {
-						return []byte(constants.ActiveLabel), nil
-					}
 					if command == "mv" && args[0] == "-f" && args[1] == activeImg && args[2] == passiveImg {
 						// we doing backup, do the "move"
 						source, _ := fs.ReadFile(activeImg)
@@ -359,6 +363,13 @@ var _ = Describe("Runtime Actions", func() {
 				)
 				Expect(err).ShouldNot(HaveOccurred())
 
+				err = fsutils.MkdirAll(config.Fs, "/proc", constants.DirPerm)
+				Expect(err).ShouldNot(HaveOccurred())
+
+				// Write proc/cmdline so we can detect what we booted from
+				err = fs.WriteFile("/proc/cmdline", []byte(constants.PassiveLabel), constants.FilePerm)
+				Expect(err).ShouldNot(HaveOccurred())
+
 				spec.Active.Size = 10
 				spec.Passive.Size = 10
 				spec.Recovery.Size = 10
@@ -373,6 +384,10 @@ var _ = Describe("Runtime Actions", func() {
 						_ = fs.WriteFile(activeImg, source, constants.FilePerm)
 						_ = fs.RemoveAll(spec.Active.File)
 					}
+					if command == "mv" && args[0] == "-f" && args[1] == activeImg && args[2] == passiveImg {
+						// If this command was called then its a complete failure as it tried to copy active into passive
+						StopTrying("Passive was overwritten").Now()
+					}
 					return []byte{}, nil
 				}
 				config.Runner = runner
@@ -436,6 +451,13 @@ var _ = Describe("Runtime Actions", func() {
 
 					spec.RecoveryUpgrade = true
 
+					err = fsutils.MkdirAll(config.Fs, "/proc", constants.DirPerm)
+					Expect(err).ShouldNot(HaveOccurred())
+
+					// Write proc/cmdline so we can detect what we booted from
+					err = fs.WriteFile("/proc/cmdline", []byte(constants.RecoveryLabel), constants.FilePerm)
+					Expect(err).ShouldNot(HaveOccurred())
+
 					runner.SideEffect = func(command string, args ...string) ([]byte, error) {
 						if command == "cat" && args[0] == "/proc/cmdline" {
 							return []byte(constants.RecoveryLabel), nil
@@ -522,6 +544,13 @@ var _ = Describe("Runtime Actions", func() {
 
 					spec.RecoveryUpgrade = true
 
+					err = fsutils.MkdirAll(config.Fs, "/proc", constants.DirPerm)
+					Expect(err).ShouldNot(HaveOccurred())
+
+					// Write proc/cmdline so we can detect what we booted from
+					err = fs.WriteFile("/proc/cmdline", []byte(constants.RecoveryLabel), constants.FilePerm)
+					Expect(err).ShouldNot(HaveOccurred())
+
 					runner.SideEffect = func(command string, args ...string) ([]byte, error) {
 						if command == "cat" && args[0] == "/proc/cmdline" {
 							return []byte(constants.RecoveryLabel), nil