feat(machine): add validation for machine names to accept only DNS labels

* feat(dns): implement extra dns names for machine in m.internal

This implements <machinename>.m.internal and <machineid>.m.internal as
names that can be queried in the cluster.

Listing all machines/services is not implemented, this may come later.

Fixes: #340

Signed-off-by: Miek Gieben <miek@miek.nl>

* Implement getting all machine address

A query for m.internal returns all ip address

There is no overflow check

Signed-off-by: Miek Gieben <miek@miek.nl>

---------

Signed-off-by: Miek Gieben <miek@miek.nl>

cmd/uc/deploy.go

@@ -305,7 +305,7 @@ const defaultFailedContainerLogsTail = 10
 // UNCLOUD_FAILED_CONTAINER_LOGS_TAIL environment variable override when set and valid.
 func failedContainerLogsTail() int {
 	if v := os.Getenv("UNCLOUD_FAILED_CONTAINER_LOGS_TAIL"); v != "" {
-		if tail, err := logs.Tail(v); err == nil && (tail == -1 || tail > 0) {
+		if tail, err := logs.ParseTail(v); err == nil && (tail == -1 || tail > 0) {
 			return tail
 		}
 	}

cmd/uc/machine/logs.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"slices"
 	"strings"
+	"time"
 
 	"github.com/psviderski/uncloud/internal/cli"
 	"github.com/psviderski/uncloud/internal/cli/completion"
@@ -64,6 +65,15 @@ If no services are specified, streams logs from the uncloud service.`,
 }
 
 func runLogs(ctx context.Context, uncli *cli.CLI, services []string, opts logs.Options) error {
+	since, until, err := logs.TimeRange(opts.Since, opts.Until, time.Now())
+	if err != nil {
+		return err
+	}
+	tail, err := opts.TailLines()
+	if err != nil {
+		return err
+	}
+
 	if len(services) == 0 {
 		services = []string{api.SystemServiceUncloud}
 	}
@@ -74,11 +84,6 @@ func runLogs(ctx context.Context, uncli *cli.CLI, services []string, opts logs.O
 		}
 	}
 
-	tail, err := logs.Tail(opts.Tail)
-	if err != nil {
-		return err
-	}
-
 	c, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
@@ -88,8 +93,8 @@ func runLogs(ctx context.Context, uncli *cli.CLI, services []string, opts logs.O
 	logsOpts := api.ServiceLogsOptions{
 		Follow:   opts.Follow,
 		Tail:     tail,
-		Since:    opts.Since,
-		Until:    opts.Until,
+		Since:    since,
+		Until:    until,
 		Machines: cli.ExpandCommaSeparatedValues(opts.Machines),
 	}
 

cmd/uc/machine/logs_test.go

@@ -0,0 +1,25 @@
+package machine
+
+import (
+	"context"
+	"testing"
+
+	"github.com/psviderski/uncloud/internal/cli/logs"
+	"github.com/stretchr/testify/require"
+)
+
+func TestRunLogsInvalidTimeFilters(t *testing.T) {
+	t.Parallel()
+
+	// Invalid filters must fail before connecting to the cluster.
+	for _, flag := range []string{"since", "until"} {
+		opts := logs.Options{}
+		if flag == "since" {
+			opts.Since = "invalid"
+		} else {
+			opts.Until = "invalid"
+		}
+		err := runLogs(context.Background(), nil, nil, opts)
+		require.ErrorContains(t, err, "invalid --"+flag+" value")
+	}
+}

cmd/uc/service/logs.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"fmt"
 	"strings"
+	"time"
 
 	mapset "github.com/deckarep/golang-set/v2"
 	"github.com/psviderski/uncloud/internal/cli"
@@ -78,6 +79,15 @@ If no services are specified, streams logs from all services defined in the Comp
 }
 
 func RunLogs(ctx context.Context, uncli *cli.CLI, args []string, opts logs.Options) error {
+	since, until, err := logs.TimeRange(opts.Since, opts.Until, time.Now())
+	if err != nil {
+		return err
+	}
+	tail, err := opts.TailLines()
+	if err != nil {
+		return err
+	}
+
 	serviceArgs, err := logs.ParseServiceArgs(args)
 	if err != nil {
 		return err
@@ -106,12 +116,6 @@ func RunLogs(ctx context.Context, uncli *cli.CLI, args []string, opts logs.Optio
 		}
 	}
 
-	// Parse tail option.
-	tail, err := logs.Tail(opts.Tail)
-	if err != nil {
-		return err
-	}
-
 	c, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
@@ -121,8 +125,8 @@ func RunLogs(ctx context.Context, uncli *cli.CLI, args []string, opts logs.Optio
 	baseOpts := api.ServiceLogsOptions{
 		Follow:   opts.Follow,
 		Tail:     tail,
-		Since:    opts.Since,
-		Until:    opts.Until,
+		Since:    since,
+		Until:    until,
 		Machines: cli.ExpandCommaSeparatedValues(opts.Machines),
 	}
 

cmd/uc/service/logs_test.go

@@ -0,0 +1,25 @@
+package service
+
+import (
+	"context"
+	"testing"
+
+	"github.com/psviderski/uncloud/internal/cli/logs"
+	"github.com/stretchr/testify/require"
+)
+
+func TestRunLogsInvalidTimeFilters(t *testing.T) {
+	t.Parallel()
+
+	// Invalid filters must fail before loading Compose files or connecting to the cluster.
+	for _, flag := range []string{"since", "until"} {
+		opts := logs.Options{}
+		if flag == "since" {
+			opts.Since = "invalid"
+		} else {
+			opts.Until = "invalid"
+		}
+		err := RunLogs(context.Background(), nil, nil, opts)
+		require.ErrorContains(t, err, "invalid --"+flag+" value")
+	}
+}

internal/cli/logs/duration.go

@@ -0,0 +1,66 @@
+package logs
+
+import (
+	"fmt"
+	"math/big"
+	"strings"
+	"time"
+)
+
+// Duration is an extended standard time.Duration that also supports days as a unit.
+type Duration = time.Duration
+
+// ParseDuration parses a Go duration with the additional unit d, meaning exactly 24 hours.
+// Days can be fractional or combined with other units, such as "1.5d" or "2d3h".
+func ParseDuration(value string) (Duration, error) {
+	if !strings.Contains(value, "d") {
+		return time.ParseDuration(value)
+	}
+
+	rest := value
+	var normalised strings.Builder
+	if len(rest) > 0 && (rest[0] == '-' || rest[0] == '+') {
+		normalised.WriteByte(rest[0])
+		rest = rest[1:]
+	}
+	for len(rest) > 0 {
+		// Each component is a decimal number followed by a unit. Only the leading sign is allowed.
+		numberEnd := 0
+		for numberEnd < len(rest) && (isDigit(rest[numberEnd]) || rest[numberEnd] == '.') {
+			numberEnd++
+		}
+		unitEnd := numberEnd
+		for unitEnd < len(rest) && !isDigit(rest[unitEnd]) && rest[unitEnd] != '.' {
+			unitEnd++
+		}
+		if numberEnd == 0 || unitEnd == numberEnd {
+			return 0, fmt.Errorf("time: invalid duration '%s'", value)
+		}
+		number, unit := rest[:numberEnd], rest[numberEnd:unitEnd]
+		if unit == "d" {
+			// Use exact decimal arithmetic to avoid float rounding and retain nanosecond precision.
+			days, ok := new(big.Rat).SetString(number)
+			if !ok {
+				return 0, fmt.Errorf("time: invalid duration '%s'", value)
+			}
+			days.Mul(days, new(big.Rat).SetInt64(int64(24*time.Hour)))
+			nanoseconds := new(big.Int).Quo(days.Num(), days.Denom())
+			normalised.WriteString(nanoseconds.String())
+			normalised.WriteString("ns")
+		} else {
+			normalised.WriteString(rest[:unitEnd])
+		}
+		rest = rest[unitEnd:]
+	}
+
+	// The standard parser validates the remaining units and checks the total for overflow.
+	duration, err := time.ParseDuration(normalised.String())
+	if err != nil {
+		return 0, fmt.Errorf("time: invalid duration '%s': %w", value, err)
+	}
+	return duration, nil
+}
+
+func isDigit(c byte) bool {
+	return c >= '0' && c <= '9'
+}

internal/cli/logs/duration_test.go

@@ -0,0 +1,62 @@
+package logs
+
+import (
+	"math"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestParseDuration(t *testing.T) {
+	t.Parallel()
+
+	for _, tt := range []struct {
+		input string
+		want  time.Duration
+	}{
+		{"0", 0},
+		{"2m30s", 150 * time.Second},
+		{"1µs", time.Microsecond},
+		{"1μs", time.Microsecond},
+		{"2d", 48 * time.Hour},
+		{"010d", 240 * time.Hour},
+		{"08d", 192 * time.Hour},
+		{"2d3h", 51 * time.Hour},
+		{"3h2d", 51 * time.Hour},
+		{"1d1d", 48 * time.Hour},
+		{"1.5d", 36 * time.Hour},
+		{".5d", 12 * time.Hour},
+		{"1.d", 24 * time.Hour},
+		{"-2d3h", -51 * time.Hour},
+		{"+2d", 48 * time.Hour},
+		{"0d", 0},
+		{"0d1ns", time.Nanosecond},
+		{"0.000000000001d", 86 * time.Nanosecond},
+		{"-0.000000000001d", -86 * time.Nanosecond},
+		{"106751d23h47m16.854775807s", time.Duration(math.MaxInt64)},
+		{"-106751d23h47m16.854775808s", time.Duration(math.MinInt64)},
+	} {
+		t.Run(tt.input, func(t *testing.T) {
+			actual, err := ParseDuration(tt.input)
+			require.NoError(t, err)
+			assert.Equal(t, tt.want, actual)
+		})
+	}
+}
+
+func TestParseDuration_Invalid(t *testing.T) {
+	t.Parallel()
+
+	for _, input := range []string{
+		"", "d", "2d3", "1..2d", ".d", "1d-2h", "1d+2h", "1d 2h", "1day", "2D", "1w",
+		"1e2d", "1d2w", "106752d", "-106752d", "106751d23h47m16.854775808s",
+		"-106751d23h47m16.854775809s", "999999999999999999999999999d",
+	} {
+		t.Run(input, func(t *testing.T) {
+			_, err := ParseDuration(input)
+			require.Error(t, err)
+		})
+	}
+}

internal/cli/logs/formatter.go

@@ -10,6 +10,7 @@ import (
 	"time"
 
 	"charm.land/lipgloss/v2"
+	"github.com/charmbracelet/colorprofile"
 	"github.com/docker/docker/pkg/stringid"
 	"github.com/psviderski/uncloud/internal/cli/tui"
 	"github.com/psviderski/uncloud/pkg/api"
@@ -24,6 +25,10 @@ type Formatter struct {
 	maxServiceWidth int
 
 	utc bool
+
+	// Cache each stream's terminal profile to avoid detection on every log entry.
+	stdout *colorprofile.Writer
+	stderr *colorprofile.Writer
 }
 
 func NewFormatter(machineNames, serviceNames []string, utc bool) *Formatter {
@@ -50,6 +55,8 @@ func NewFormatter(machineNames, serviceNames []string, utc bool) *Formatter {
 		maxMachineWidth: maxMachineWidth,
 		maxServiceWidth: maxServiceWidth,
 		utc:             utc,
+		stdout:          colorprofile.NewWriter(os.Stdout, os.Environ()),
+		stderr:          colorprofile.NewWriter(os.Stderr, os.Environ()),
 	}
 }
 
@@ -138,9 +145,9 @@ func (f *Formatter) PrintEntry(entry api.ServiceLogEntry) {
 
 	// Print to appropriate stream.
 	if entry.Stream == api.LogStreamStderr {
-		fmt.Fprint(os.Stderr, output.String())
+		fmt.Fprint(f.stderr, output.String())
 	} else {
-		fmt.Print(output.String())
+		fmt.Fprint(f.stdout, output.String())
 	}
 }
 
@@ -148,8 +155,7 @@ func (f *Formatter) PrintEntry(entry api.ServiceLogEntry) {
 func (f *Formatter) printError(entry api.ServiceLogEntry) {
 	if entry.Metadata.ServiceName == "" {
 		msg := fmt.Sprintf("ERROR: %v", entry.Err)
-		style := lipgloss.NewStyle().Bold(true).Foreground(lipgloss.BrightRed)
-		fmt.Fprintln(os.Stderr, style.Render(msg))
+		fmt.Fprintln(f.stderr, tui.BoldRed.Render(msg))
 		return
 	}
 
@@ -171,8 +177,7 @@ func (f *Formatter) printError(entry api.ServiceLogEntry) {
 		msg += fmt.Sprintf(": %v", entry.Err)
 	}
 
-	style := lipgloss.NewStyle().Bold(true).Foreground(lipgloss.BrightYellow)
-	fmt.Fprintln(os.Stderr, style.Render(msg))
+	fmt.Fprintln(f.stderr, tui.BoldYellow.Render(msg))
 }
 
 // palette is available colors for machine/service differentiation.

internal/cli/logs/logs.go

@@ -19,6 +19,18 @@ type Options struct {
 	Machines []string
 }
 
+// TailLines resolves the default tail limit after parsing flags. An empty opts.Tail means the user
+// did not specify a limit, so --since can select all matching logs without overriding an explicit --tail.
+func (opts Options) TailLines() (int, error) {
+	if opts.Tail == "" {
+		if opts.Since != "" {
+			return -1, nil
+		}
+		return 100, nil
+	}
+	return ParseTail(opts.Tail)
+}
+
 func Flags(options *Options) *pflag.FlagSet {
 	set := &pflag.FlagSet{}
 
@@ -29,14 +41,17 @@ func Flags(options *Options) *pflag.FlagSet {
 	set.StringVar(&options.Since, "since", "",
 		"Show logs generated on or after the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.\n"+
 			"Examples:\n"+
-			"  --since 2m30s                      Relative duration (2 minutes 30 seconds ago)\n"+
-			"  --since 1h                         Relative duration (1 hour ago)\n"+
-			"  --since 2025-11-24                 RFC 3339 date only (midnight using local timezone)\n"+
-			"  --since 2024-05-14T22:50:00        RFC 3339 date/time using local timezone\n"+
+			"  --since 1h45m                      Relative duration (1 hour 45 minutes ago)\n"+
+			"                                     Supported units: d (day = 24h), h (hour), m (minute),\n"+
+			"                                                      s (second), ms (millisecond),\n"+
+			"                                                      us/µs (microsecond), ns (nanosecond)\n"+
+			"  --since 2025-11-24                 RFC 3339 date only (midnight using client local timezone)\n"+
+			"  --since 2024-05-14T22:50:00        RFC 3339 date/time using client local timezone\n"+
 			"  --since 2024-01-31T10:30:00Z       RFC 3339 date/time in UTC\n"+
 			"  --since 1763953966                 Unix timestamp (seconds since January 1, 1970)")
-	set.StringVarP(&options.Tail, "tail", "n", "100",
-		"Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs.")
+	set.StringVarP(&options.Tail, "tail", "n", "",
+		"Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs.\n"+
+			"Defaults to 100, or 'all' when --since is set.")
 	set.StringVar(&options.Until, "until", "",
 		"Show logs generated before the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.\n"+
 			"See --since for examples.")
@@ -46,7 +61,7 @@ func Flags(options *Options) *pflag.FlagSet {
 	return set
 }
 
-func Tail(tail string) (int, error) {
+func ParseTail(tail string) (int, error) {
 	if tail == "all" {
 		return -1, nil
 	}

internal/cli/logs/logs_test.go

@@ -7,6 +7,42 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
+func TestOptionsTailLines(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name string
+		args []string
+		want int
+	}{
+		{"default", nil, 100},
+		{"follow", []string{"-f"}, 100},
+		{"since", []string{"--since", "1h"}, -1},
+		{"since and follow", []string{"--since", "1h", "-f"}, -1},
+		{"until only", []string{"--until", "1h"}, 100},
+		{"time range", []string{"--since", "3h", "--until", "1h"}, -1},
+		{"explicit default with since", []string{"--since", "1h", "--tail", "100"}, 100},
+		{"explicit limit with since", []string{"--since", "1h", "-n", "20"}, 20},
+		{"explicit limit before since", []string{"-n", "20", "--since", "1h"}, 20},
+		{"explicit all", []string{"--tail", "all"}, -1},
+		{"explicit zero with since", []string{"--since", "1h", "-n", "0", "-f"}, 0},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			var options Options
+			require.NoError(t, Flags(&options).Parse(tt.args))
+			tail, err := options.TailLines()
+			require.NoError(t, err)
+			assert.Equal(t, tt.want, tail)
+		})
+	}
+
+	var options Options
+	require.NoError(t, Flags(&options).Parse([]string{"--since", "1h", "--tail", "invalid"}))
+	_, err := options.TailLines()
+	require.ErrorContains(t, err, "invalid --tail value")
+}
+
 func TestParseServiceArgs(t *testing.T) {
 	t.Parallel()
 

internal/cli/logs/time.go

@@ -0,0 +1,72 @@
+package logs
+
+import (
+	"fmt"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// TimeRange resolves log filters that could be RFC 3339, timestamp, or relative duration to UTC timestamps.
+// Dates without a timezone use now's location. Relative durations are computed from now.
+func TimeRange(since, until string, now time.Time) (string, string, error) {
+	var err error
+	since, err = timestamp(since, now)
+	if err != nil {
+		return "", "", fmt.Errorf("invalid --since value: %w", err)
+	}
+	until, err = timestamp(until, now)
+	if err != nil {
+		return "", "", fmt.Errorf("invalid --until value: %w", err)
+	}
+
+	return since, until, nil
+}
+
+func timestamp(value string, now time.Time) (string, error) {
+	if value == "" {
+		return "", nil
+	}
+	// A bare zero is the Unix epoch, matching Docker's log filters.
+	if duration, err := ParseDuration(value); value != "0" && err == nil {
+		return now.Add(-duration).UTC().Format(time.RFC3339Nano), nil
+	}
+
+	// Keep Docker's supported date layouts, but use the location's offset at the requested date.
+	for _, layout := range []string{
+		time.RFC3339Nano,
+		"2006-01-02T15:04Z07:00",
+		"2006-01-02T15Z07:00",
+		"2006-01-02Z07:00",
+		"2006-01-02T15:04:05.999999999",
+		"2006-01-02T15:04",
+		"2006-01-02T15",
+		"2006-01-02",
+	} {
+		if t, err := time.ParseInLocation(layout, value, now.Location()); err == nil {
+			return t.UTC().Format(time.RFC3339Nano), nil
+		}
+	}
+
+	seconds, fraction, hasFraction := strings.Cut(value, ".")
+	sec, err := strconv.ParseInt(seconds, 10, 64)
+	if err != nil {
+		return "", fmt.Errorf("failed to parse '%s' as a time or duration", value)
+	}
+
+	var nsec int64
+	if hasFraction {
+		if len(fraction) == 0 || len(fraction) > 9 || strings.ContainsAny(fraction, "+-") {
+			return "", fmt.Errorf("invalid Unix timestamp fraction in '%s'", value)
+		}
+		nsec, err = strconv.ParseInt(fraction+strings.Repeat("0", 9-len(fraction)), 10, 64)
+		if err != nil {
+			return "", fmt.Errorf("invalid Unix timestamp fraction in '%s': %w", value, err)
+		}
+	}
+	t := time.Unix(sec, nsec).UTC()
+	if t.Year() < 0 || t.Year() > 9999 {
+		return "", fmt.Errorf("Unix timestamp '%s' is outside the RFC 3339 date range", value)
+	}
+	return t.Format(time.RFC3339Nano), nil
+}

internal/cli/logs/time_test.go

@@ -0,0 +1,87 @@
+package logs
+
+import (
+	"testing"
+	"time"
+
+	timetypes "github.com/docker/docker/api/types/time"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestTimeRange(t *testing.T) {
+	t.Parallel()
+
+	location, err := time.LoadLocation("Australia/Sydney")
+	require.NoError(t, err)
+	// October is daylight-saving time, but July timestamps must use the winter offset.
+	// Daylight saving starts on 4 October. Day durations must still mean 24 elapsed hours.
+	now := time.Date(2026, 10, 5, 12, 0, 0, 123456789, location)
+	tests := []struct {
+		input string
+		want  string
+	}{
+		{"", ""},
+		{"2026-07-01", "2026-06-30T14:00:00Z"},
+		{"2026-07-01T10", "2026-07-01T00:00:00Z"},
+		{"2026-07-01T10:30", "2026-07-01T00:30:00Z"},
+		{"2026-07-01T10:30:45.123456789", "2026-07-01T00:30:45.123456789Z"},
+		{"2026-01-01T10:00:00", "2025-12-31T23:00:00Z"},
+		{"2026-07-01T10:30:45Z", "2026-07-01T10:30:45Z"},
+		{"2026-07-01T10:30:45+02:00", "2026-07-01T08:30:45Z"},
+		{"2026-07-01T10:30:45-04:00", "2026-07-01T14:30:45Z"},
+		{"2026-07-01T10+02:00", "2026-07-01T08:00:00Z"},
+		{"2026-07-01T10:30+02:00", "2026-07-01T08:30:00Z"},
+		{"2026-07-01+02:00", "2026-06-30T22:00:00Z"},
+		{"1763953966", "2025-11-24T03:12:46Z"},
+		{"1763953966.000000001", "2025-11-24T03:12:46.000000001Z"},
+		{"0", "1970-01-01T00:00:00Z"},
+		{"2m30s", "2026-10-05T00:57:30.123456789Z"},
+		{"1d", "2026-10-04T01:00:00.123456789Z"},
+		{"2d", "2026-10-03T01:00:00.123456789Z"},
+		{"2d3h", "2026-10-02T22:00:00.123456789Z"},
+		{"1.5d", "2026-10-03T13:00:00.123456789Z"},
+		{"-2d", "2026-10-07T01:00:00.123456789Z"},
+		{"-1h", "2026-10-05T02:00:00.123456789Z"},
+	}
+	for _, tt := range tests {
+		t.Run(tt.input, func(t *testing.T) {
+			since, until, err := TimeRange(tt.input, tt.input, now)
+			require.NoError(t, err)
+			assert.Equal(t, tt.want, since)
+			assert.Equal(t, tt.want, until)
+
+			if since != "" {
+				// The server-side Docker SDK must preserve the cutoff even in another timezone.
+				actual, err := timetypes.GetTimestamp(since, now.In(time.UTC))
+				require.NoError(t, err)
+				expected, err := time.Parse(time.RFC3339Nano, tt.want)
+				require.NoError(t, err)
+				sec, nsec, err := timetypes.ParseTimestamps(actual, 0)
+				require.NoError(t, err)
+				assert.True(t, expected.Equal(time.Unix(sec, nsec)))
+			}
+		})
+	}
+
+	since, until, err := TimeRange("3h", "1h30m", now)
+	require.NoError(t, err)
+	assert.Equal(t, "2026-10-04T22:00:00.123456789Z", since)
+	assert.Equal(t, "2026-10-04T23:30:00.123456789Z", until)
+}
+
+func TestTimeRange_Invalid(t *testing.T) {
+	t.Parallel()
+
+	for _, input := range []string{
+		"invalid", "2026-02-30", "2026-01-01T25:00:00", "1763953966.xyz",
+		"1763953966.1234567890", "253402300800",
+	} {
+		t.Run(input, func(t *testing.T) {
+			_, _, err := TimeRange(input, "", time.Now())
+			require.ErrorContains(t, err, "invalid --since value")
+			_, _, err = TimeRange("", input, time.Now())
+			require.ErrorContains(t, err, "invalid --until value")
+		})
+	}
+}

internal/cli/tui/print.go

@@ -3,9 +3,11 @@ package tui
 import (
 	"fmt"
 	"os"
+
+	"charm.land/lipgloss/v2"
 )
 
 func PrintWarning(msg string) {
 	styledMsg := BoldYellow.Render(fmt.Sprintf("WARNING: %s", msg))
-	fmt.Fprintln(os.Stderr, styledMsg)
+	lipgloss.Fprintln(os.Stderr, styledMsg)
 }

internal/journal/journal.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"io"
 	"os/exec"
+	"time"
 
 	"github.com/psviderski/uncloud/pkg/api"
 )
@@ -31,11 +32,11 @@ func logs(ctx context.Context, unit string, opts api.ServiceLogsOptions) (io.Rea
 
 	if opts.Since != "" {
 		args = append(args, "-S")
-		args = append(args, opts.Since)
+		args = append(args, journalTimestamp(opts.Since))
 	}
 	if opts.Until != "" {
 		args = append(args, "-U")
-		args = append(args, opts.Until)
+		args = append(args, journalTimestamp(opts.Until))
 	}
 
 	cmd := commandContext(ctx, journalctl, args...)
@@ -51,6 +52,18 @@ func logs(ctx context.Context, unit string, opts api.ServiceLogsOptions) (io.Rea
 	return p, cmd.Wait, nil
 }
 
+// journalTimestamp formats normalised client timestamps for journalctl versions that do not
+// accept RFC 3339 timezone suffixes. Support for timestamps containing T and Z was added in
+// systemd 255 (6 December 2023). Debian 12 ships systemd 252, so it still needs this conversion.
+// Journald timestamps have microsecond precision.
+func journalTimestamp(value string) string {
+	if t, err := time.Parse(time.RFC3339Nano, value); err == nil {
+		return t.UTC().Format("2006-01-02 15:04:05.999999") + " UTC"
+	}
+	// Preserve raw filters from older clients and SDK callers.
+	return value
+}
+
 // follow synchronously follows the io.Reader, writing each new journal entry to channel.
 // It stops when the reader is exhausted or the context is cancelled.
 func follow(ctx context.Context, reader io.Reader, outCh chan api.LogEntry) {

internal/journal/logs_test.go

@@ -115,6 +115,9 @@ func TestEntry(t *testing.T) {
 }
 
 func TestLogs(t *testing.T) {
+	originalCommandContext := commandContext
+	t.Cleanup(func() { commandContext = originalCommandContext })
+
 	commandContext = func(ctx context.Context, _ string, _ ...string) *exec.Cmd {
 		return exec.CommandContext(ctx, "/usr/bin/tail", "testdata/logs")
 	}
@@ -150,3 +153,38 @@ func TestLogs(t *testing.T) {
 	// Still six because heartbeats are not written here and Tail is ignored as the command is overridden.
 	assert.Equal(t, 6, i)
 }
+
+func TestLogs_TimeFilters(t *testing.T) {
+	originalCommandContext := commandContext
+	t.Cleanup(func() { commandContext = originalCommandContext })
+
+	var args []string
+	commandContext = func(ctx context.Context, command string, commandArgs ...string) *exec.Cmd {
+		assert.Equal(t, "journalctl", command)
+		args = commandArgs
+		return exec.CommandContext(ctx, "/usr/bin/tail", "testdata/logs")
+	}
+
+	ch, err := Logs(context.Background(), "uncloud", api.ServiceLogsOptions{
+		Since: "2026-07-01T10:30:45.123456789+10:00",
+		Until: "2026-07-01T01:30:45Z",
+	})
+	require.NoError(t, err)
+	for range ch {
+	}
+	assert.Equal(t, []string{
+		"-u", "uncloud", "--no-hostname", "-n", "0", "-o", "short-unix",
+		"-S", "2026-07-01 00:30:45.123456 UTC",
+		"-U", "2026-07-01 01:30:45 UTC",
+	}, args)
+
+	// Older clients and SDK callers can still pass journalctl's native filters.
+	ch, err = Logs(context.Background(), "uncloud", api.ServiceLogsOptions{Since: "1h ago", Until: "today"})
+	require.NoError(t, err)
+	for range ch {
+	}
+	assert.Equal(t, []string{
+		"-u", "uncloud", "--no-hostname", "-n", "0", "-o", "short-unix",
+		"-S", "1h ago", "-U", "today",
+	}, args)
+}

internal/machine/cluster/cluster.go

@@ -14,6 +14,7 @@ import (
 	"github.com/psviderski/uncloud/internal/machine/network"
 	"github.com/psviderski/uncloud/internal/machine/store"
 	"github.com/psviderski/uncloud/internal/secret"
+	"github.com/psviderski/uncloud/pkg/api"
 	"google.golang.org/grpc/codes"
 	"google.golang.org/grpc/status"
 	"google.golang.org/protobuf/types/known/emptypb"
@@ -87,6 +88,11 @@ func (c *Cluster) AddMachine(ctx context.Context, req *pb.AddMachineRequest) (*p
 func (c *Cluster) AddMachineWithoutReadyCheck(
 	ctx context.Context, req *pb.AddMachineRequest,
 ) (*pb.AddMachineResponse, error) {
+	if req.Name != "" {
+		if err := api.ValidateMachineName(req.Name); err != nil {
+			return nil, status.Error(codes.InvalidArgument, err.Error())
+		}
+	}
 	if req.Network == nil {
 		return nil, status.Error(codes.InvalidArgument, "network not set")
 	}

internal/machine/cluster/cluster_test.go

@@ -0,0 +1,28 @@
+package cluster
+
+import (
+	"context"
+	"testing"
+
+	"github.com/psviderski/uncloud/api/pb"
+	"github.com/stretchr/testify/require"
+	"google.golang.org/grpc/codes"
+	"google.golang.org/grpc/status"
+)
+
+func TestAddMachine_InvalidName(t *testing.T) {
+	t.Parallel()
+
+	ready := make(chan struct{})
+	close(ready)
+	// Invalid names must be rejected before any store access.
+	c := NewCluster(nil, nil, nil, ready)
+	for _, name := range []string{"VPS1", "worker.example", "rr", "nearest", "c337f00600de51ef4375c9a9a267dba5"} {
+		t.Run("add", func(t *testing.T) {
+			resp, err := c.AddMachine(context.Background(), &pb.AddMachineRequest{Name: name})
+			require.Nil(t, resp)
+			require.Equal(t, codes.InvalidArgument, status.Code(err))
+			require.ErrorContains(t, err, "invalid machine name")
+		})
+	}
+}

internal/machine/cluster/machine.go

@@ -6,6 +6,7 @@ import (
 	"strings"
 
 	"github.com/psviderski/uncloud/internal/secret"
+	"github.com/psviderski/uncloud/pkg/api"
 )
 
 // NewMachineID generates a new unique machine ID.
@@ -27,7 +28,7 @@ func NewRandomMachineName() (string, error) {
 // a numeric suffix ("-1", "-2", etc.) if needed.
 func DefaultMachineName(hostname string, existing []string) (string, error) {
 	name := machineNameFromHostname(hostname)
-	if name == "" {
+	if api.ValidateMachineName(name) != nil {
 		var err error
 		if name, err = NewRandomMachineName(); err != nil {
 			return "", err
@@ -38,7 +39,9 @@ func DefaultMachineName(hostname string, existing []string) (string, error) {
 		return name, nil
 	}
 	for i := 1; ; i++ {
-		candidate := fmt.Sprintf("%s-%d", name, i)
+		suffix := fmt.Sprintf("-%d", i)
+		base := strings.TrimRight(name[:min(len(name), 63-len(suffix))], "-")
+		candidate := base + suffix
 		if !slices.Contains(existing, candidate) {
 			return candidate, nil
 		}
@@ -55,12 +58,13 @@ func machineNameFromHostname(hostname string) string {
 	var b strings.Builder
 	for _, r := range label {
 		switch {
-		case r >= 'a' && r <= 'z', r >= '0' && r <= '9', r == '-', r == '_':
+		case r >= 'a' && r <= 'z', r >= '0' && r <= '9', r == '-':
 			b.WriteRune(r)
 		default:
 			b.WriteRune('-')
 		}
 	}
 	// Trim leading and trailing hyphens that may result from the sanitisation above.
-	return strings.Trim(b.String(), "-")
+	name := strings.Trim(b.String(), "-")
+	return strings.TrimRight(name[:min(len(name), 63)], "-")
 }

internal/machine/cluster/machine_test.go

@@ -1,8 +1,10 @@
 package cluster
 
 import (
+	"strings"
 	"testing"
 
+	"github.com/psviderski/uncloud/pkg/api"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
@@ -18,9 +20,11 @@ func TestMachineNameFromHostname(t *testing.T) {
 		{"simple", "web", "web"},
 		{"fqdn uses first label", "web-1.example.com", "web-1"},
 		{"uppercase lowercased", "Web-Server", "web-server"},
-		{"invalid chars replaced", "host_name@1", "host_name-1"},
-		{"trim surrounding hyphens", "-_host_-", "_host_"},
+		{"invalid chars replaced", "host_name@1", "host-name-1"},
+		{"trim surrounding hyphens", "-_host_-", "host"},
 		{"whitespace trimmed", "  myhost  ", "myhost"},
+		{"long hostname truncated", strings.Repeat("a", 64), strings.Repeat("a", 63)},
+		{"truncation trims trailing hyphen", strings.Repeat("a", 62) + "-b", strings.Repeat("a", 62)},
 		{"empty", "", ""},
 		{"only invalid chars", "@#", ""},
 		{"dot only", ".example.com", ""},
@@ -37,10 +41,12 @@ func TestDefaultMachineName(t *testing.T) {
 	t.Parallel()
 
 	t.Run("falls back to random", func(t *testing.T) {
-		// An empty or fully invalid hostname falls back to a random "machine-xxxx" name.
-		got, err := DefaultMachineName("***", nil)
-		require.NoError(t, err)
-		assert.Regexp(t, `^machine-[a-zA-Z0-9]{4}$`, got)
+		for _, hostname := range []string{"***", "rr", "nearest", strings.Repeat("a", 32)} {
+			got, err := DefaultMachineName(hostname, nil)
+			require.NoError(t, err)
+			assert.Regexp(t, `^machine-[a-z0-9]{4}$`, got)
+			require.NoError(t, api.ValidateMachineName(got))
+		}
 	})
 
 	tests := []struct {
@@ -49,10 +55,45 @@ func TestDefaultMachineName(t *testing.T) {
 		existing []string
 		want     string
 	}{
-		{"from hostname", "web-1.example.com", nil, "web-1"},
-		{"dedup against existing", "web", []string{"web"}, "web-1"},
-		{"dedup multiple", "web", []string{"web", "web-1"}, "web-2"},
-		{"sanitized", "My_Host", nil, "my_host"},
+		{
+			name:     "from hostname",
+			hostname: "web-1.example.com",
+			want:     "web-1",
+		},
+		{
+			name:     "dedup against existing",
+			hostname: "web",
+			existing: []string{"web"},
+			want:     "web-1",
+		},
+		{
+			name:     "dedup multiple",
+			hostname: "web",
+			existing: []string{"web", "web-1"},
+			want:     "web-2",
+		},
+		{
+			name:     "sanitized",
+			hostname: "My_Host",
+			want:     "my-host",
+		},
+		{
+			name:     "long hostname",
+			hostname: strings.Repeat("a", 64),
+			want:     strings.Repeat("a", 63),
+		},
+		{
+			name:     "dedup maximum length",
+			hostname: strings.Repeat("a", 63),
+			existing: []string{strings.Repeat("a", 63)},
+			want:     strings.Repeat("a", 61) + "-1",
+		},
+		{
+			name:     "dedup trims trailing hyphen",
+			hostname: strings.Repeat("a", 60) + "-bb",
+			existing: []string{strings.Repeat("a", 60) + "-bb"},
+			want:     strings.Repeat("a", 60) + "-1",
+		},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
@@ -60,6 +101,7 @@ func TestDefaultMachineName(t *testing.T) {
 			got, err := DefaultMachineName(tt.hostname, tt.existing)
 			require.NoError(t, err)
 			assert.Equal(t, tt.want, got)
+			require.NoError(t, api.ValidateMachineName(got))
 		})
 	}
 }

internal/machine/dns/resolver.go

@@ -10,6 +10,8 @@ import (
 	"sync"
 	"time"
 
+	"github.com/psviderski/uncloud/api/pb"
+	"github.com/psviderski/uncloud/internal/machine/network"
 	"github.com/psviderski/uncloud/internal/machine/store"
 )
 
@@ -19,7 +21,9 @@ type ClusterResolver struct {
 	store *store.Store
 	// serviceIPs maps service names to container IPs.
 	serviceIPs map[string][]netip.Addr
-	// mu protects the serviceIPs map.
+	// machineIPs maps machine IDs and names to their IP.
+	machineIPs map[string]netip.Addr
+	// mu protects the serviceIPs and machineIPs map.
 	mu sync.RWMutex
 	// lastUpdate tracks when records were last updated.
 	lastUpdate time.Time
@@ -46,6 +50,13 @@ func (r *ClusterResolver) Run(ctx context.Context) error {
 	// TODO: implement machine membership check using Corrossion Admin client to filter available containers.
 	r.updateServiceIPs(containers)
 
+	machines, mchanges, err := r.store.SubscribeMachines(ctx)
+	if err != nil {
+		return fmt.Errorf("subscribe to machine changes: %w", err)
+	}
+	r.log.Info("Subscribed to machine changes in the clsuter to keep machine DNS records updated.")
+	r.updateMachineIPs(machines)
+
 	for {
 		select {
 		case _, ok := <-changes:
@@ -59,9 +70,22 @@ func (r *ClusterResolver) Run(ctx context.Context) error {
 				r.log.Error("Failed to list containers.", "err", err)
 				continue
 			}
-
 			// TODO: implement machine membership check using Corrossion Admin client to filter available containers.
 			r.updateServiceIPs(containers)
+
+		case _, ok := <-mchanges:
+			if !ok {
+				return fmt.Errorf("machine subscription failed")
+			}
+			r.log.Debug("Cluster machines changed, updating DNS records.")
+
+			machines, err := r.store.ListMachines(ctx)
+			if err != nil {
+				r.log.Error("Failed to list machines.", "err", err)
+				continue
+			}
+			r.updateMachineIPs(machines)
+
 		case <-ctx.Done():
 			return nil
 		}
@@ -121,19 +145,49 @@ func (r *ClusterResolver) updateServiceIPs(containers []store.ContainerRecord) {
 	r.log.Info("DNS records updated.", "services", len(newServiceIPs)/3, "containers", containersCount)
 }
 
-// Resolve returns IP addresses of the service containers.
-func (r *ClusterResolver) Resolve(serviceName string) []netip.Addr {
+func (r *ClusterResolver) updateMachineIPs(machines []*pb.MachineInfo) {
+	newMachineIPs := make(map[string]netip.Addr, len(machines))
+
+	for _, machine := range machines {
+		subnet, err := machine.Network.Subnet.ToPrefix()
+		if err != nil {
+			continue
+		}
+		addr := network.MachineIP(subnet)
+		newMachineIPs[machine.Name+".m"] = addr
+		newMachineIPs[machine.Id+".m"] = addr
+	}
+	r.mu.Lock()
+	r.machineIPs = newMachineIPs
+	r.mu.Unlock()
+
+	r.log.Info("DNS records updated.", "machines", len(machines))
+}
+
+// Resolve returns IP addresses of the service containers or machines.
+func (r *ClusterResolver) Resolve(name string) []netip.Addr {
 	r.mu.RLock()
 	defer r.mu.RUnlock()
 
-	ips, ok := r.serviceIPs[serviceName]
-	if !ok || len(ips) == 0 {
-		return nil
+	// Return a copy of the IPs slice to prevent modification of the original.
+	ips, ok := r.serviceIPs[name]
+	if ok && len(ips) > 0 {
+		return slices.Clone(ips)
 	}
 
-	// Return a copy of the IPs slice to prevent modification of the original.
-	ipsCopy := make([]netip.Addr, len(ips))
-	copy(ipsCopy, ips)
+	ip, ok := r.machineIPs[name]
+	if ok {
+		return slices.Clone([]netip.Addr{ip})
+	}
 
-	return ipsCopy
+	if name == "m" { // all machines subdomain
+		// collect all unique IP address
+		uniq := map[string]netip.Addr{}
+		for _, addr := range r.machineIPs {
+			uniq[addr.String()] = addr
+		}
+		return slices.Clone(slices.Collect(maps.Values(uniq)))
+	}
+
+	return nil
 }

internal/machine/dns/resolver_test.go

@@ -1,11 +1,13 @@
 package dns
 
 import (
+	"net/netip"
 	"reflect"
 	"testing"
 
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/network"
+	"github.com/psviderski/uncloud/api/pb"
 	"github.com/psviderski/uncloud/internal/machine/store"
 	"github.com/psviderski/uncloud/pkg/api"
 	"github.com/stretchr/testify/assert"
@@ -75,3 +77,32 @@ func newRecord(serviceID, serviceName, ip, machineID string) store.ContainerReco
 		MachineID: machineID,
 	}
 }
+
+func TestClusterResolver_UpdateMachineIPs(t *testing.T) {
+	t.Parallel()
+
+	machines := []*pb.MachineInfo{
+		newMachineRecord("x0y0z0", "mach-1", "10.210.0.0/24"),
+		newMachineRecord("x1y1z1", "mach-2", "10.210.1.0/24"),
+		newMachineRecord("x2y2z2", "mach-3", "10.210.2.0/24"),
+	}
+
+	r := NewClusterResolver(nil)
+	r.updateMachineIPs(machines)
+
+	assert.NotEmpty(t, r.Resolve("mach-1.m"))
+	assert.NotEmpty(t, r.Resolve("mach-3.m"))
+	assert.NotEmpty(t, r.Resolve("x0y0z0.m"))
+
+	assert.Equal(t, 3, len(r.Resolve("m")))
+}
+
+func newMachineRecord(machineID, machineName, prefix string) *pb.MachineInfo {
+	return &pb.MachineInfo{
+		Id:   machineID,
+		Name: machineName,
+		Network: &pb.NetworkConfig{
+			Subnet: pb.NewIPPrefix(netip.MustParsePrefix(prefix)),
+		},
+	}
+}

internal/machine/dns/server.go

@@ -293,13 +293,13 @@ func (s *Server) forwardRequest(req *dns.Msg, proto string) (*dns.Msg, error) {
 // handleAQuery processes an A query for the internal domain and returns A records for the requested name.
 // The internal domain suffix is already stripped from the name. An empty list is returned if no records are found.
 func (s *Server) handleAQuery(name string) []dns.RR {
-	serviceName, mode := extractModeFromDomain(trimInternalDomain(name))
-	ips := s.resolver.Resolve(serviceName)
+	unname, mode := extractModeFromDomain(trimInternalDomain(name))
+	ips := s.resolver.Resolve(unname)
 	if len(ips) == 0 {
-		s.log.Debug("Failed to resolve service name.", "service", serviceName)
+		s.log.Debug("Failed to resolve internal name.", "name", name)
 		return nil
 	}
-	s.log.Debug("Resolved service name.", "service", serviceName, "ips", ips)
+	s.log.Debug("Resolved service internal name.", "name", name, "ips", ips)
 
 	if len(ips) > 1 {
 		// Shuffle the IPs to approximate round-robin.

internal/machine/machine.go

@@ -836,6 +836,11 @@ func (m *Machine) InitCluster(ctx context.Context, req *pb.InitClusterRequest) (
 	if m.Initialised() {
 		return nil, status.Error(codes.FailedPrecondition, "machine is already configured as a cluster member")
 	}
+	if req.MachineName != "" {
+		if err := api.ValidateMachineName(req.MachineName); err != nil {
+			return nil, status.Error(codes.InvalidArgument, err.Error())
+		}
+	}
 
 	clusterNetwork, err := req.Network.ToPrefix()
 	if err != nil {
@@ -1220,8 +1225,8 @@ func (m *Machine) applyMachineUpdate(ctx context.Context, req *pb.UpdateMachineR
 	defer m.state.mu.Unlock()
 
 	if req.Name != nil {
-		if *req.Name == "" {
-			return status.Error(codes.InvalidArgument, "machine name cannot be empty")
+		if err := api.ValidateMachineName(*req.Name); err != nil {
+			return status.Error(codes.InvalidArgument, err.Error())
 		}
 		// Check for duplicate names across the cluster, excluding this machine.
 		if *req.Name != m.state.Name {

internal/machine/machine_test.go

@@ -0,0 +1,38 @@
+package machine
+
+import (
+	"context"
+	"testing"
+
+	"github.com/psviderski/uncloud/api/pb"
+	"github.com/stretchr/testify/require"
+	"google.golang.org/grpc/codes"
+	"google.golang.org/grpc/status"
+)
+
+func TestInitCluster_InvalidMachineName(t *testing.T) {
+	t.Parallel()
+
+	for _, name := range []string{"VPS1", "rr", "nearest", "c337f00600de51ef4375c9a9a267dba5"} {
+		// Invalid names must be rejected before initializing cluster state.
+		m := &Machine{state: &State{}}
+
+		resp, err := m.InitCluster(context.Background(), &pb.InitClusterRequest{MachineName: name})
+		require.Nil(t, resp)
+		require.Equal(t, codes.InvalidArgument, status.Code(err))
+		require.ErrorContains(t, err, "invalid machine name")
+	}
+}
+
+func TestUpdateMachine_InvalidName(t *testing.T) {
+	t.Parallel()
+
+	for _, name := range []string{"", "VPS1", "worker.example", "rr", "nearest", "c337f00600de51ef4375c9a9a267dba5"} {
+		m := &Machine{state: &State{ID: "machine-id", Name: "worker"}}
+		resp, err := m.UpdateMachine(context.Background(), &pb.UpdateMachineRequest{Name: &name})
+		require.Nil(t, resp)
+		require.Equal(t, codes.InvalidArgument, status.Code(err))
+		require.ErrorContains(t, err, "invalid machine name")
+		require.Equal(t, "worker", m.state.Name)
+	}
+}

internal/machine/store/container.go

@@ -23,6 +23,10 @@ const (
 	// SyncStatusOutdated indicates that a container record may be outdated, for example, due to being unable
 	// to retrieve the container's state from the Docker daemon or when the machine is being stopped or restarted.
 	SyncStatusOutdated = "outdated"
+
+	// containerChangesDebounceInterval defines how long to wait before notifying subscribers about container changes.
+	// Multiple changes within this window are grouped into a single notification to prevent system overload.
+	containerChangesDebounceInterval = 250 * time.Millisecond
 )
 
 type ContainerRecord struct {
@@ -267,6 +271,13 @@ func (s *Store) SubscribeContainers(ctx context.Context) ([]ContainerRecord, <-c
 	changes := make(chan struct{})
 	go func() {
 		defer close(changes)
+		// Coalesce bursts of rapid-fire row-level change events (e.g. from frequent health-check updates
+		// across many services) into a single signal per debounce window, instead of forwarding one signal
+		// per event. Every subscriber of this channel (e.g. the Caddy and DNS reconcilers) otherwise reruns
+		// its full reconciliation on every single event, which can burn significant CPU across the cluster
+		// when there's a lot of container churn.
+		var debouncer *time.Timer
+		var debouncerCh <-chan time.Time
 		for {
 			select {
 			case <-ctx.Done():
@@ -279,8 +290,21 @@ func (s *Store) SubscribeContainers(ctx context.Context) ([]ContainerRecord, <-c
 					}
 					return
 				}
-				// Just signal that there is a change in the containers list.
-				changes <- struct{}{}
+				if debouncerCh == nil {
+					if debouncer == nil {
+						debouncer = time.NewTimer(containerChangesDebounceInterval)
+					} else {
+						debouncer.Reset(containerChangesDebounceInterval)
+					}
+					debouncerCh = debouncer.C
+				}
+			case <-debouncerCh:
+				select {
+				case changes <- struct{}{}:
+				case <-ctx.Done():
+					return
+				}
+				debouncerCh = nil
 			}
 		}
 	}()

pkg/api/machine.go

@@ -1,12 +1,33 @@
 package api
 
 import (
+	"fmt"
 	"net/netip"
 	"strings"
 
 	"github.com/psviderski/uncloud/api/pb"
 )
 
+// ValidateMachineName checks that a machine name is a lowercase DNS label and doesn't conflict with
+// internal DNS query modes or machine IDs.
+func ValidateMachineName(name string) error {
+	if !DNSLabelRegex.MatchString(name) {
+		return fmt.Errorf("invalid machine name %q: must be 1-63 characters, lowercase letters, numbers, "+
+			"and hyphens only, starting and ending with a letter or number", name)
+	}
+
+	switch name {
+	case "rr", "nearest":
+		return fmt.Errorf("invalid machine name %q: reserved for internal DNS query modes", name)
+	}
+	if IDRegex.MatchString(name) {
+		return fmt.Errorf(
+			"invalid machine name %q: must not match the machine ID format (32 hexadecimal characters)", name)
+	}
+
+	return nil
+}
+
 // MachineFilter defines criteria to filter machines in ListMachines.
 type MachineFilter struct {
 	// Available filters machines that are not DOWN.

pkg/api/machine_test.go

@@ -3,6 +3,7 @@ package api
 import (
 	"encoding/json"
 	"net/netip"
+	"strings"
 	"testing"
 
 	"github.com/psviderski/uncloud/api/pb"
@@ -10,6 +11,52 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
+func TestValidateMachineName(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name    string
+		input   string
+		wantErr string
+	}{
+		{name: "single letter", input: "a"},
+		{name: "single digit", input: "1"},
+		{name: "two characters", input: "a1"},
+		{name: "generated name", input: "machine-ab12"},
+		{name: "hyphens and digits", input: "1-worker-2"},
+		{name: "maximum length", input: strings.Repeat("a", 63)},
+		{name: "maximum length with hyphens", input: "a" + strings.Repeat("-", 61) + "1"},
+		{name: "machine namespace label", input: "m"},
+		{name: "mode prefix", input: "nearest-worker"},
+		{name: "short hexadecimal name", input: strings.Repeat("a", 31)},
+		{name: "long hexadecimal name", input: strings.Repeat("a", 33)},
+		{name: "non-hexadecimal 32 characters", input: strings.Repeat("g", 32)},
+		{name: "empty", wantErr: "must be 1-63 characters"},
+		{name: "too long", input: strings.Repeat("a", 64), wantErr: "must be 1-63 characters"},
+		{name: "uppercase", input: "VPS1", wantErr: "lowercase letters"},
+		{name: "leading hyphen", input: "-worker", wantErr: "starting and ending"},
+		{name: "trailing hyphen", input: "worker-", wantErr: "starting and ending"},
+		{name: "underscore", input: "worker_1", wantErr: "hyphens only"},
+		{name: "dot", input: "worker.example", wantErr: "hyphens only"},
+		{name: "space", input: "worker 1", wantErr: "hyphens only"},
+		{name: "leading whitespace", input: " worker", wantErr: "hyphens only"},
+		{name: "trailing whitespace", input: "worker\t", wantErr: "hyphens only"},
+		{name: "non-ASCII", input: "wörker", wantErr: "lowercase letters"},
+		{name: "round-robin mode", input: "rr", wantErr: "reserved for internal DNS query modes"},
+		{name: "nearest mode", input: "nearest", wantErr: "reserved for internal DNS query modes"},
+		{name: "machine ID", input: "c337f00600de51ef4375c9a9a267dba5", wantErr: "machine ID format"},
+	}
+
+	for _, tt := range tests {
+		err := ValidateMachineName(tt.input)
+		if tt.wantErr == "" {
+			require.NoError(t, err)
+		} else {
+			require.ErrorContains(t, err, tt.wantErr)
+		}
+	}
+}
+
 func TestMachineMembersList_Info(t *testing.T) {
 	t.Parallel()
 

pkg/api/service.go

@@ -41,12 +41,30 @@ const (
 )
 
 var (
-	serviceIDRegexp = regexp.MustCompile("^[0-9a-f]{32}$")
-	dnsLabelRegexp  = regexp.MustCompile(`^[a-z0-9]([-a-z0-9]*[a-z0-9])?$`)
+	IDRegex       = regexp.MustCompile("^[0-9a-f]{32}$")
+	DNSLabelRegex = regexp.MustCompile(`^[a-z0-9]([-a-z0-9]{0,61}[a-z0-9])?$`)
 )
 
 func ValidateServiceID(id string) bool {
-	return serviceIDRegexp.MatchString(id)
+	return IDRegex.MatchString(id)
+}
+
+// ValidateServiceName checks that a service name is a lowercase DNS label and doesn't conflict with
+// the machine DNS namespace or service IDs.
+func ValidateServiceName(name string) error {
+	if !DNSLabelRegex.MatchString(name) {
+		return fmt.Errorf("invalid service name %q: must be 1-63 characters, lowercase letters, numbers, "+
+			"and hyphens only, starting and ending with a letter or number", name)
+	}
+	if name == "m" {
+		return fmt.Errorf("invalid service name %q: reserved for the machine DNS namespace", name)
+	}
+	if IDRegex.MatchString(name) {
+		return fmt.Errorf(
+			"invalid service name %q: must not match the service ID format (32 hexadecimal characters)", name)
+	}
+
+	return nil
 }
 
 // ServiceSpec defines the desired state of a service.
@@ -147,12 +165,8 @@ func (s *ServiceSpec) Validate() error {
 	}
 
 	if s.Name != "" {
-		if len(s.Name) > 63 {
-			return fmt.Errorf("service name too long (max 63 characters): %q", s.Name)
-		}
-		if !dnsLabelRegexp.MatchString(s.Name) {
-			return fmt.Errorf("invalid service name: %q. must be 1-63 characters, lowercase letters, numbers, "+
-				"and dashes only; must start and end with a letter or number", s.Name)
+		if err := ValidateServiceName(s.Name); err != nil {
+			return err
 		}
 	}
 

pkg/api/service_test.go

@@ -2,6 +2,7 @@ package api
 
 import (
 	"os"
+	"strings"
 	"testing"
 
 	"github.com/docker/docker/api/types/container"
@@ -9,6 +10,59 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
+func TestServiceSpec_Validate_Name(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name    string
+		input   string
+		wantErr string
+	}{
+		{name: "empty allows generated name"},
+		{name: "single character", input: "a"},
+		{name: "single digit", input: "1"},
+		{name: "two characters", input: "a1"},
+		{name: "hyphens and digits", input: "1-web-2"},
+		{name: "consecutive hyphens", input: "web--1"},
+		{name: "maximum length", input: strings.Repeat("a", 63)},
+		{name: "maximum length with hyphens", input: "a" + strings.Repeat("-", 61) + "1"},
+		{name: "round-robin mode is a valid service name", input: "rr"},
+		{name: "nearest mode is a valid service name", input: "nearest"},
+		{name: "machine namespace prefix", input: "m-service"},
+		{name: "short hexadecimal name", input: strings.Repeat("a", 31)},
+		{name: "long hexadecimal name", input: strings.Repeat("a", 33)},
+		{name: "non-hexadecimal 32 characters", input: strings.Repeat("g", 32)},
+		{name: "too long", input: strings.Repeat("a", 64), wantErr: "must be 1-63 characters"},
+		{name: "uppercase", input: "WEB1", wantErr: "lowercase letters"},
+		{name: "leading hyphen", input: "-web", wantErr: "starting and ending"},
+		{name: "trailing hyphen", input: "web-", wantErr: "starting and ending"},
+		{name: "hyphen only", input: "-", wantErr: "starting and ending"},
+		{name: "underscore", input: "web_1", wantErr: "hyphens only"},
+		{name: "dot", input: "web.example", wantErr: "hyphens only"},
+		{name: "slash", input: "web/api", wantErr: "hyphens only"},
+		{name: "space", input: "web 1", wantErr: "hyphens only"},
+		{name: "leading whitespace", input: " web", wantErr: "hyphens only"},
+		{name: "trailing whitespace", input: "web ", wantErr: "hyphens only"},
+		{name: "tab", input: "web\t1", wantErr: "hyphens only"},
+		{name: "newline", input: "web\n", wantErr: "hyphens only"},
+		{name: "non-ASCII", input: "wéb", wantErr: "lowercase letters"},
+		{name: "machine DNS namespace", input: "m", wantErr: "reserved for the machine DNS namespace"},
+		{name: "service ID", input: "c337f00600de51ef4375c9a9a267dba5", wantErr: "service ID format"},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+			spec := ServiceSpec{Name: tt.input, Container: ContainerSpec{Image: "nginx:latest"}}
+			err := spec.Validate()
+			if tt.wantErr != "" {
+				require.ErrorContains(t, err, tt.wantErr)
+			} else {
+				require.NoError(t, err)
+			}
+		})
+	}
+}
+
 func TestServiceSpec_Validate_CaddyAndPorts(t *testing.T) {
 	tests := []struct {
 		name    string

website/docs/3-concepts/2-ingress/4-cluster-storage-for-caddy.md

@@ -15,6 +15,15 @@ includes it and configure Caddy to use it.
 
 ## Enabling cluster storage
 
+:::warning Certificate migration
+
+Caddy doesn't migrate existing certificates automatically when you change its storage backend. **All certificates will
+need to be reissued unless you migrate your existing storage first.** Use the experimental
+[`caddy storage export` and `caddy storage import`](https://caddyserver.com/docs/command-line#caddy-storage)
+commands with the old and new configs to transfer the storage contents.
+
+:::
+
 :::info Requirements
 
 Cluster storage requires Uncloud **v0.21.0 or newer** for both the `uc` CLI and the daemon on every cluster machine.
@@ -37,7 +46,7 @@ Create a global Caddyfile, or add `storage uncloud` to the global options in you
 Deploy Caddy with the pre-built module image and your global config:
 
 ```shell
-uc caddy deploy --image ghcr.io/unlabs-dev/caddy-uncloud:0.1.1 --caddyfile global.Caddyfile
+uc caddy deploy --image ghcr.io/unlabs-dev/caddy-uncloud:0.1.3 --caddyfile global.Caddyfile
 ```
 
 See the [module README](https://github.com/unlabs-dev/caddy-uncloud#usage) for custom image builds, Compose deployment,
@@ -55,4 +64,10 @@ List issued certificates in cluster storage with [`uc caddy cert ls`](../../9-cl
 
 ```shell
 uc caddy cert ls
+
+ID             NAME                            ISSUER                    EXPIRES
+c111e23615f7   dns.uncloud.run                 Let's Encrypt             2026-12-31 (2 months)
+b5dcd2a03e1b   nginx.2t5ex2.uncld.dev          Let's Encrypt             2026-12-31 (2 months)
+70707fd2fea2   test-staging.2t5ex2.uncld.dev   Let's Encrypt (staging)   2026-12-31 (2 months)
+c5f9301e1c06   uncloud.run                     Let's Encrypt             2026-12-31 (2 months)
 ```

website/docs/3-concepts/6-services/1-internal-dns.md

@@ -3,6 +3,7 @@
 Services can be addressed on the internal WireGuard network by service name, service ID, or a machine-scoped service name:
 
 ## Service name
+
 ```
 $ nslookup nats.internal
 Server:         127.0.0.11
@@ -30,6 +31,7 @@ Address: 10.210.1.4
 ```
 
 ## Service ID
+
 ```
 $ nslookup 3ecb3a8bbec5fd3f46efb056a934714a.internal
 Server:         127.0.0.11
@@ -40,6 +42,7 @@ Address: 10.210.0.4
 ```
 
 ## Machine ID scoped service name
+
 ```
 $ nslookup 0903f0ee483aa97d559eeeaac5e22283.m.nats.internal
 Server:         127.0.0.11
@@ -64,7 +67,8 @@ Address: 10.210.1.4
 
 Additionally, the IP ordering preference can be specified with a `rr` (round-robin) or `nearest` subdomain prefix.
 
-### `rr` (round-robin) *current default*
+### `rr` (round-robin) _current default_
+
 Randomly shuffled order on each lookup.
 
 ```
@@ -96,9 +100,11 @@ Address: 10.210.0.3
 ```
 
 ## Nearest scope
+
 Returns machine-local instances first.
 
 `machine-a`:
+
 ```
 $ nslookup nearest.worker.internal
 Server:         127.0.0.11
@@ -115,6 +121,7 @@ Address: 10.210.1.4
 ```
 
 `machine-b`:
+
 ```
 $ nslookup nearest.worker.internal
 Server:         127.0.0.11
@@ -131,3 +138,28 @@ Address: 10.210.0.4
 ```
 
 The prefixes can be used with service ID and machine-scoped service names, as well (e.g. `nearest.3ecb3a8bbec5fd3f46efb056a934714a.internal` or `rr.0903f0ee483aa97d559eeeaac5e22283.m.worker.internal`).
+
+## Machine name
+
+Machines can be addressed too on the internal WireGuard network, either by name of by ID, these names are
+available from the `m.internal` zone.
+
+```
+$ nslookup machine-1.m.internal
+Server:         127.0.0.11
+Address:        127.0.0.11#53
+
+Name:   machine-1.m.internal
+Address: 10.210.0.1
+```
+
+## Machine ID
+
+```
+$ nslookup c337f00600de51ef4375c9a9a267dba5.m.internal
+Server:         127.0.0.11
+Address:        127.0.0.11#53
+
+Name:   c337f00600de51ef4375c9a9a267dba5.m.internal
+Address: 10.210.0.1
+```

website/docs/9-cli-reference/uc_caddy_logs.md

@@ -21,13 +21,16 @@ uc caddy logs [flags]
   -m, --machine strings   Filter logs by machine name or ID. Can be specified multiple times or as a comma-separated list.
       --since string      Show logs generated on or after the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.
                           Examples:
-                            --since 2m30s                      Relative duration (2 minutes 30 seconds ago)
-                            --since 1h                         Relative duration (1 hour ago)
-                            --since 2025-11-24                 RFC 3339 date only (midnight using local timezone)
-                            --since 2024-05-14T22:50:00        RFC 3339 date/time using local timezone
+                            --since 1h45m                      Relative duration (1 hour 45 minutes ago)
+                                                               Supported units: d (day = 24h), h (hour), m (minute),
+                                                                                s (second), ms (millisecond),
+                                                                                us/µs (microsecond), ns (nanosecond)
+                            --since 2025-11-24                 RFC 3339 date only (midnight using client local timezone)
+                            --since 2024-05-14T22:50:00        RFC 3339 date/time using client local timezone
                             --since 2024-01-31T10:30:00Z       RFC 3339 date/time in UTC
                             --since 1763953966                 Unix timestamp (seconds since January 1, 1970)
-  -n, --tail string       Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs. (default "100")
+  -n, --tail string       Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs.
+                          Defaults to 100, or 'all' when --since is set.
       --until string      Show logs generated before the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.
                           See --since for examples.
       --utc               Print timestamps in UTC instead of local timezone.

website/docs/9-cli-reference/uc_logs.md

@@ -56,13 +56,16 @@ uc logs [SERVICE[/CONTAINER]...] [flags]
   -m, --machine strings   Filter logs by machine name or ID. Can be specified multiple times or as a comma-separated list.
       --since string      Show logs generated on or after the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.
                           Examples:
-                            --since 2m30s                      Relative duration (2 minutes 30 seconds ago)
-                            --since 1h                         Relative duration (1 hour ago)
-                            --since 2025-11-24                 RFC 3339 date only (midnight using local timezone)
-                            --since 2024-05-14T22:50:00        RFC 3339 date/time using local timezone
+                            --since 1h45m                      Relative duration (1 hour 45 minutes ago)
+                                                               Supported units: d (day = 24h), h (hour), m (minute),
+                                                                                s (second), ms (millisecond),
+                                                                                us/µs (microsecond), ns (nanosecond)
+                            --since 2025-11-24                 RFC 3339 date only (midnight using client local timezone)
+                            --since 2024-05-14T22:50:00        RFC 3339 date/time using client local timezone
                             --since 2024-01-31T10:30:00Z       RFC 3339 date/time in UTC
                             --since 1763953966                 Unix timestamp (seconds since January 1, 1970)
-  -n, --tail string       Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs. (default "100")
+  -n, --tail string       Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs.
+                          Defaults to 100, or 'all' when --since is set.
       --until string      Show logs generated before the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.
                           See --since for examples.
       --utc               Print timestamps in UTC instead of local timezone.

website/docs/9-cli-reference/uc_machine_logs.md

@@ -52,13 +52,16 @@ uc machine logs [SERVICE...] [flags]
   -m, --machine strings   Filter logs by machine name or ID. Can be specified multiple times or as a comma-separated list.
       --since string      Show logs generated on or after the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.
                           Examples:
-                            --since 2m30s                      Relative duration (2 minutes 30 seconds ago)
-                            --since 1h                         Relative duration (1 hour ago)
-                            --since 2025-11-24                 RFC 3339 date only (midnight using local timezone)
-                            --since 2024-05-14T22:50:00        RFC 3339 date/time using local timezone
+                            --since 1h45m                      Relative duration (1 hour 45 minutes ago)
+                                                               Supported units: d (day = 24h), h (hour), m (minute),
+                                                                                s (second), ms (millisecond),
+                                                                                us/µs (microsecond), ns (nanosecond)
+                            --since 2025-11-24                 RFC 3339 date only (midnight using client local timezone)
+                            --since 2024-05-14T22:50:00        RFC 3339 date/time using client local timezone
                             --since 2024-01-31T10:30:00Z       RFC 3339 date/time in UTC
                             --since 1763953966                 Unix timestamp (seconds since January 1, 1970)
-  -n, --tail string       Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs. (default "100")
+  -n, --tail string       Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs.
+                          Defaults to 100, or 'all' when --since is set.
       --until string      Show logs generated before the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.
                           See --since for examples.
       --utc               Print timestamps in UTC instead of local timezone.

website/docs/9-cli-reference/uc_service_logs.md

@@ -56,13 +56,16 @@ uc service logs [SERVICE[/CONTAINER]...] [flags]
   -m, --machine strings   Filter logs by machine name or ID. Can be specified multiple times or as a comma-separated list.
       --since string      Show logs generated on or after the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.
                           Examples:
-                            --since 2m30s                      Relative duration (2 minutes 30 seconds ago)
-                            --since 1h                         Relative duration (1 hour ago)
-                            --since 2025-11-24                 RFC 3339 date only (midnight using local timezone)
-                            --since 2024-05-14T22:50:00        RFC 3339 date/time using local timezone
+                            --since 1h45m                      Relative duration (1 hour 45 minutes ago)
+                                                               Supported units: d (day = 24h), h (hour), m (minute),
+                                                                                s (second), ms (millisecond),
+                                                                                us/µs (microsecond), ns (nanosecond)
+                            --since 2025-11-24                 RFC 3339 date only (midnight using client local timezone)
+                            --since 2024-05-14T22:50:00        RFC 3339 date/time using client local timezone
                             --since 2024-01-31T10:30:00Z       RFC 3339 date/time in UTC
                             --since 1763953966                 Unix timestamp (seconds since January 1, 1970)
-  -n, --tail string       Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs. (default "100")
+  -n, --tail string       Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs.
+                          Defaults to 100, or 'all' when --since is set.
       --until string      Show logs generated before the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.
                           See --since for examples.
       --utc               Print timestamps in UTC instead of local timezone.