feat: proxmox idlewatcher (#88)

* feat: idle sleep for proxmox LXCs

* refactor: replace deprecated docker api types

* chore(api): remove debug task list endpoint

* refactor: move servemux to gphttp/servemux; favicon.go to v1/favicon

* refactor: introduce Pool interface, move agent_pool to agent module

* refactor: simplify api code

* feat: introduce debug api

* refactor: remove net.URL and net.CIDR types, improved unmarshal handling

* chore: update Makefile for debug build tag, update README

* chore: add gperr.Unwrap method

* feat: relative time and duration formatting

* chore: add ROOT_DIR environment variable, refactor

* migration: move homepage override and icon cache to $BASE_DIR/data, add migration code

* fix: nil dereference on marshalling service health

* fix: wait for route deletion

* chore: enhance tasks debuggability

* feat: stdout access logger and MultiWriter

* fix(agent): remove agent properly on verify error

* fix(metrics): disk exclusion logic and added corresponding tests

* chore: update schema and prettify, fix package.json and Makefile

* fix: I/O buffer not being shrunk before putting back to pool

* feat: enhanced error handling module

* chore: deps upgrade

* feat: better value formatting and handling

---------

Co-authored-by: yusing <yusing@6uo.me>

internal/docker/client.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"maps"
 	"net"
 	"net/http"
 	"sync"
@@ -14,16 +15,15 @@ import (
 	"github.com/docker/docker/client"
 	"github.com/yusing/go-proxy/agent/pkg/agent"
 	"github.com/yusing/go-proxy/internal/common"
-	config "github.com/yusing/go-proxy/internal/config/types"
 	"github.com/yusing/go-proxy/internal/logging"
 	"github.com/yusing/go-proxy/internal/task"
+	"github.com/yusing/go-proxy/internal/utils/strutils"
 )
 
 type (
 	SharedClient struct {
 		*client.Client
 
-		key      string
 		refCount uint32
 		closedOn int64
 
@@ -66,7 +66,7 @@ func initClientCleaner() {
 		defer clientMapMu.Unlock()
 
 		for _, c := range clientMap {
-			delete(clientMap, c.key)
+			delete(clientMap, c.Key())
 			c.Client.Close()
 		}
 	})
@@ -80,30 +80,20 @@ func closeTimedOutClients() {
 
 	for _, c := range clientMap {
 		if atomic.LoadUint32(&c.refCount) == 0 && now-atomic.LoadInt64(&c.closedOn) > clientTTLSecs {
-			delete(clientMap, c.key)
+			delete(clientMap, c.Key())
 			c.Client.Close()
-			logging.Debug().Str("host", c.key).Msg("docker client closed")
+			logging.Debug().Str("host", c.DaemonHost()).Msg("docker client closed")
 		}
 	}
 }
 
-func (c *SharedClient) Address() string {
-	return c.addr
-}
+func Clients() map[string]*SharedClient {
+	clientMapMu.RLock()
+	defer clientMapMu.RUnlock()
 
-func (c *SharedClient) CheckConnection(ctx context.Context) error {
-	conn, err := c.dial(ctx)
-	if err != nil {
-		return err
-	}
-	conn.Close()
-	return nil
-}
-
-// if the client is still referenced, this is no-op.
-func (c *SharedClient) Close() {
-	atomic.StoreInt64(&c.closedOn, time.Now().Unix())
-	atomic.AddUint32(&c.refCount, ^uint32(0))
+	clients := make(map[string]*SharedClient, len(clientMap))
+	maps.Copy(clients, clientMap)
+	return clients
 }
 
 // NewClient creates a new Docker client connection to the specified host.
@@ -134,7 +124,7 @@ func NewClient(host string) (*SharedClient, error) {
 	var dial func(ctx context.Context) (net.Conn, error)
 
 	if agent.IsDockerHostAgent(host) {
-		cfg, ok := config.GetInstance().GetAgent(host)
+		cfg, ok := agent.Agents.Get(host)
 		if !ok {
 			panic(fmt.Errorf("agent %q not found", host))
 		}
@@ -187,7 +177,6 @@ func NewClient(host string) (*SharedClient, error) {
 
 	c := &SharedClient{
 		Client:   client,
-		key:      host,
 		refCount: 1,
 		addr:     addr,
 		dial:     dial,
@@ -197,9 +186,44 @@ func NewClient(host string) (*SharedClient, error) {
 	if c.dial == nil {
 		c.dial = client.Dialer()
 	}
+	if c.addr == "" {
+		c.addr = c.Client.DaemonHost()
+	}
 
 	defer logging.Debug().Str("host", host).Msg("docker client initialized")
 
-	clientMap[c.key] = c
+	clientMap[c.Key()] = c
 	return c, nil
 }
+
+func (c *SharedClient) Key() string {
+	return c.DaemonHost()
+}
+
+func (c *SharedClient) Address() string {
+	return c.addr
+}
+
+func (c *SharedClient) CheckConnection(ctx context.Context) error {
+	conn, err := c.dial(ctx)
+	if err != nil {
+		return err
+	}
+	conn.Close()
+	return nil
+}
+
+// if the client is still referenced, this is no-op.
+func (c *SharedClient) Close() {
+	atomic.StoreInt64(&c.closedOn, time.Now().Unix())
+	atomic.AddUint32(&c.refCount, ^uint32(0))
+}
+
+func (c *SharedClient) MarshalMap() map[string]any {
+	return map[string]any{
+		"host":      c.DaemonHost(),
+		"addr":      c.addr,
+		"ref_count": c.refCount,
+		"closed_on": strutils.FormatUnixTime(c.closedOn),
+	}
+}

internal/docker/container.go

@@ -8,16 +8,17 @@ import (
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/go-connections/nat"
 	"github.com/yusing/go-proxy/agent/pkg/agent"
-	config "github.com/yusing/go-proxy/internal/config/types"
+	"github.com/yusing/go-proxy/internal/gperr"
+	idlewatcher "github.com/yusing/go-proxy/internal/idlewatcher/types"
 	"github.com/yusing/go-proxy/internal/logging"
-	U "github.com/yusing/go-proxy/internal/utils"
+	"github.com/yusing/go-proxy/internal/utils"
 	"github.com/yusing/go-proxy/internal/utils/strutils"
 )
 
 type (
 	PortMapping = map[int]*container.Port
 	Container   struct {
-		_ U.NoCopy
+		_ utils.NoCopy
 
 		DockerHost    string          `json:"docker_host"`
 		Image         *ContainerImage `json:"image"`
@@ -26,7 +27,8 @@ type (
 
 		Agent *agent.AgentConfig `json:"agent"`
 
-		Labels map[string]string `json:"-"`
+		RouteConfig       map[string]string   `json:"route_config"`
+		IdlewatcherConfig *idlewatcher.Config `json:"idlewatcher_config"`
 
 		Mounts []string `json:"mounts"`
 
@@ -35,16 +37,10 @@ type (
 		PublicHostname     string      `json:"public_hostname"`
 		PrivateHostname    string      `json:"private_hostname"`
 
-		Aliases       []string `json:"aliases"`
-		IsExcluded    bool     `json:"is_excluded"`
-		IsExplicit    bool     `json:"is_explicit"`
-		IdleTimeout   string   `json:"idle_timeout,omitempty"`
-		WakeTimeout   string   `json:"wake_timeout,omitempty"`
-		StopMethod    string   `json:"stop_method,omitempty"`
-		StopTimeout   string   `json:"stop_timeout,omitempty"` // stop_method = "stop" only
-		StopSignal    string   `json:"stop_signal,omitempty"`  // stop_method = "stop" | "kill" only
-		StartEndpoint string   `json:"start_endpoint,omitempty"`
-		Running       bool     `json:"running"`
+		Aliases    []string `json:"aliases"`
+		IsExcluded bool     `json:"is_excluded"`
+		IsExplicit bool     `json:"is_explicit"`
+		Running    bool     `json:"running"`
 	}
 	ContainerImage struct {
 		Author string `json:"author,omitempty"`
@@ -69,21 +65,15 @@ func FromDocker(c *container.Summary, dockerHost string) (res *Container) {
 		PublicPortMapping:  helper.getPublicPortMapping(),
 		PrivatePortMapping: helper.getPrivatePortMapping(),
 
-		Aliases:       helper.getAliases(),
-		IsExcluded:    strutils.ParseBool(helper.getDeleteLabel(LabelExclude)),
-		IsExplicit:    isExplicit,
-		IdleTimeout:   helper.getDeleteLabel(LabelIdleTimeout),
-		WakeTimeout:   helper.getDeleteLabel(LabelWakeTimeout),
-		StopMethod:    helper.getDeleteLabel(LabelStopMethod),
-		StopTimeout:   helper.getDeleteLabel(LabelStopTimeout),
-		StopSignal:    helper.getDeleteLabel(LabelStopSignal),
-		StartEndpoint: helper.getDeleteLabel(LabelStartEndpoint),
-		Running:       c.Status == "running" || c.State == "running",
+		Aliases:    helper.getAliases(),
+		IsExcluded: strutils.ParseBool(helper.getDeleteLabel(LabelExclude)),
+		IsExplicit: isExplicit,
+		Running:    c.Status == "running" || c.State == "running",
 	}
 
 	if agent.IsDockerHostAgent(dockerHost) {
 		var ok bool
-		res.Agent, ok = config.GetInstance().GetAgent(dockerHost)
+		res.Agent, ok = agent.Agents.Get(dockerHost)
 		if !ok {
 			logging.Error().Msgf("agent %q not found", dockerHost)
 		}
@@ -91,6 +81,7 @@ func FromDocker(c *container.Summary, dockerHost string) (res *Container) {
 
 	res.setPrivateHostname(helper)
 	res.setPublicHostname()
+	res.loadDeleteIdlewatcherLabels(helper)
 
 	for lbl := range c.Labels {
 		if strings.HasPrefix(lbl, NSProxy+".") {
@@ -200,3 +191,31 @@ func (c *Container) setPrivateHostname(helper containerHelper) {
 		return
 	}
 }
+
+func (c *Container) loadDeleteIdlewatcherLabels(helper containerHelper) {
+	cfg := map[string]any{
+		"idle_timeout":   helper.getDeleteLabel(LabelIdleTimeout),
+		"wake_timeout":   helper.getDeleteLabel(LabelWakeTimeout),
+		"stop_method":    helper.getDeleteLabel(LabelStopMethod),
+		"stop_timeout":   helper.getDeleteLabel(LabelStopTimeout),
+		"stop_signal":    helper.getDeleteLabel(LabelStopSignal),
+		"start_endpoint": helper.getDeleteLabel(LabelStartEndpoint),
+	}
+	// set only if idlewatcher is enabled
+	idleTimeout := cfg["idle_timeout"]
+	if idleTimeout != "" {
+		idwCfg := &idlewatcher.Config{
+			Docker: &idlewatcher.DockerConfig{
+				DockerHost:    c.DockerHost,
+				ContainerID:   c.ContainerID,
+				ContainerName: c.ContainerName,
+			},
+		}
+		err := utils.MapUnmarshalValidate(cfg, idwCfg)
+		if err != nil {
+			gperr.LogWarn("invalid idlewatcher config", gperr.PrependSubject(c.ContainerName, err))
+		} else {
+			c.IdlewatcherConfig = idwCfg
+		}
+	}
+}

internal/docker/inspect.go

@@ -24,5 +24,5 @@ func (c *SharedClient) Inspect(containerID string) (*Container, error) {
 	if err != nil {
 		return nil, err
 	}
-	return FromInspectResponse(json, c.key), nil
+	return FromInspectResponse(json, c.DaemonHost()), nil
 }