Admin: misc improvements on admin server and workers. EC now works. (#7055)

* initial design

* added simulation as tests

* reorganized the codebase to move the simulation framework and tests into their own dedicated package

* integration test. ec worker task

* remove "enhanced" reference

* start master, volume servers, filer

Current Status
 Master: Healthy and running (port 9333)
 Filer: Healthy and running (port 8888)
 Volume Servers: All 6 servers running (ports 8080-8085)
🔄 Admin/Workers: Will start when dependencies are ready

* generate write load

* tasks are assigned

* admin start wtih grpc port. worker has its own working directory

* Update .gitignore

* working worker and admin. Task detection is not working yet.

* compiles, detection uses volumeSizeLimitMB from master

* compiles

* worker retries connecting to admin

* build and restart

* rendering pending tasks

* skip task ID column

* sticky worker id

* test canScheduleTaskNow

* worker reconnect to admin

* clean up logs

* worker register itself first

* worker can run ec work and report status

but:
1. one volume should not be repeatedly worked on.
2. ec shards needs to be distributed and source data should be deleted.

* move ec task logic

* listing ec shards

* local copy, ec. Need to distribute.

* ec is mostly working now

* distribution of ec shards needs improvement
* need configuration to enable ec

* show ec volumes

* interval field UI component

* rename

* integration test with vauuming

* garbage percentage threshold

* fix warning

* display ec shard sizes

* fix ec volumes list

* Update ui.go

* show default values

* ensure correct default value

* MaintenanceConfig use ConfigField

* use schema defined defaults

* config

* reduce duplication

* refactor to use BaseUIProvider

* each task register its schema

* checkECEncodingCandidate use ecDetector

* use vacuumDetector

* use volumeSizeLimitMB

* remove

remove

* remove unused

* refactor

* use new framework

* remove v2 reference

* refactor

* left menu can scroll now

* The maintenance manager was not being initialized when no data directory was configured for persistent storage.

* saving config

* Update task_config_schema_templ.go

* enable/disable tasks

* protobuf encoded task configurations

* fix system settings

* use ui component

* remove logs

* interface{} Reduction

* reduce interface{}

* reduce interface{}

* avoid from/to map

* reduce interface{}

* refactor

* keep it DRY

* added logging

* debug messages

* debug level

* debug

* show the log caller line

* use configured task policy

* log level

* handle admin heartbeat response

* Update worker.go

* fix EC rack and dc count

* Report task status to admin server

* fix task logging, simplify interface checking, use erasure_coding constants

* factor in empty volume server during task planning

* volume.list adds disk id

* track disk id also

* fix locking scheduled and manual scanning

* add active topology

* simplify task detector

* ec task completed, but shards are not showing up

* implement ec in ec_typed.go

* adjust log level

* dedup

* implementing ec copying shards and only ecx files

* use disk id when distributing ec shards

🎯 Planning: ActiveTopology creates DestinationPlan with specific TargetDisk
📦 Task Creation: maintenance_integration.go creates ECDestination with DiskId
🚀 Task Execution: EC task passes DiskId in VolumeEcShardsCopyRequest
💾 Volume Server: Receives disk_id and stores shards on specific disk (vs.store.Locations[req.DiskId])
📂 File System: EC shards and metadata land in the exact disk directory planned

* Delete original volume from all locations

* clean up existing shard locations

* local encoding and distributing

* Update docker/admin_integration/EC-TESTING-README.md

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* check volume id range

* simplify

* fix tests

* fix types

* clean up logs and tests

---------

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This commit is contained in:
Chris Lu
2025-07-30 12:38:03 -07:00
committed by GitHub
parent 64198dad83
commit 891a2fb6eb
130 changed files with 27737 additions and 4429 deletions

View File

@@ -0,0 +1,190 @@
package vacuum
import (
"fmt"
"github.com/seaweedfs/seaweedfs/weed/admin/config"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/pb/worker_pb"
"github.com/seaweedfs/seaweedfs/weed/worker/tasks/base"
)
// Config extends BaseConfig with vacuum-specific settings
type Config struct {
base.BaseConfig
GarbageThreshold float64 `json:"garbage_threshold"`
MinVolumeAgeSeconds int `json:"min_volume_age_seconds"`
MinIntervalSeconds int `json:"min_interval_seconds"`
}
// NewDefaultConfig creates a new default vacuum configuration
func NewDefaultConfig() *Config {
return &Config{
BaseConfig: base.BaseConfig{
Enabled: true,
ScanIntervalSeconds: 2 * 60 * 60, // 2 hours
MaxConcurrent: 2,
},
GarbageThreshold: 0.3, // 30%
MinVolumeAgeSeconds: 24 * 60 * 60, // 24 hours
MinIntervalSeconds: 7 * 24 * 60 * 60, // 7 days
}
}
// ToTaskPolicy converts configuration to a TaskPolicy protobuf message
func (c *Config) ToTaskPolicy() *worker_pb.TaskPolicy {
return &worker_pb.TaskPolicy{
Enabled: c.Enabled,
MaxConcurrent: int32(c.MaxConcurrent),
RepeatIntervalSeconds: int32(c.ScanIntervalSeconds),
CheckIntervalSeconds: int32(c.ScanIntervalSeconds),
TaskConfig: &worker_pb.TaskPolicy_VacuumConfig{
VacuumConfig: &worker_pb.VacuumTaskConfig{
GarbageThreshold: float64(c.GarbageThreshold),
MinVolumeAgeHours: int32(c.MinVolumeAgeSeconds / 3600), // Convert seconds to hours
MinIntervalSeconds: int32(c.MinIntervalSeconds),
},
},
}
}
// FromTaskPolicy loads configuration from a TaskPolicy protobuf message
func (c *Config) FromTaskPolicy(policy *worker_pb.TaskPolicy) error {
if policy == nil {
return fmt.Errorf("policy is nil")
}
// Set general TaskPolicy fields
c.Enabled = policy.Enabled
c.MaxConcurrent = int(policy.MaxConcurrent)
c.ScanIntervalSeconds = int(policy.RepeatIntervalSeconds) // Direct seconds-to-seconds mapping
// Set vacuum-specific fields from the task config
if vacuumConfig := policy.GetVacuumConfig(); vacuumConfig != nil {
c.GarbageThreshold = float64(vacuumConfig.GarbageThreshold)
c.MinVolumeAgeSeconds = int(vacuumConfig.MinVolumeAgeHours * 3600) // Convert hours to seconds
c.MinIntervalSeconds = int(vacuumConfig.MinIntervalSeconds)
}
return nil
}
// LoadConfigFromPersistence loads configuration from the persistence layer if available
func LoadConfigFromPersistence(configPersistence interface{}) *Config {
config := NewDefaultConfig()
// Try to load from persistence if available
if persistence, ok := configPersistence.(interface {
LoadVacuumTaskPolicy() (*worker_pb.TaskPolicy, error)
}); ok {
if policy, err := persistence.LoadVacuumTaskPolicy(); err == nil && policy != nil {
if err := config.FromTaskPolicy(policy); err == nil {
glog.V(1).Infof("Loaded vacuum configuration from persistence")
return config
}
}
}
glog.V(1).Infof("Using default vacuum configuration")
return config
}
// GetConfigSpec returns the configuration schema for vacuum tasks
func GetConfigSpec() base.ConfigSpec {
return base.ConfigSpec{
Fields: []*config.Field{
{
Name: "enabled",
JSONName: "enabled",
Type: config.FieldTypeBool,
DefaultValue: true,
Required: false,
DisplayName: "Enable Vacuum Tasks",
Description: "Whether vacuum tasks should be automatically created",
HelpText: "Toggle this to enable or disable automatic vacuum task generation",
InputType: "checkbox",
CSSClasses: "form-check-input",
},
{
Name: "scan_interval_seconds",
JSONName: "scan_interval_seconds",
Type: config.FieldTypeInterval,
DefaultValue: 2 * 60 * 60,
MinValue: 10 * 60,
MaxValue: 24 * 60 * 60,
Required: true,
DisplayName: "Scan Interval",
Description: "How often to scan for volumes needing vacuum",
HelpText: "The system will check for volumes that need vacuuming at this interval",
Placeholder: "2",
Unit: config.UnitHours,
InputType: "interval",
CSSClasses: "form-control",
},
{
Name: "max_concurrent",
JSONName: "max_concurrent",
Type: config.FieldTypeInt,
DefaultValue: 2,
MinValue: 1,
MaxValue: 10,
Required: true,
DisplayName: "Max Concurrent Tasks",
Description: "Maximum number of vacuum tasks that can run simultaneously",
HelpText: "Limits the number of vacuum operations running at the same time to control system load",
Placeholder: "2 (default)",
Unit: config.UnitCount,
InputType: "number",
CSSClasses: "form-control",
},
{
Name: "garbage_threshold",
JSONName: "garbage_threshold",
Type: config.FieldTypeFloat,
DefaultValue: 0.3,
MinValue: 0.0,
MaxValue: 1.0,
Required: true,
DisplayName: "Garbage Percentage Threshold",
Description: "Trigger vacuum when garbage ratio exceeds this percentage",
HelpText: "Volumes with more deleted content than this threshold will be vacuumed",
Placeholder: "0.30 (30%)",
Unit: config.UnitNone,
InputType: "number",
CSSClasses: "form-control",
},
{
Name: "min_volume_age_seconds",
JSONName: "min_volume_age_seconds",
Type: config.FieldTypeInterval,
DefaultValue: 24 * 60 * 60,
MinValue: 1 * 60 * 60,
MaxValue: 7 * 24 * 60 * 60,
Required: true,
DisplayName: "Minimum Volume Age",
Description: "Only vacuum volumes older than this duration",
HelpText: "Prevents vacuuming of recently created volumes that may still be actively written to",
Placeholder: "24",
Unit: config.UnitHours,
InputType: "interval",
CSSClasses: "form-control",
},
{
Name: "min_interval_seconds",
JSONName: "min_interval_seconds",
Type: config.FieldTypeInterval,
DefaultValue: 7 * 24 * 60 * 60,
MinValue: 1 * 24 * 60 * 60,
MaxValue: 30 * 24 * 60 * 60,
Required: true,
DisplayName: "Minimum Interval",
Description: "Minimum time between vacuum operations on the same volume",
HelpText: "Prevents excessive vacuuming of the same volume by enforcing a minimum wait time",
Placeholder: "7",
Unit: config.UnitDays,
InputType: "interval",
CSSClasses: "form-control",
},
},
}
}

View File

@@ -0,0 +1,112 @@
package vacuum
import (
"time"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/worker/tasks/base"
"github.com/seaweedfs/seaweedfs/weed/worker/types"
)
// Detection implements the detection logic for vacuum tasks
func Detection(metrics []*types.VolumeHealthMetrics, clusterInfo *types.ClusterInfo, config base.TaskConfig) ([]*types.TaskDetectionResult, error) {
if !config.IsEnabled() {
return nil, nil
}
vacuumConfig := config.(*Config)
var results []*types.TaskDetectionResult
minVolumeAge := time.Duration(vacuumConfig.MinVolumeAgeSeconds) * time.Second
debugCount := 0
skippedDueToGarbage := 0
skippedDueToAge := 0
for _, metric := range metrics {
// Check if volume needs vacuum
if metric.GarbageRatio >= vacuumConfig.GarbageThreshold && metric.Age >= minVolumeAge {
priority := types.TaskPriorityNormal
if metric.GarbageRatio > 0.6 {
priority = types.TaskPriorityHigh
}
result := &types.TaskDetectionResult{
TaskType: types.TaskTypeVacuum,
VolumeID: metric.VolumeID,
Server: metric.Server,
Collection: metric.Collection,
Priority: priority,
Reason: "Volume has excessive garbage requiring vacuum",
ScheduleAt: time.Now(),
}
results = append(results, result)
} else {
// Debug why volume was not selected
if debugCount < 5 { // Limit debug output to first 5 volumes
if metric.GarbageRatio < vacuumConfig.GarbageThreshold {
skippedDueToGarbage++
}
if metric.Age < minVolumeAge {
skippedDueToAge++
}
}
debugCount++
}
}
// Log debug summary if no tasks were created
if len(results) == 0 && len(metrics) > 0 {
totalVolumes := len(metrics)
glog.Infof("VACUUM: No tasks created for %d volumes. Threshold=%.2f%%, MinAge=%s. Skipped: %d (garbage<threshold), %d (age<minimum)",
totalVolumes, vacuumConfig.GarbageThreshold*100, minVolumeAge, skippedDueToGarbage, skippedDueToAge)
// Show details for first few volumes
for i, metric := range metrics {
if i >= 3 { // Limit to first 3 volumes
break
}
glog.Infof("VACUUM: Volume %d: garbage=%.2f%% (need ≥%.2f%%), age=%s (need ≥%s)",
metric.VolumeID, metric.GarbageRatio*100, vacuumConfig.GarbageThreshold*100,
metric.Age.Truncate(time.Minute), minVolumeAge.Truncate(time.Minute))
}
}
return results, nil
}
// Scheduling implements the scheduling logic for vacuum tasks
func Scheduling(task *types.Task, runningTasks []*types.Task, availableWorkers []*types.Worker, config base.TaskConfig) bool {
vacuumConfig := config.(*Config)
// Count running vacuum tasks
runningVacuumCount := 0
for _, runningTask := range runningTasks {
if runningTask.Type == types.TaskTypeVacuum {
runningVacuumCount++
}
}
// Check concurrency limit
if runningVacuumCount >= vacuumConfig.MaxConcurrent {
return false
}
// Check for available workers with vacuum capability
for _, worker := range availableWorkers {
if worker.CurrentLoad < worker.MaxConcurrent {
for _, capability := range worker.Capabilities {
if capability == types.TaskTypeVacuum {
return true
}
}
}
}
return false
}
// CreateTask creates a new vacuum task instance
func CreateTask(params types.TaskParams) (types.TaskInterface, error) {
// Create and return the vacuum task using existing Task type
return NewTask(params.Server, params.VolumeID), nil
}

View File

@@ -1,314 +0,0 @@
package vacuum
import (
"fmt"
"html/template"
"strconv"
"time"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/worker/types"
)
// UIProvider provides the UI for vacuum task configuration
type UIProvider struct {
detector *VacuumDetector
scheduler *VacuumScheduler
}
// NewUIProvider creates a new vacuum UI provider
func NewUIProvider(detector *VacuumDetector, scheduler *VacuumScheduler) *UIProvider {
return &UIProvider{
detector: detector,
scheduler: scheduler,
}
}
// GetTaskType returns the task type
func (ui *UIProvider) GetTaskType() types.TaskType {
return types.TaskTypeVacuum
}
// GetDisplayName returns the human-readable name
func (ui *UIProvider) GetDisplayName() string {
return "Volume Vacuum"
}
// GetDescription returns a description of what this task does
func (ui *UIProvider) GetDescription() string {
return "Reclaims disk space by removing deleted files from volumes"
}
// GetIcon returns the icon CSS class for this task type
func (ui *UIProvider) GetIcon() string {
return "fas fa-broom text-primary"
}
// VacuumConfig represents the vacuum configuration
type VacuumConfig struct {
Enabled bool `json:"enabled"`
GarbageThreshold float64 `json:"garbage_threshold"`
ScanIntervalSeconds int `json:"scan_interval_seconds"`
MaxConcurrent int `json:"max_concurrent"`
MinVolumeAgeSeconds int `json:"min_volume_age_seconds"`
MinIntervalSeconds int `json:"min_interval_seconds"`
}
// Helper functions for duration conversion
func secondsToDuration(seconds int) time.Duration {
return time.Duration(seconds) * time.Second
}
func durationToSeconds(d time.Duration) int {
return int(d.Seconds())
}
// formatDurationForUser formats seconds as a user-friendly duration string
func formatDurationForUser(seconds int) string {
d := secondsToDuration(seconds)
if d < time.Minute {
return fmt.Sprintf("%ds", seconds)
}
if d < time.Hour {
return fmt.Sprintf("%.0fm", d.Minutes())
}
if d < 24*time.Hour {
return fmt.Sprintf("%.1fh", d.Hours())
}
return fmt.Sprintf("%.1fd", d.Hours()/24)
}
// RenderConfigForm renders the configuration form HTML
func (ui *UIProvider) RenderConfigForm(currentConfig interface{}) (template.HTML, error) {
config := ui.getCurrentVacuumConfig()
// Build form using the FormBuilder helper
form := types.NewFormBuilder()
// Detection Settings
form.AddCheckboxField(
"enabled",
"Enable Vacuum Tasks",
"Whether vacuum tasks should be automatically created",
config.Enabled,
)
form.AddNumberField(
"garbage_threshold",
"Garbage Threshold (%)",
"Trigger vacuum when garbage ratio exceeds this percentage (0.0-1.0)",
config.GarbageThreshold,
true,
)
form.AddDurationField(
"scan_interval",
"Scan Interval",
"How often to scan for volumes needing vacuum",
secondsToDuration(config.ScanIntervalSeconds),
true,
)
form.AddDurationField(
"min_volume_age",
"Minimum Volume Age",
"Only vacuum volumes older than this duration",
secondsToDuration(config.MinVolumeAgeSeconds),
true,
)
// Scheduling Settings
form.AddNumberField(
"max_concurrent",
"Max Concurrent Tasks",
"Maximum number of vacuum tasks that can run simultaneously",
float64(config.MaxConcurrent),
true,
)
form.AddDurationField(
"min_interval",
"Minimum Interval",
"Minimum time between vacuum operations on the same volume",
secondsToDuration(config.MinIntervalSeconds),
true,
)
// Generate organized form sections using Bootstrap components
html := `
<div class="row">
<div class="col-12">
<div class="card mb-4">
<div class="card-header">
<h5 class="mb-0">
<i class="fas fa-search me-2"></i>
Detection Settings
</h5>
</div>
<div class="card-body">
` + string(form.Build()) + `
</div>
</div>
</div>
</div>
<script>
function resetForm() {
if (confirm('Reset all vacuum settings to defaults?')) {
// Reset to default values
document.querySelector('input[name="enabled"]').checked = true;
document.querySelector('input[name="garbage_threshold"]').value = '0.3';
document.querySelector('input[name="scan_interval"]').value = '30m';
document.querySelector('input[name="min_volume_age"]').value = '1h';
document.querySelector('input[name="max_concurrent"]').value = '2';
document.querySelector('input[name="min_interval"]').value = '6h';
}
}
</script>
`
return template.HTML(html), nil
}
// ParseConfigForm parses form data into configuration
func (ui *UIProvider) ParseConfigForm(formData map[string][]string) (interface{}, error) {
config := &VacuumConfig{}
// Parse enabled checkbox
config.Enabled = len(formData["enabled"]) > 0 && formData["enabled"][0] == "on"
// Parse garbage threshold
if thresholdStr := formData["garbage_threshold"]; len(thresholdStr) > 0 {
if threshold, err := strconv.ParseFloat(thresholdStr[0], 64); err != nil {
return nil, fmt.Errorf("invalid garbage threshold: %w", err)
} else if threshold < 0 || threshold > 1 {
return nil, fmt.Errorf("garbage threshold must be between 0.0 and 1.0")
} else {
config.GarbageThreshold = threshold
}
}
// Parse scan interval
if intervalStr := formData["scan_interval"]; len(intervalStr) > 0 {
if interval, err := time.ParseDuration(intervalStr[0]); err != nil {
return nil, fmt.Errorf("invalid scan interval: %w", err)
} else {
config.ScanIntervalSeconds = durationToSeconds(interval)
}
}
// Parse min volume age
if ageStr := formData["min_volume_age"]; len(ageStr) > 0 {
if age, err := time.ParseDuration(ageStr[0]); err != nil {
return nil, fmt.Errorf("invalid min volume age: %w", err)
} else {
config.MinVolumeAgeSeconds = durationToSeconds(age)
}
}
// Parse max concurrent
if concurrentStr := formData["max_concurrent"]; len(concurrentStr) > 0 {
if concurrent, err := strconv.Atoi(concurrentStr[0]); err != nil {
return nil, fmt.Errorf("invalid max concurrent: %w", err)
} else if concurrent < 1 {
return nil, fmt.Errorf("max concurrent must be at least 1")
} else {
config.MaxConcurrent = concurrent
}
}
// Parse min interval
if intervalStr := formData["min_interval"]; len(intervalStr) > 0 {
if interval, err := time.ParseDuration(intervalStr[0]); err != nil {
return nil, fmt.Errorf("invalid min interval: %w", err)
} else {
config.MinIntervalSeconds = durationToSeconds(interval)
}
}
return config, nil
}
// GetCurrentConfig returns the current configuration
func (ui *UIProvider) GetCurrentConfig() interface{} {
return ui.getCurrentVacuumConfig()
}
// ApplyConfig applies the new configuration
func (ui *UIProvider) ApplyConfig(config interface{}) error {
vacuumConfig, ok := config.(*VacuumConfig)
if !ok {
return fmt.Errorf("invalid config type, expected *VacuumConfig")
}
// Apply to detector
if ui.detector != nil {
ui.detector.SetEnabled(vacuumConfig.Enabled)
ui.detector.SetGarbageThreshold(vacuumConfig.GarbageThreshold)
ui.detector.SetScanInterval(secondsToDuration(vacuumConfig.ScanIntervalSeconds))
ui.detector.SetMinVolumeAge(secondsToDuration(vacuumConfig.MinVolumeAgeSeconds))
}
// Apply to scheduler
if ui.scheduler != nil {
ui.scheduler.SetEnabled(vacuumConfig.Enabled)
ui.scheduler.SetMaxConcurrent(vacuumConfig.MaxConcurrent)
ui.scheduler.SetMinInterval(secondsToDuration(vacuumConfig.MinIntervalSeconds))
}
glog.V(1).Infof("Applied vacuum configuration: enabled=%v, threshold=%.1f%%, scan_interval=%s, max_concurrent=%d",
vacuumConfig.Enabled, vacuumConfig.GarbageThreshold*100, formatDurationForUser(vacuumConfig.ScanIntervalSeconds), vacuumConfig.MaxConcurrent)
return nil
}
// getCurrentVacuumConfig gets the current configuration from detector and scheduler
func (ui *UIProvider) getCurrentVacuumConfig() *VacuumConfig {
config := &VacuumConfig{
// Default values (fallback if detectors/schedulers are nil)
Enabled: true,
GarbageThreshold: 0.3,
ScanIntervalSeconds: 30 * 60,
MinVolumeAgeSeconds: 1 * 60 * 60,
MaxConcurrent: 2,
MinIntervalSeconds: 6 * 60 * 60,
}
// Get current values from detector
if ui.detector != nil {
config.Enabled = ui.detector.IsEnabled()
config.GarbageThreshold = ui.detector.GetGarbageThreshold()
config.ScanIntervalSeconds = durationToSeconds(ui.detector.ScanInterval())
config.MinVolumeAgeSeconds = durationToSeconds(ui.detector.GetMinVolumeAge())
}
// Get current values from scheduler
if ui.scheduler != nil {
config.MaxConcurrent = ui.scheduler.GetMaxConcurrent()
config.MinIntervalSeconds = durationToSeconds(ui.scheduler.GetMinInterval())
}
return config
}
// RegisterUI registers the vacuum UI provider with the UI registry
func RegisterUI(uiRegistry *types.UIRegistry, detector *VacuumDetector, scheduler *VacuumScheduler) {
uiProvider := NewUIProvider(detector, scheduler)
uiRegistry.RegisterUI(uiProvider)
glog.V(1).Infof("✅ Registered vacuum task UI provider")
}
// Example: How to get the UI provider for external use
func GetUIProvider(uiRegistry *types.UIRegistry) *UIProvider {
provider := uiRegistry.GetProvider(types.TaskTypeVacuum)
if provider == nil {
return nil
}
if vacuumProvider, ok := provider.(*UIProvider); ok {
return vacuumProvider
}
return nil
}

View File

@@ -1,60 +1,184 @@
package vacuum
import (
"context"
"fmt"
"io"
"time"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/pb"
"github.com/seaweedfs/seaweedfs/weed/pb/volume_server_pb"
"github.com/seaweedfs/seaweedfs/weed/worker/tasks"
"github.com/seaweedfs/seaweedfs/weed/worker/types"
"google.golang.org/grpc"
"google.golang.org/grpc/credentials/insecure"
)
// Task implements vacuum operation to reclaim disk space
type Task struct {
*tasks.BaseTask
server string
volumeID uint32
server string
volumeID uint32
garbageThreshold float64
}
// NewTask creates a new vacuum task instance
func NewTask(server string, volumeID uint32) *Task {
task := &Task{
BaseTask: tasks.NewBaseTask(types.TaskTypeVacuum),
server: server,
volumeID: volumeID,
BaseTask: tasks.NewBaseTask(types.TaskTypeVacuum),
server: server,
volumeID: volumeID,
garbageThreshold: 0.3, // Default 30% threshold
}
return task
}
// Execute executes the vacuum task
// Execute performs the vacuum operation
func (t *Task) Execute(params types.TaskParams) error {
glog.Infof("Starting vacuum task for volume %d on server %s", t.volumeID, t.server)
// Use BaseTask.ExecuteTask to handle logging initialization
return t.ExecuteTask(context.Background(), params, t.executeImpl)
}
// Simulate vacuum operation with progress updates
steps := []struct {
name string
duration time.Duration
progress float64
}{
{"Scanning volume", 1 * time.Second, 20},
{"Identifying deleted files", 2 * time.Second, 50},
{"Compacting data", 3 * time.Second, 80},
{"Finalizing vacuum", 1 * time.Second, 100},
// executeImpl is the actual vacuum implementation
func (t *Task) executeImpl(ctx context.Context, params types.TaskParams) error {
t.LogInfo("Starting vacuum for volume %d on server %s", t.volumeID, t.server)
// Parse garbage threshold from typed parameters
if params.TypedParams != nil {
if vacuumParams := params.TypedParams.GetVacuumParams(); vacuumParams != nil {
t.garbageThreshold = vacuumParams.GarbageThreshold
t.LogWithFields("INFO", "Using garbage threshold from parameters", map[string]interface{}{
"threshold": t.garbageThreshold,
})
}
}
for _, step := range steps {
if t.IsCancelled() {
return fmt.Errorf("vacuum task cancelled")
// Convert server address to gRPC address and use proper dial option
grpcAddress := pb.ServerToGrpcAddress(t.server)
var dialOpt grpc.DialOption = grpc.WithTransportCredentials(insecure.NewCredentials())
if params.GrpcDialOption != nil {
dialOpt = params.GrpcDialOption
}
conn, err := grpc.NewClient(grpcAddress, dialOpt)
if err != nil {
t.LogError("Failed to connect to volume server %s: %v", t.server, err)
return fmt.Errorf("failed to connect to volume server %s: %v", t.server, err)
}
defer conn.Close()
client := volume_server_pb.NewVolumeServerClient(conn)
// Step 1: Check vacuum eligibility
t.SetProgress(10.0)
t.LogDebug("Checking vacuum eligibility for volume %d", t.volumeID)
checkResp, err := client.VacuumVolumeCheck(ctx, &volume_server_pb.VacuumVolumeCheckRequest{
VolumeId: t.volumeID,
})
if err != nil {
t.LogError("Vacuum check failed for volume %d: %v", t.volumeID, err)
return fmt.Errorf("vacuum check failed for volume %d: %v", t.volumeID, err)
}
// Check if garbage ratio meets threshold
if checkResp.GarbageRatio < t.garbageThreshold {
t.LogWarning("Volume %d garbage ratio %.2f%% is below threshold %.2f%%, skipping vacuum",
t.volumeID, checkResp.GarbageRatio*100, t.garbageThreshold*100)
return fmt.Errorf("volume %d garbage ratio %.2f%% is below threshold %.2f%%, skipping vacuum",
t.volumeID, checkResp.GarbageRatio*100, t.garbageThreshold*100)
}
t.LogWithFields("INFO", "Volume eligible for vacuum", map[string]interface{}{
"volume_id": t.volumeID,
"garbage_ratio": checkResp.GarbageRatio,
"threshold": t.garbageThreshold,
"garbage_percent": checkResp.GarbageRatio * 100,
})
// Step 2: Compact volume
t.SetProgress(30.0)
t.LogInfo("Starting compact for volume %d", t.volumeID)
compactStream, err := client.VacuumVolumeCompact(ctx, &volume_server_pb.VacuumVolumeCompactRequest{
VolumeId: t.volumeID,
})
if err != nil {
t.LogError("Vacuum compact failed for volume %d: %v", t.volumeID, err)
return fmt.Errorf("vacuum compact failed for volume %d: %v", t.volumeID, err)
}
// Process compact stream and track progress
var processedBytes int64
var totalBytes int64
for {
resp, err := compactStream.Recv()
if err != nil {
if err == io.EOF {
break
}
t.LogError("Vacuum compact stream error for volume %d: %v", t.volumeID, err)
return fmt.Errorf("vacuum compact stream error for volume %d: %v", t.volumeID, err)
}
glog.V(1).Infof("Vacuum task step: %s", step.name)
t.SetProgress(step.progress)
processedBytes = resp.ProcessedBytes
if resp.LoadAvg_1M > 0 {
totalBytes = int64(resp.LoadAvg_1M) // This is a rough approximation
}
// Simulate work
time.Sleep(step.duration)
// Update progress based on processed bytes (30% to 70% of total progress)
if totalBytes > 0 {
compactProgress := float64(processedBytes) / float64(totalBytes)
if compactProgress > 1.0 {
compactProgress = 1.0
}
progress := 30.0 + (compactProgress * 40.0) // 30% to 70%
t.SetProgress(progress)
}
t.LogWithFields("DEBUG", "Volume compact progress", map[string]interface{}{
"volume_id": t.volumeID,
"processed_bytes": processedBytes,
"total_bytes": totalBytes,
"compact_progress": fmt.Sprintf("%.1f%%", (float64(processedBytes)/float64(totalBytes))*100),
})
}
glog.Infof("Vacuum task completed for volume %d on server %s", t.volumeID, t.server)
// Step 3: Commit vacuum changes
t.SetProgress(80.0)
t.LogInfo("Committing vacuum for volume %d", t.volumeID)
commitResp, err := client.VacuumVolumeCommit(ctx, &volume_server_pb.VacuumVolumeCommitRequest{
VolumeId: t.volumeID,
})
if err != nil {
t.LogError("Vacuum commit failed for volume %d: %v", t.volumeID, err)
return fmt.Errorf("vacuum commit failed for volume %d: %v", t.volumeID, err)
}
// Step 4: Cleanup temporary files
t.SetProgress(90.0)
t.LogInfo("Cleaning up vacuum files for volume %d", t.volumeID)
_, err = client.VacuumVolumeCleanup(ctx, &volume_server_pb.VacuumVolumeCleanupRequest{
VolumeId: t.volumeID,
})
if err != nil {
// Log warning but don't fail the task
t.LogWarning("Vacuum cleanup warning for volume %d: %v", t.volumeID, err)
}
t.SetProgress(100.0)
newVolumeSize := commitResp.VolumeSize
t.LogWithFields("INFO", "Successfully completed vacuum", map[string]interface{}{
"volume_id": t.volumeID,
"server": t.server,
"new_volume_size": newVolumeSize,
"garbage_reclaimed": true,
})
return nil
}
@@ -71,9 +195,20 @@ func (t *Task) Validate(params types.TaskParams) error {
// EstimateTime estimates the time needed for the task
func (t *Task) EstimateTime(params types.TaskParams) time.Duration {
// Base time for vacuum operation
baseTime := 25 * time.Second
// Base time for vacuum operations - varies by volume size and garbage ratio
// Typically vacuum is faster than EC encoding
baseTime := 5 * time.Minute
// Could adjust based on volume size or usage patterns
// Use default estimation since volume size is not available in typed params
return baseTime
}
// GetProgress returns the current progress
func (t *Task) GetProgress() float64 {
return t.BaseTask.GetProgress()
}
// Cancel cancels the task
func (t *Task) Cancel() error {
return t.BaseTask.Cancel()
}

View File

@@ -1,132 +0,0 @@
package vacuum
import (
"time"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/worker/types"
)
// VacuumDetector implements vacuum task detection using code instead of schemas
type VacuumDetector struct {
enabled bool
garbageThreshold float64
minVolumeAge time.Duration
scanInterval time.Duration
}
// Compile-time interface assertions
var (
_ types.TaskDetector = (*VacuumDetector)(nil)
_ types.PolicyConfigurableDetector = (*VacuumDetector)(nil)
)
// NewVacuumDetector creates a new simple vacuum detector
func NewVacuumDetector() *VacuumDetector {
return &VacuumDetector{
enabled: true,
garbageThreshold: 0.3,
minVolumeAge: 24 * time.Hour,
scanInterval: 30 * time.Minute,
}
}
// GetTaskType returns the task type
func (d *VacuumDetector) GetTaskType() types.TaskType {
return types.TaskTypeVacuum
}
// ScanForTasks scans for volumes that need vacuum operations
func (d *VacuumDetector) ScanForTasks(volumeMetrics []*types.VolumeHealthMetrics, clusterInfo *types.ClusterInfo) ([]*types.TaskDetectionResult, error) {
if !d.enabled {
return nil, nil
}
var results []*types.TaskDetectionResult
for _, metric := range volumeMetrics {
// Check if volume needs vacuum
if metric.GarbageRatio >= d.garbageThreshold && metric.Age >= d.minVolumeAge {
// Higher priority for volumes with more garbage
priority := types.TaskPriorityNormal
if metric.GarbageRatio > 0.6 {
priority = types.TaskPriorityHigh
}
result := &types.TaskDetectionResult{
TaskType: types.TaskTypeVacuum,
VolumeID: metric.VolumeID,
Server: metric.Server,
Collection: metric.Collection,
Priority: priority,
Reason: "Volume has excessive garbage requiring vacuum",
Parameters: map[string]interface{}{
"garbage_ratio": metric.GarbageRatio,
"volume_age": metric.Age.String(),
},
ScheduleAt: time.Now(),
}
results = append(results, result)
}
}
glog.V(2).Infof("Vacuum detector found %d volumes needing vacuum", len(results))
return results, nil
}
// ScanInterval returns how often this detector should scan
func (d *VacuumDetector) ScanInterval() time.Duration {
return d.scanInterval
}
// IsEnabled returns whether this detector is enabled
func (d *VacuumDetector) IsEnabled() bool {
return d.enabled
}
// Configuration setters
func (d *VacuumDetector) SetEnabled(enabled bool) {
d.enabled = enabled
}
func (d *VacuumDetector) SetGarbageThreshold(threshold float64) {
d.garbageThreshold = threshold
}
func (d *VacuumDetector) SetScanInterval(interval time.Duration) {
d.scanInterval = interval
}
func (d *VacuumDetector) SetMinVolumeAge(age time.Duration) {
d.minVolumeAge = age
}
// GetGarbageThreshold returns the current garbage threshold
func (d *VacuumDetector) GetGarbageThreshold() float64 {
return d.garbageThreshold
}
// GetMinVolumeAge returns the minimum volume age
func (d *VacuumDetector) GetMinVolumeAge() time.Duration {
return d.minVolumeAge
}
// GetScanInterval returns the scan interval
func (d *VacuumDetector) GetScanInterval() time.Duration {
return d.scanInterval
}
// ConfigureFromPolicy configures the detector based on the maintenance policy
func (d *VacuumDetector) ConfigureFromPolicy(policy interface{}) {
// Type assert to the maintenance policy type we expect
if maintenancePolicy, ok := policy.(interface {
GetVacuumEnabled() bool
GetVacuumGarbageRatio() float64
}); ok {
d.SetEnabled(maintenancePolicy.GetVacuumEnabled())
d.SetGarbageThreshold(maintenancePolicy.GetVacuumGarbageRatio())
} else {
glog.V(1).Infof("Could not configure vacuum detector from policy: unsupported policy type")
}
}

View File

@@ -2,80 +2,71 @@ package vacuum
import (
"fmt"
"time"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/worker/tasks"
"github.com/seaweedfs/seaweedfs/weed/worker/tasks/base"
"github.com/seaweedfs/seaweedfs/weed/worker/types"
)
// Factory creates vacuum task instances
type Factory struct {
*tasks.BaseTaskFactory
}
// NewFactory creates a new vacuum task factory
func NewFactory() *Factory {
return &Factory{
BaseTaskFactory: tasks.NewBaseTaskFactory(
types.TaskTypeVacuum,
[]string{"vacuum", "storage"},
"Vacuum operation to reclaim disk space by removing deleted files",
),
}
}
// Create creates a new vacuum task instance
func (f *Factory) Create(params types.TaskParams) (types.TaskInterface, error) {
// Validate parameters
if params.VolumeID == 0 {
return nil, fmt.Errorf("volume_id is required")
}
if params.Server == "" {
return nil, fmt.Errorf("server is required")
}
task := NewTask(params.Server, params.VolumeID)
task.SetEstimatedDuration(task.EstimateTime(params))
return task, nil
}
// Shared detector and scheduler instances
var (
sharedDetector *VacuumDetector
sharedScheduler *VacuumScheduler
)
// getSharedInstances returns the shared detector and scheduler instances
func getSharedInstances() (*VacuumDetector, *VacuumScheduler) {
if sharedDetector == nil {
sharedDetector = NewVacuumDetector()
}
if sharedScheduler == nil {
sharedScheduler = NewVacuumScheduler()
}
return sharedDetector, sharedScheduler
}
// GetSharedInstances returns the shared detector and scheduler instances (public access)
func GetSharedInstances() (*VacuumDetector, *VacuumScheduler) {
return getSharedInstances()
}
// Global variable to hold the task definition for configuration updates
var globalTaskDef *base.TaskDefinition
// Auto-register this task when the package is imported
func init() {
factory := NewFactory()
tasks.AutoRegister(types.TaskTypeVacuum, factory)
RegisterVacuumTask()
// Get shared instances for all registrations
detector, scheduler := getSharedInstances()
// Register with types registry
tasks.AutoRegisterTypes(func(registry *types.TaskRegistry) {
registry.RegisterTask(detector, scheduler)
})
// Register with UI registry using the same instances
tasks.AutoRegisterUI(func(uiRegistry *types.UIRegistry) {
RegisterUI(uiRegistry, detector, scheduler)
})
// Register config updater
tasks.AutoRegisterConfigUpdater(types.TaskTypeVacuum, UpdateConfigFromPersistence)
}
// RegisterVacuumTask registers the vacuum task with the new architecture
func RegisterVacuumTask() {
// Create configuration instance
config := NewDefaultConfig()
// Create complete task definition
taskDef := &base.TaskDefinition{
Type: types.TaskTypeVacuum,
Name: "vacuum",
DisplayName: "Volume Vacuum",
Description: "Reclaims disk space by removing deleted files from volumes",
Icon: "fas fa-broom text-primary",
Capabilities: []string{"vacuum", "storage"},
Config: config,
ConfigSpec: GetConfigSpec(),
CreateTask: CreateTask,
DetectionFunc: Detection,
ScanInterval: 2 * time.Hour,
SchedulingFunc: Scheduling,
MaxConcurrent: 2,
RepeatInterval: 7 * 24 * time.Hour,
}
// Store task definition globally for configuration updates
globalTaskDef = taskDef
// Register everything with a single function call!
base.RegisterTask(taskDef)
}
// UpdateConfigFromPersistence updates the vacuum configuration from persistence
func UpdateConfigFromPersistence(configPersistence interface{}) error {
if globalTaskDef == nil {
return fmt.Errorf("vacuum task not registered")
}
// Load configuration from persistence
newConfig := LoadConfigFromPersistence(configPersistence)
if newConfig == nil {
return fmt.Errorf("failed to load configuration from persistence")
}
// Update the task definition's config
globalTaskDef.Config = newConfig
glog.V(1).Infof("Updated vacuum task configuration from persistence")
return nil
}

View File

@@ -1,111 +0,0 @@
package vacuum
import (
"time"
"github.com/seaweedfs/seaweedfs/weed/worker/types"
)
// VacuumScheduler implements vacuum task scheduling using code instead of schemas
type VacuumScheduler struct {
enabled bool
maxConcurrent int
minInterval time.Duration
}
// Compile-time interface assertions
var (
_ types.TaskScheduler = (*VacuumScheduler)(nil)
)
// NewVacuumScheduler creates a new simple vacuum scheduler
func NewVacuumScheduler() *VacuumScheduler {
return &VacuumScheduler{
enabled: true,
maxConcurrent: 2,
minInterval: 6 * time.Hour,
}
}
// GetTaskType returns the task type
func (s *VacuumScheduler) GetTaskType() types.TaskType {
return types.TaskTypeVacuum
}
// CanScheduleNow determines if a vacuum task can be scheduled right now
func (s *VacuumScheduler) CanScheduleNow(task *types.Task, runningTasks []*types.Task, availableWorkers []*types.Worker) bool {
// Check if scheduler is enabled
if !s.enabled {
return false
}
// Check concurrent limit
runningVacuumCount := 0
for _, runningTask := range runningTasks {
if runningTask.Type == types.TaskTypeVacuum {
runningVacuumCount++
}
}
if runningVacuumCount >= s.maxConcurrent {
return false
}
// Check if there's an available worker with vacuum capability
for _, worker := range availableWorkers {
if worker.CurrentLoad < worker.MaxConcurrent {
for _, capability := range worker.Capabilities {
if capability == types.TaskTypeVacuum {
return true
}
}
}
}
return false
}
// GetPriority returns the priority for this task
func (s *VacuumScheduler) GetPriority(task *types.Task) types.TaskPriority {
// Could adjust priority based on task parameters
if params, ok := task.Parameters["garbage_ratio"].(float64); ok {
if params > 0.8 {
return types.TaskPriorityHigh
}
}
return task.Priority
}
// GetMaxConcurrent returns max concurrent tasks of this type
func (s *VacuumScheduler) GetMaxConcurrent() int {
return s.maxConcurrent
}
// GetDefaultRepeatInterval returns the default interval to wait before repeating vacuum tasks
func (s *VacuumScheduler) GetDefaultRepeatInterval() time.Duration {
return s.minInterval
}
// IsEnabled returns whether this scheduler is enabled
func (s *VacuumScheduler) IsEnabled() bool {
return s.enabled
}
// Configuration setters
func (s *VacuumScheduler) SetEnabled(enabled bool) {
s.enabled = enabled
}
func (s *VacuumScheduler) SetMaxConcurrent(max int) {
s.maxConcurrent = max
}
func (s *VacuumScheduler) SetMinInterval(interval time.Duration) {
s.minInterval = interval
}
// GetMinInterval returns the minimum interval
func (s *VacuumScheduler) GetMinInterval() time.Duration {
return s.minInterval
}