perf(monitor): improve monitor query and chart performance for large ranges (#13874)

* perf(monitor): improve monitor query and chart performance for large ranges

- enable LTTB sampling in the shared line chart to bound rendered points
- empty io/network in monitor search now means no name filter, returning
  all device rows in one query instead of fanning out per device
- merge device names recorded in the monitor db into io/network/gpu
  options so removed devices stay selectable; move logic to service layer
- add composite indexes (name, created_at) on monitor_ios/monitor_networks
  and (product_name, created_at) on monitor_gpus
- stop migrating the unused MonitorGPU model into MonitorDB
- fix the memory param which was compared against "mem" and never matched

* chore(monitor): remove monitor test fixtures

---------

Co-authored-by: ssongliu <sloooop1x@gmail.com>
This commit is contained in:
CN_CoreSteb
2026-09-29 15:34:16 +08:00
committed by GitHub
co-authored by ssongliu
parent d0183f460e
commit 2189d9229a
6 changed files with 108 additions and 30 deletions
+2 -20
View File
@@ -1,13 +1,9 @@
package v2
import (
"sort"
"github.com/1Panel-dev/1Panel/agent/app/api/v2/helper"
"github.com/1Panel-dev/1Panel/agent/app/dto"
"github.com/gin-gonic/gin"
"github.com/shirou/gopsutil/v4/disk"
"github.com/shirou/gopsutil/v4/net"
)
// @Tags Monitor
@@ -96,14 +92,7 @@ func (b *BaseApi) UpdateMonitorSetting(c *gin.Context) {
// @Security Timestamp
// @Router /hosts/monitor/netoptions [get]
func (b *BaseApi) GetNetworkOptions(c *gin.Context) {
netStat, _ := net.IOCounters(true)
var options []string
options = append(options, "all")
for _, net := range netStat {
options = append(options, net.Name)
}
sort.Strings(options)
helper.SuccessWithData(c, options)
helper.SuccessWithData(c, monitorService.LoadNetworkOptions())
}
// @Tags Monitor
@@ -113,12 +102,5 @@ func (b *BaseApi) GetNetworkOptions(c *gin.Context) {
// @Security Timestamp
// @Router /hosts/monitor/iooptions [get]
func (b *BaseApi) GetIOOptions(c *gin.Context) {
diskStat, _ := disk.IOCounters()
var options []string
options = append(options, "all")
for _, net := range diskStat {
options = append(options, net.Name)
}
sort.Strings(options)
helper.SuccessWithData(c, options)
helper.SuccessWithData(c, monitorService.LoadIOOptions())
}
+14
View File
@@ -29,6 +29,8 @@ type IMonitorRepo interface {
GetGPUDevices() ([]model.MonitorGPU, error)
GetIO(opts ...DBOption) ([]model.MonitorIO, error)
GetNetwork(opts ...DBOption) ([]model.MonitorNetwork, error)
GetIONames() ([]string, error)
GetNetworkNames() ([]string, error)
CreateMonitorBase(model model.MonitorBase) error
BatchCreateMonitorGPU(list []model.MonitorGPU) error
@@ -86,6 +88,18 @@ func (u *MonitorRepo) GetNetwork(opts ...DBOption) ([]model.MonitorNetwork, erro
err := db.Find(&data).Error
return data, err
}
func (u *MonitorRepo) GetIONames() ([]string, error) {
var names []string
err := global.MonitorDB.Model(&model.MonitorIO{}).Distinct().Pluck("name", &names).Error
return names, err
}
func (u *MonitorRepo) GetNetworkNames() ([]string, error) {
var names []string
err := global.MonitorDB.Model(&model.MonitorNetwork{}).Distinct().Pluck("name", &names).Error
return names, err
}
func (u *MonitorRepo) CreateMonitorBase(model model.MonitorBase) error {
return global.MonitorDB.Create(&model).Error
}
+55 -3
View File
@@ -52,6 +52,8 @@ type IMonitorService interface {
UpdateSetting(key, value string) error
CleanData(monitorType string) error
LoadIOOptions() []string
LoadNetworkOptions() []string
LoadVLLMMonitorData(req dto.MonitorVLLMSearch) (dto.MonitorVLLMData, error)
LoadVLLMCurrent(ctx context.Context, req dto.MonitorVLLMCurrent) (model.MonitorVLLM, error)
CleanVLLMMonitor(req dto.MonitorVLLMClean) error
@@ -92,7 +94,7 @@ func (m *MonitorService) LoadMonitorData(req dto.MonitorSearch) ([]dto.MonitorDa
base.TopCPUItems = processes
base.TopCPU = ""
}
if req.Param == "all" || req.Param == "mem" {
if req.Param == "all" || req.Param == "memory" {
var processes []dto.Process
_ = json.Unmarshal([]byte(base.TopMem), &processes)
base.TopMemItems = processes
@@ -103,7 +105,11 @@ func (m *MonitorService) LoadMonitorData(req dto.MonitorSearch) ([]dto.MonitorDa
data = append(data, itemData)
}
if req.Param == "all" || req.Param == "io" {
bases, err := monitorRepo.GetIO(repo.WithByName(req.IO), repo.WithByCreatedAt(req.StartTime, req.EndTime))
ioOpts := []repo.DBOption{repo.WithByCreatedAt(req.StartTime, req.EndTime)}
if len(req.IO) != 0 {
ioOpts = append(ioOpts, repo.WithByName(req.IO))
}
bases, err := monitorRepo.GetIO(ioOpts...)
if err != nil {
return nil, err
}
@@ -117,7 +123,11 @@ func (m *MonitorService) LoadMonitorData(req dto.MonitorSearch) ([]dto.MonitorDa
data = append(data, itemData)
}
if req.Param == "all" || req.Param == "network" {
bases, err := monitorRepo.GetNetwork(repo.WithByName(req.Network), repo.WithByCreatedAt(req.StartTime, req.EndTime))
netOpts := []repo.DBOption{repo.WithByCreatedAt(req.StartTime, req.EndTime)}
if len(req.Network) != 0 {
netOpts = append(netOpts, repo.WithByName(req.Network))
}
bases, err := monitorRepo.GetNetwork(netOpts...)
if err != nil {
return nil, err
}
@@ -133,6 +143,37 @@ func (m *MonitorService) LoadMonitorData(req dto.MonitorSearch) ([]dto.MonitorDa
return data, nil
}
func (m *MonitorService) LoadIOOptions() []string {
optionSet := make(map[string]struct{})
if diskStat, err := disk.IOCounters(); err == nil {
for _, item := range diskStat {
optionSet[item.Name] = struct{}{}
}
}
// union with names recorded in the monitor db so removed devices stay selectable
if names, err := monitorRepo.GetIONames(); err == nil {
for _, name := range names {
optionSet[name] = struct{}{}
}
}
return sortedMonitorOptions(optionSet)
}
func (m *MonitorService) LoadNetworkOptions() []string {
optionSet := make(map[string]struct{})
if netStat, err := net.IOCounters(true); err == nil {
for _, item := range netStat {
optionSet[item.Name] = struct{}{}
}
}
if names, err := monitorRepo.GetNetworkNames(); err == nil {
for _, name := range names {
optionSet[name] = struct{}{}
}
}
return sortedMonitorOptions(optionSet)
}
func (m *MonitorService) LoadGPUOptions() dto.MonitorGPUOptions {
var data dto.MonitorGPUOptions
exist, client := accelerator.New()
@@ -942,3 +983,14 @@ func loadHostSysPath() string {
}
return hostSys
}
func sortedMonitorOptions(optionSet map[string]struct{}) []string {
options := make([]string, 0, len(optionSet))
for name := range optionSet {
if len(name) != 0 && name != "all" {
options = append(options, name)
}
}
sort.Strings(options)
return append([]string{"all"}, options...)
}
+33 -1
View File
@@ -4,6 +4,7 @@ import (
"os"
"os/exec"
"strings"
"time"
"github.com/1Panel-dev/1Panel/agent/app/dto"
"github.com/1Panel-dev/1Panel/agent/app/model"
@@ -13,6 +14,7 @@ import (
"github.com/1Panel-dev/1Panel/agent/global"
"github.com/1Panel-dev/1Panel/agent/utils/alert_push"
"github.com/1Panel-dev/1Panel/agent/utils/xpack"
"gorm.io/gorm"
)
func Init() {
@@ -174,6 +176,36 @@ func initAlertTask() {
}
func initMonitorDB() {
_ = global.MonitorDB.AutoMigrate(&model.MonitorBase{}, &model.MonitorNetwork{}, &model.MonitorGPU{}, &model.MonitorIO{})
_ = global.MonitorDB.AutoMigrate(&model.MonitorBase{}, &model.MonitorNetwork{}, &model.MonitorIO{})
_ = global.GPUMonitorDB.AutoMigrate(&model.MonitorGPU{})
_ = global.TaskDB.AutoMigrate(&model.Task{})
// building indexes on large monitor tables can take seconds, keep it off the startup path;
// WAL mode leaves readers unblocked and busy_timeout covers the collector's inserts meanwhile
go ensureMonitorIndexes()
}
func ensureMonitorIndexes() {
indexes := []struct {
db *gorm.DB
stmt string
}{
// created_at alone serves unfiltered range queries and retention cleanup
{global.MonitorDB, "CREATE INDEX IF NOT EXISTS idx_monitor_bases_created ON monitor_bases(created_at)"},
{global.MonitorDB, "CREATE INDEX IF NOT EXISTS idx_monitor_ios_created ON monitor_ios(created_at)"},
{global.MonitorDB, "CREATE INDEX IF NOT EXISTS idx_monitor_networks_created ON monitor_networks(created_at)"},
{global.GPUMonitorDB, "CREATE INDEX IF NOT EXISTS idx_monitor_gpus_created ON monitor_gpus(created_at)"},
// (name, created_at) serves per-device range queries and distinct name lookups
{global.MonitorDB, "CREATE INDEX IF NOT EXISTS idx_monitor_ios_name_created ON monitor_ios(name, created_at)"},
{global.MonitorDB, "CREATE INDEX IF NOT EXISTS idx_monitor_networks_name_created ON monitor_networks(name, created_at)"},
{global.GPUMonitorDB, "CREATE INDEX IF NOT EXISTS idx_monitor_gpus_product_created ON monitor_gpus(product_name, created_at)"},
}
start := time.Now()
for _, index := range indexes {
if err := index.db.Exec(index.stmt).Error; err != nil {
global.LOG.Warnf("create monitor index failed, stmt: %s, err: %v", index.stmt, err)
}
}
if elapsed := time.Since(start); elapsed > time.Second {
global.LOG.Infof("monitor indexes ready, took %s", elapsed)
}
}
@@ -139,6 +139,7 @@ function initChart() {
data: item?.data,
showSymbol: item.showSymbol ?? false,
symbolSize: item.symbolSize,
sampling: 'lttb',
yAxisIndex: item.yAxisIndex ? 1 : null,
});
});
@@ -227,6 +228,7 @@ function initChart() {
};
const tooltip = props.option.tooltip || defaultTooltip;
const option = {
animation: (props.option.xData?.length ?? 0) <= 3000,
title: [
{
left: props.option.titleLeft ?? 'center',
@@ -327,11 +327,7 @@ const searchGlobal = () => {
timeRangeMemory.value = timeRangeGlobal.value;
timeRangeIO.value = timeRangeGlobal.value;
timeRangeNetwork.value = timeRangeGlobal.value;
search('load');
search('cpu');
search('memory');
search('io');
search('network');
search('all');
};
const search = async (param: string) => {
@@ -355,7 +351,7 @@ const search = async (param: string) => {
searchInfo.network = networkChoose.value || 'all';
break;
case 'all':
searchTime.value = timeRangeNetwork.value;
searchTime.value = timeRangeGlobal.value;
searchInfo.io = ioChoose.value || 'all';
searchInfo.network = networkChoose.value || 'all';
break;