mirror of
https://github.com/henrygd/beszel.git
synced 2026-04-16 18:01:50 +02:00
Compare commits
6 Commits
fd4ac60908
...
755-xpu-sm
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
14f7480915 | ||
|
|
aab5725d82 | ||
|
|
e94a1cd421 | ||
|
|
73c1a1b208 | ||
|
|
0526c88ce0 | ||
|
|
a2e9056a00 |
@@ -4,42 +4,36 @@ package agent
|
||||
import (
|
||||
"beszel"
|
||||
"beszel/internal/entities/system"
|
||||
"context"
|
||||
"log/slog"
|
||||
"os"
|
||||
"strings"
|
||||
"sync"
|
||||
"time"
|
||||
|
||||
"github.com/shirou/gopsutil/v4/common"
|
||||
)
|
||||
|
||||
type Agent struct {
|
||||
sync.Mutex // Used to lock agent while collecting data
|
||||
debug bool // true if LOG_LEVEL is set to debug
|
||||
zfs bool // true if system has arcstats
|
||||
memCalc string // Memory calculation formula
|
||||
fsNames []string // List of filesystem device names being monitored
|
||||
fsStats map[string]*system.FsStats // Keeps track of disk stats for each filesystem
|
||||
netInterfaces map[string]struct{} // Stores all valid network interfaces
|
||||
netIoStats system.NetIoStats // Keeps track of bandwidth usage
|
||||
dockerManager *dockerManager // Manages Docker API requests
|
||||
sensorsContext context.Context // Sensors context to override sys location
|
||||
sensorsWhitelist map[string]struct{} // List of sensors to monitor
|
||||
primarySensor string // Value of PRIMARY_SENSOR env var
|
||||
systemInfo system.Info // Host system info
|
||||
gpuManager *GPUManager // Manages GPU data
|
||||
cache *SessionCache // Cache for system stats based on primary session ID
|
||||
sync.Mutex // Used to lock agent while collecting data
|
||||
debug bool // true if LOG_LEVEL is set to debug
|
||||
zfs bool // true if system has arcstats
|
||||
memCalc string // Memory calculation formula
|
||||
fsNames []string // List of filesystem device names being monitored
|
||||
fsStats map[string]*system.FsStats // Keeps track of disk stats for each filesystem
|
||||
netInterfaces map[string]struct{} // Stores all valid network interfaces
|
||||
netIoStats system.NetIoStats // Keeps track of bandwidth usage
|
||||
dockerManager *dockerManager // Manages Docker API requests
|
||||
sensorConfig *SensorConfig // Sensors config
|
||||
systemInfo system.Info // Host system info
|
||||
gpuManager *GPUManager // Manages GPU data
|
||||
cache *SessionCache // Cache for system stats based on primary session ID
|
||||
}
|
||||
|
||||
func NewAgent() *Agent {
|
||||
agent := &Agent{
|
||||
sensorsContext: context.Background(),
|
||||
fsStats: make(map[string]*system.FsStats),
|
||||
cache: NewSessionCache(69 * time.Second),
|
||||
fsStats: make(map[string]*system.FsStats),
|
||||
cache: NewSessionCache(69 * time.Second),
|
||||
}
|
||||
agent.memCalc, _ = GetEnv("MEM_CALC")
|
||||
agent.primarySensor, _ = GetEnv("PRIMARY_SENSOR")
|
||||
agent.sensorConfig = agent.newSensorConfig()
|
||||
// Set up slog with a log level determined by the LOG_LEVEL env var
|
||||
if logLevelStr, exists := GetEnv("LOG_LEVEL"); exists {
|
||||
switch strings.ToLower(logLevelStr) {
|
||||
@@ -55,24 +49,6 @@ func NewAgent() *Agent {
|
||||
|
||||
slog.Debug(beszel.Version)
|
||||
|
||||
// Set sensors context (allows overriding sys location for sensors)
|
||||
if sysSensors, exists := GetEnv("SYS_SENSORS"); exists {
|
||||
slog.Info("SYS_SENSORS", "path", sysSensors)
|
||||
agent.sensorsContext = context.WithValue(agent.sensorsContext,
|
||||
common.EnvKey, common.EnvMap{common.HostSysEnvKey: sysSensors},
|
||||
)
|
||||
}
|
||||
|
||||
// Set sensors whitelist
|
||||
if sensors, exists := GetEnv("SENSORS"); exists {
|
||||
agent.sensorsWhitelist = make(map[string]struct{})
|
||||
for sensor := range strings.SplitSeq(sensors, ",") {
|
||||
if sensor != "" {
|
||||
agent.sensorsWhitelist[sensor] = struct{}{}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// initialize system info / docker manager
|
||||
agent.initializeSystemInfo()
|
||||
agent.initializeDiskInfo()
|
||||
|
||||
@@ -4,6 +4,7 @@ import (
|
||||
"beszel/internal/entities/system"
|
||||
"bufio"
|
||||
"bytes"
|
||||
"encoding/csv"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"os/exec"
|
||||
@@ -21,11 +22,13 @@ const (
|
||||
nvidiaSmiCmd = "nvidia-smi"
|
||||
rocmSmiCmd = "rocm-smi"
|
||||
tegraStatsCmd = "tegrastats"
|
||||
xpuSmiCmd = "xpu-smi"
|
||||
|
||||
// Polling intervals
|
||||
nvidiaSmiInterval = "4" // in seconds
|
||||
tegraStatsInterval = "3700" // in milliseconds
|
||||
rocmSmiInterval = 4300 * time.Millisecond
|
||||
xpuSmiInterval = 4
|
||||
|
||||
// Command retry and timeout constants
|
||||
retryWaitTime = 5 * time.Second
|
||||
@@ -41,10 +44,11 @@ const (
|
||||
// GPUManager manages data collection for GPUs (either Nvidia or AMD)
|
||||
type GPUManager struct {
|
||||
sync.Mutex
|
||||
nvidiaSmi bool
|
||||
rocmSmi bool
|
||||
tegrastats bool
|
||||
GpuDataMap map[string]*system.GPUData
|
||||
nvidiaSmi bool
|
||||
rocmSmi bool
|
||||
tegrastats bool
|
||||
intelXpuSmi bool
|
||||
GpuDataMap map[string]*system.GPUData
|
||||
}
|
||||
|
||||
// RocmSmiJson represents the JSON structure of rocm-smi output
|
||||
@@ -160,6 +164,59 @@ func (gm *GPUManager) getJetsonParser() func(output []byte) bool {
|
||||
}
|
||||
}
|
||||
|
||||
func (gm *GPUManager) parseIntelData(output []byte) bool {
|
||||
gm.Lock()
|
||||
defer gm.Unlock()
|
||||
reader := csv.NewReader(bytes.NewReader(output))
|
||||
records, err := reader.ReadAll()
|
||||
if err != nil {
|
||||
slog.Warn("Failed to parse Intel GPU data", "err", err)
|
||||
return false
|
||||
}
|
||||
|
||||
header := []string{"Timestamp", "DeviceId", "GPU Power (W)", "GPU Frequency (MHz)", "GPU Memory Utilization (%)", "GPU Memory Used (MiB)"}
|
||||
gpuData := &system.GPUData{Name: "GPU"}
|
||||
gm.GpuDataMap["0"] = gpuData
|
||||
|
||||
for _, record := range records {
|
||||
if strings.Join(record, ",") == strings.Join(header, ",") {
|
||||
slog.Debug("Skipping header", "header", record)
|
||||
continue
|
||||
}
|
||||
var memoryUtilization *float64
|
||||
var memoryUsed *float64
|
||||
for i, field := range header {
|
||||
if field == "Timestamp" {
|
||||
continue
|
||||
}
|
||||
stripped := strings.TrimSpace(record[i])
|
||||
value, err := strconv.ParseFloat(stripped, 64)
|
||||
if err != nil {
|
||||
slog.Warn("Failed to parse field", "field", field, "value", stripped, "err", err)
|
||||
continue
|
||||
}
|
||||
|
||||
switch field {
|
||||
case "GPU Power (W)":
|
||||
gpuData.Power += value
|
||||
case "GPU Frequency (MHz)":
|
||||
gpuData.Usage += value
|
||||
case "GPU Memory Utilization (%)":
|
||||
memoryUtilization = &value
|
||||
case "GPU Memory Used (MiB)":
|
||||
memoryUsed = &value
|
||||
}
|
||||
}
|
||||
if memoryUtilization != nil && memoryUsed != nil {
|
||||
gpuData.MemoryUsed = *memoryUsed
|
||||
gpuData.MemoryTotal = (*memoryUsed / *memoryUtilization) * 100 // convert to total memory
|
||||
}
|
||||
}
|
||||
gpuData.Count++
|
||||
|
||||
return true
|
||||
}
|
||||
|
||||
// parseNvidiaData parses the output of nvidia-smi and updates the GPUData map
|
||||
func (gm *GPUManager) parseNvidiaData(output []byte) bool {
|
||||
gm.Lock()
|
||||
@@ -278,10 +335,14 @@ func (gm *GPUManager) detectGPUs() error {
|
||||
gm.tegrastats = true
|
||||
gm.nvidiaSmi = false
|
||||
}
|
||||
if gm.nvidiaSmi || gm.rocmSmi || gm.tegrastats {
|
||||
fmt.Println("Looking for gpus")
|
||||
if _, err := exec.LookPath(xpuSmiCmd); err == nil {
|
||||
gm.intelXpuSmi = true
|
||||
}
|
||||
if gm.nvidiaSmi || gm.rocmSmi || gm.tegrastats || gm.intelXpuSmi {
|
||||
return nil
|
||||
}
|
||||
return fmt.Errorf("no GPU found - install nvidia-smi, rocm-smi, or tegrastats")
|
||||
return fmt.Errorf("no GPU found - install nvidia-smi, rocm-smi, intel_gpu_top, or tegrastats")
|
||||
}
|
||||
|
||||
// startCollector starts the appropriate GPU data collector based on the command
|
||||
@@ -318,6 +379,10 @@ func (gm *GPUManager) startCollector(command string) {
|
||||
time.Sleep(rocmSmiInterval)
|
||||
}
|
||||
}()
|
||||
case xpuSmiCmd:
|
||||
collector.cmdArgs = []string{"dump", "-d", "-1", "-m", "1,2,5,18", "-i", strconv.Itoa(xpuSmiInterval)}
|
||||
collector.parse = gm.parseIntelData
|
||||
go collector.start()
|
||||
}
|
||||
}
|
||||
|
||||
@@ -338,6 +403,9 @@ func NewGPUManager() (*GPUManager, error) {
|
||||
if gm.tegrastats {
|
||||
gm.startCollector(tegraStatsCmd)
|
||||
}
|
||||
if gm.intelXpuSmi {
|
||||
gm.startCollector(xpuSmiCmd)
|
||||
}
|
||||
|
||||
return &gm, nil
|
||||
}
|
||||
|
||||
143
beszel/internal/agent/sensors.go
Normal file
143
beszel/internal/agent/sensors.go
Normal file
@@ -0,0 +1,143 @@
|
||||
package agent
|
||||
|
||||
import (
|
||||
"beszel/internal/entities/system"
|
||||
"context"
|
||||
"log/slog"
|
||||
"path"
|
||||
"strconv"
|
||||
"strings"
|
||||
|
||||
"github.com/shirou/gopsutil/v4/common"
|
||||
"github.com/shirou/gopsutil/v4/sensors"
|
||||
)
|
||||
|
||||
type SensorConfig struct {
|
||||
context context.Context
|
||||
sensors map[string]struct{}
|
||||
primarySensor string
|
||||
isBlacklist bool
|
||||
hasWildcards bool
|
||||
skipCollection bool
|
||||
}
|
||||
|
||||
func (a *Agent) newSensorConfig() *SensorConfig {
|
||||
primarySensor, _ := GetEnv("PRIMARY_SENSOR")
|
||||
sysSensors, _ := GetEnv("SYS_SENSORS")
|
||||
sensorsEnvVal, sensorsSet := GetEnv("SENSORS")
|
||||
skipCollection := sensorsSet && sensorsEnvVal == ""
|
||||
|
||||
return a.newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal, skipCollection)
|
||||
}
|
||||
|
||||
// newSensorConfigWithEnv creates a SensorConfig with the provided environment variables
|
||||
// sensorsSet indicates if the SENSORS environment variable was explicitly set (even to empty string)
|
||||
func (a *Agent) newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal string, skipCollection bool) *SensorConfig {
|
||||
config := &SensorConfig{
|
||||
context: context.Background(),
|
||||
primarySensor: primarySensor,
|
||||
skipCollection: skipCollection,
|
||||
sensors: make(map[string]struct{}),
|
||||
}
|
||||
|
||||
// Set sensors context (allows overriding sys location for sensors)
|
||||
if sysSensors != "" {
|
||||
slog.Info("SYS_SENSORS", "path", sysSensors)
|
||||
config.context = context.WithValue(config.context,
|
||||
common.EnvKey, common.EnvMap{common.HostSysEnvKey: sysSensors},
|
||||
)
|
||||
}
|
||||
|
||||
// handle blacklist
|
||||
if strings.HasPrefix(sensorsEnvVal, "-") {
|
||||
config.isBlacklist = true
|
||||
sensorsEnvVal = sensorsEnvVal[1:]
|
||||
}
|
||||
|
||||
for sensor := range strings.SplitSeq(sensorsEnvVal, ",") {
|
||||
sensor = strings.TrimSpace(sensor)
|
||||
if sensor != "" {
|
||||
config.sensors[sensor] = struct{}{}
|
||||
if strings.Contains(sensor, "*") {
|
||||
config.hasWildcards = true
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return config
|
||||
}
|
||||
|
||||
// updateTemperatures updates the agent with the latest sensor temperatures
|
||||
func (a *Agent) updateTemperatures(systemStats *system.Stats) {
|
||||
// skip if sensors whitelist is set to empty string
|
||||
if a.sensorConfig.skipCollection {
|
||||
slog.Debug("Skipping temperature collection")
|
||||
return
|
||||
}
|
||||
|
||||
// reset high temp
|
||||
a.systemInfo.DashboardTemp = 0
|
||||
|
||||
// get sensor data
|
||||
temps, _ := sensors.TemperaturesWithContext(a.sensorConfig.context)
|
||||
slog.Debug("Temperature", "sensors", temps)
|
||||
|
||||
// return if no sensors
|
||||
if len(temps) == 0 {
|
||||
return
|
||||
}
|
||||
|
||||
systemStats.Temperatures = make(map[string]float64, len(temps))
|
||||
for i, sensor := range temps {
|
||||
// skip if temperature is unreasonable
|
||||
if sensor.Temperature <= 0 || sensor.Temperature >= 200 {
|
||||
continue
|
||||
}
|
||||
sensorName := sensor.SensorKey
|
||||
if _, ok := systemStats.Temperatures[sensorName]; ok {
|
||||
// if key already exists, append int to key
|
||||
sensorName = sensorName + "_" + strconv.Itoa(i)
|
||||
}
|
||||
// skip if not in whitelist or blacklist
|
||||
if !isValidSensor(sensorName, a.sensorConfig) {
|
||||
continue
|
||||
}
|
||||
// set dashboard temperature
|
||||
if a.sensorConfig.primarySensor == "" {
|
||||
a.systemInfo.DashboardTemp = max(a.systemInfo.DashboardTemp, sensor.Temperature)
|
||||
} else if a.sensorConfig.primarySensor == sensorName {
|
||||
a.systemInfo.DashboardTemp = sensor.Temperature
|
||||
}
|
||||
systemStats.Temperatures[sensorName] = twoDecimals(sensor.Temperature)
|
||||
}
|
||||
}
|
||||
|
||||
// isValidSensor checks if a sensor is valid based on the sensor name and the sensor config
|
||||
func isValidSensor(sensorName string, config *SensorConfig) bool {
|
||||
// if no sensors configured, everything is valid
|
||||
if len(config.sensors) == 0 {
|
||||
return true
|
||||
}
|
||||
|
||||
// Exact match - return true if whitelist, false if blacklist
|
||||
if _, exactMatch := config.sensors[sensorName]; exactMatch {
|
||||
return !config.isBlacklist
|
||||
}
|
||||
|
||||
// If no wildcards, return true if blacklist, false if whitelist
|
||||
if !config.hasWildcards {
|
||||
return config.isBlacklist
|
||||
}
|
||||
|
||||
// Check for wildcard patterns
|
||||
for pattern := range config.sensors {
|
||||
if !strings.Contains(pattern, "*") {
|
||||
continue
|
||||
}
|
||||
if match, _ := path.Match(pattern, sensorName); match {
|
||||
return !config.isBlacklist
|
||||
}
|
||||
}
|
||||
|
||||
return config.isBlacklist
|
||||
}
|
||||
374
beszel/internal/agent/sensors_test.go
Normal file
374
beszel/internal/agent/sensors_test.go
Normal file
@@ -0,0 +1,374 @@
|
||||
//go:build testing
|
||||
// +build testing
|
||||
|
||||
package agent
|
||||
|
||||
import (
|
||||
"context"
|
||||
"os"
|
||||
"testing"
|
||||
|
||||
"github.com/shirou/gopsutil/v4/common"
|
||||
"github.com/stretchr/testify/assert"
|
||||
"github.com/stretchr/testify/require"
|
||||
)
|
||||
|
||||
func TestIsValidSensor(t *testing.T) {
|
||||
tests := []struct {
|
||||
name string
|
||||
sensorName string
|
||||
config *SensorConfig
|
||||
expectedValid bool
|
||||
}{
|
||||
{
|
||||
name: "Whitelist - sensor in list",
|
||||
sensorName: "cpu_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"cpu_temp": {}},
|
||||
isBlacklist: false,
|
||||
},
|
||||
expectedValid: true,
|
||||
},
|
||||
{
|
||||
name: "Whitelist - sensor not in list",
|
||||
sensorName: "gpu_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"cpu_temp": {}},
|
||||
isBlacklist: false,
|
||||
},
|
||||
expectedValid: false,
|
||||
},
|
||||
{
|
||||
name: "Blacklist - sensor in list",
|
||||
sensorName: "cpu_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"cpu_temp": {}},
|
||||
isBlacklist: true,
|
||||
},
|
||||
expectedValid: false,
|
||||
},
|
||||
{
|
||||
name: "Blacklist - sensor not in list",
|
||||
sensorName: "gpu_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"cpu_temp": {}},
|
||||
isBlacklist: true,
|
||||
},
|
||||
expectedValid: true,
|
||||
},
|
||||
{
|
||||
name: "Whitelist with wildcard - matching pattern",
|
||||
sensorName: "core_0_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"core_*_temp": {}},
|
||||
isBlacklist: false,
|
||||
hasWildcards: true,
|
||||
},
|
||||
expectedValid: true,
|
||||
},
|
||||
{
|
||||
name: "Whitelist with wildcard - non-matching pattern",
|
||||
sensorName: "gpu_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"core_*_temp": {}},
|
||||
isBlacklist: false,
|
||||
hasWildcards: true,
|
||||
},
|
||||
expectedValid: false,
|
||||
},
|
||||
{
|
||||
name: "Blacklist with wildcard - matching pattern",
|
||||
sensorName: "core_0_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"core_*_temp": {}},
|
||||
isBlacklist: true,
|
||||
hasWildcards: true,
|
||||
},
|
||||
expectedValid: false,
|
||||
},
|
||||
{
|
||||
name: "Blacklist with wildcard - non-matching pattern",
|
||||
sensorName: "gpu_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"core_*_temp": {}},
|
||||
isBlacklist: true,
|
||||
hasWildcards: true,
|
||||
},
|
||||
expectedValid: true,
|
||||
},
|
||||
{
|
||||
name: "No sensors configured",
|
||||
sensorName: "any_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{},
|
||||
isBlacklist: false,
|
||||
hasWildcards: false,
|
||||
skipCollection: false,
|
||||
},
|
||||
expectedValid: true,
|
||||
},
|
||||
{
|
||||
name: "Mixed patterns in whitelist - exact match",
|
||||
sensorName: "cpu_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"cpu_temp": {}, "core_*_temp": {}},
|
||||
isBlacklist: false,
|
||||
hasWildcards: true,
|
||||
},
|
||||
expectedValid: true,
|
||||
},
|
||||
{
|
||||
name: "Mixed patterns in whitelist - wildcard match",
|
||||
sensorName: "core_1_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"cpu_temp": {}, "core_*_temp": {}},
|
||||
isBlacklist: false,
|
||||
hasWildcards: true,
|
||||
},
|
||||
expectedValid: true,
|
||||
},
|
||||
{
|
||||
name: "Mixed patterns in blacklist - exact match",
|
||||
sensorName: "cpu_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"cpu_temp": {}, "core_*_temp": {}},
|
||||
isBlacklist: true,
|
||||
hasWildcards: true,
|
||||
},
|
||||
expectedValid: false,
|
||||
},
|
||||
{
|
||||
name: "Mixed patterns in blacklist - wildcard match",
|
||||
sensorName: "core_1_temp",
|
||||
config: &SensorConfig{
|
||||
sensors: map[string]struct{}{"cpu_temp": {}, "core_*_temp": {}},
|
||||
isBlacklist: true,
|
||||
hasWildcards: true,
|
||||
},
|
||||
expectedValid: false,
|
||||
},
|
||||
}
|
||||
|
||||
for _, tt := range tests {
|
||||
t.Run(tt.name, func(t *testing.T) {
|
||||
result := isValidSensor(tt.sensorName, tt.config)
|
||||
assert.Equal(t, tt.expectedValid, result, "isValidSensor(%q, config) returned unexpected result", tt.sensorName)
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
func TestNewSensorConfigWithEnv(t *testing.T) {
|
||||
agent := &Agent{}
|
||||
|
||||
tests := []struct {
|
||||
name string
|
||||
primarySensor string
|
||||
sysSensors string
|
||||
sensors string
|
||||
skipCollection bool
|
||||
expectedConfig *SensorConfig
|
||||
}{
|
||||
{
|
||||
name: "Empty configuration",
|
||||
primarySensor: "",
|
||||
sysSensors: "",
|
||||
sensors: "",
|
||||
expectedConfig: &SensorConfig{
|
||||
context: context.Background(),
|
||||
primarySensor: "",
|
||||
sensors: map[string]struct{}{},
|
||||
isBlacklist: false,
|
||||
hasWildcards: false,
|
||||
skipCollection: false,
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "Explicitly set to empty string",
|
||||
primarySensor: "",
|
||||
sysSensors: "",
|
||||
sensors: "",
|
||||
skipCollection: true,
|
||||
expectedConfig: &SensorConfig{
|
||||
context: context.Background(),
|
||||
primarySensor: "",
|
||||
sensors: map[string]struct{}{},
|
||||
isBlacklist: false,
|
||||
hasWildcards: false,
|
||||
skipCollection: true,
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "Primary sensor only - should create sensor map",
|
||||
primarySensor: "cpu_temp",
|
||||
sysSensors: "",
|
||||
sensors: "",
|
||||
expectedConfig: &SensorConfig{
|
||||
context: context.Background(),
|
||||
primarySensor: "cpu_temp",
|
||||
sensors: map[string]struct{}{},
|
||||
isBlacklist: false,
|
||||
hasWildcards: false,
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "Whitelist sensors",
|
||||
primarySensor: "cpu_temp",
|
||||
sysSensors: "",
|
||||
sensors: "cpu_temp,gpu_temp",
|
||||
expectedConfig: &SensorConfig{
|
||||
context: context.Background(),
|
||||
primarySensor: "cpu_temp",
|
||||
sensors: map[string]struct{}{
|
||||
"cpu_temp": {},
|
||||
"gpu_temp": {},
|
||||
},
|
||||
isBlacklist: false,
|
||||
hasWildcards: false,
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "Blacklist sensors",
|
||||
primarySensor: "cpu_temp",
|
||||
sysSensors: "",
|
||||
sensors: "-cpu_temp,gpu_temp",
|
||||
expectedConfig: &SensorConfig{
|
||||
context: context.Background(),
|
||||
primarySensor: "cpu_temp",
|
||||
sensors: map[string]struct{}{
|
||||
"cpu_temp": {},
|
||||
"gpu_temp": {},
|
||||
},
|
||||
isBlacklist: true,
|
||||
hasWildcards: false,
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "Sensors with wildcard",
|
||||
primarySensor: "cpu_temp",
|
||||
sysSensors: "",
|
||||
sensors: "cpu_*,gpu_temp",
|
||||
expectedConfig: &SensorConfig{
|
||||
context: context.Background(),
|
||||
primarySensor: "cpu_temp",
|
||||
sensors: map[string]struct{}{
|
||||
"cpu_*": {},
|
||||
"gpu_temp": {},
|
||||
},
|
||||
isBlacklist: false,
|
||||
hasWildcards: true,
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "Sensors with whitespace",
|
||||
primarySensor: "cpu_temp",
|
||||
sysSensors: "",
|
||||
sensors: "cpu_*, gpu_temp",
|
||||
expectedConfig: &SensorConfig{
|
||||
context: context.Background(),
|
||||
primarySensor: "cpu_temp",
|
||||
sensors: map[string]struct{}{
|
||||
"cpu_*": {},
|
||||
"gpu_temp": {},
|
||||
},
|
||||
isBlacklist: false,
|
||||
hasWildcards: true,
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "With SYS_SENSORS path",
|
||||
primarySensor: "cpu_temp",
|
||||
sysSensors: "/custom/path",
|
||||
sensors: "cpu_temp",
|
||||
expectedConfig: &SensorConfig{
|
||||
primarySensor: "cpu_temp",
|
||||
sensors: map[string]struct{}{
|
||||
"cpu_temp": {},
|
||||
},
|
||||
isBlacklist: false,
|
||||
hasWildcards: false,
|
||||
},
|
||||
},
|
||||
}
|
||||
|
||||
for _, tt := range tests {
|
||||
t.Run(tt.name, func(t *testing.T) {
|
||||
result := agent.newSensorConfigWithEnv(tt.primarySensor, tt.sysSensors, tt.sensors, tt.skipCollection)
|
||||
|
||||
// Check primary sensor
|
||||
assert.Equal(t, tt.expectedConfig.primarySensor, result.primarySensor)
|
||||
|
||||
// Check sensor map
|
||||
if tt.expectedConfig.sensors == nil {
|
||||
assert.Nil(t, result.sensors)
|
||||
} else {
|
||||
assert.Equal(t, len(tt.expectedConfig.sensors), len(result.sensors))
|
||||
for sensor := range tt.expectedConfig.sensors {
|
||||
_, exists := result.sensors[sensor]
|
||||
assert.True(t, exists, "Sensor %s should exist in the result", sensor)
|
||||
}
|
||||
}
|
||||
|
||||
// Check flags
|
||||
assert.Equal(t, tt.expectedConfig.isBlacklist, result.isBlacklist)
|
||||
assert.Equal(t, tt.expectedConfig.hasWildcards, result.hasWildcards)
|
||||
|
||||
// Check context
|
||||
if tt.sysSensors != "" {
|
||||
// Verify context contains correct values
|
||||
envMap, ok := result.context.Value(common.EnvKey).(common.EnvMap)
|
||||
require.True(t, ok, "Context should contain EnvMap")
|
||||
sysPath, ok := envMap[common.HostSysEnvKey]
|
||||
require.True(t, ok, "EnvMap should contain HostSysEnvKey")
|
||||
assert.Equal(t, tt.sysSensors, sysPath)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
func TestNewSensorConfig(t *testing.T) {
|
||||
// Save original environment variables
|
||||
originalPrimary, hasPrimary := os.LookupEnv("BESZEL_AGENT_PRIMARY_SENSOR")
|
||||
originalSys, hasSys := os.LookupEnv("BESZEL_AGENT_SYS_SENSORS")
|
||||
originalSensors, hasSensors := os.LookupEnv("BESZEL_AGENT_SENSORS")
|
||||
|
||||
// Restore environment variables after the test
|
||||
defer func() {
|
||||
// Clean up test environment variables
|
||||
os.Unsetenv("BESZEL_AGENT_PRIMARY_SENSOR")
|
||||
os.Unsetenv("BESZEL_AGENT_SYS_SENSORS")
|
||||
os.Unsetenv("BESZEL_AGENT_SENSORS")
|
||||
|
||||
// Restore original values if they existed
|
||||
if hasPrimary {
|
||||
os.Setenv("BESZEL_AGENT_PRIMARY_SENSOR", originalPrimary)
|
||||
}
|
||||
if hasSys {
|
||||
os.Setenv("BESZEL_AGENT_SYS_SENSORS", originalSys)
|
||||
}
|
||||
if hasSensors {
|
||||
os.Setenv("BESZEL_AGENT_SENSORS", originalSensors)
|
||||
}
|
||||
}()
|
||||
|
||||
// Set test environment variables
|
||||
os.Setenv("BESZEL_AGENT_PRIMARY_SENSOR", "test_primary")
|
||||
os.Setenv("BESZEL_AGENT_SYS_SENSORS", "/test/path")
|
||||
os.Setenv("BESZEL_AGENT_SENSORS", "test_sensor1,test_*,test_sensor3")
|
||||
|
||||
agent := &Agent{}
|
||||
result := agent.newSensorConfig()
|
||||
|
||||
// Verify results
|
||||
assert.Equal(t, "test_primary", result.primarySensor)
|
||||
assert.NotNil(t, result.sensors)
|
||||
assert.Equal(t, 3, len(result.sensors))
|
||||
assert.True(t, result.hasWildcards)
|
||||
assert.False(t, result.isBlacklist)
|
||||
|
||||
// Check that sys sensors path is in context
|
||||
envMap, ok := result.context.Value(common.EnvKey).(common.EnvMap)
|
||||
require.True(t, ok, "Context should contain EnvMap")
|
||||
sysPath, ok := envMap[common.HostSysEnvKey]
|
||||
require.True(t, ok, "EnvMap should contain HostSysEnvKey")
|
||||
assert.Equal(t, "/test/path", sysPath)
|
||||
}
|
||||
@@ -16,7 +16,6 @@ import (
|
||||
"github.com/shirou/gopsutil/v4/host"
|
||||
"github.com/shirou/gopsutil/v4/mem"
|
||||
psutilNet "github.com/shirou/gopsutil/v4/net"
|
||||
"github.com/shirou/gopsutil/v4/sensors"
|
||||
)
|
||||
|
||||
// Sets initial / non-changing values about the host system
|
||||
@@ -200,16 +199,24 @@ func (a *Agent) getSystemStats() system.Stats {
|
||||
if systemStats.Temperatures == nil {
|
||||
systemStats.Temperatures = make(map[string]float64, len(gpuData))
|
||||
}
|
||||
highestTemp := 0.0
|
||||
for _, gpu := range gpuData {
|
||||
if gpu.Temperature > 0 {
|
||||
systemStats.Temperatures[gpu.Name] = gpu.Temperature
|
||||
if a.primarySensor == gpu.Name {
|
||||
if a.sensorConfig.primarySensor == gpu.Name {
|
||||
a.systemInfo.DashboardTemp = gpu.Temperature
|
||||
}
|
||||
if gpu.Temperature > highestTemp {
|
||||
highestTemp = gpu.Temperature
|
||||
}
|
||||
}
|
||||
// update high gpu percent for dashboard
|
||||
a.systemInfo.GpuPct = max(a.systemInfo.GpuPct, gpu.Usage)
|
||||
}
|
||||
// use highest temp for dashboard temp if dashboard temp is unset
|
||||
if a.systemInfo.DashboardTemp == 0 {
|
||||
a.systemInfo.DashboardTemp = highestTemp
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@@ -224,52 +231,6 @@ func (a *Agent) getSystemStats() system.Stats {
|
||||
return systemStats
|
||||
}
|
||||
|
||||
func (a *Agent) updateTemperatures(systemStats *system.Stats) {
|
||||
// skip if sensors whitelist is set to empty string
|
||||
if a.sensorsWhitelist != nil && len(a.sensorsWhitelist) == 0 {
|
||||
slog.Debug("Skipping temperature collection")
|
||||
return
|
||||
}
|
||||
|
||||
// reset high temp
|
||||
a.systemInfo.DashboardTemp = 0
|
||||
|
||||
// get sensor data
|
||||
temps, _ := sensors.TemperaturesWithContext(a.sensorsContext)
|
||||
slog.Debug("Temperature", "sensors", temps)
|
||||
|
||||
// return if no sensors
|
||||
if len(temps) == 0 {
|
||||
return
|
||||
}
|
||||
|
||||
systemStats.Temperatures = make(map[string]float64, len(temps))
|
||||
for i, sensor := range temps {
|
||||
// skip if temperature is unreasonable
|
||||
if sensor.Temperature <= 0 || sensor.Temperature >= 200 {
|
||||
continue
|
||||
}
|
||||
sensorName := sensor.SensorKey
|
||||
if _, ok := systemStats.Temperatures[sensorName]; ok {
|
||||
// if key already exists, append int to key
|
||||
sensorName = sensorName + "_" + strconv.Itoa(i)
|
||||
}
|
||||
// skip if not in whitelist
|
||||
if a.sensorsWhitelist != nil {
|
||||
if _, nameInWhitelist := a.sensorsWhitelist[sensorName]; !nameInWhitelist {
|
||||
continue
|
||||
}
|
||||
}
|
||||
// set dashboard temperature
|
||||
if a.primarySensor == "" {
|
||||
a.systemInfo.DashboardTemp = max(a.systemInfo.DashboardTemp, sensor.Temperature)
|
||||
} else if a.primarySensor == sensorName {
|
||||
a.systemInfo.DashboardTemp = sensor.Temperature
|
||||
}
|
||||
systemStats.Temperatures[sensorName] = twoDecimals(sensor.Temperature)
|
||||
}
|
||||
}
|
||||
|
||||
// Returns the size of the ZFS ARC memory cache in bytes
|
||||
func getARCSize() (uint64, error) {
|
||||
file, err := os.Open("/proc/spl/kstat/zfs/arcstats")
|
||||
|
||||
@@ -41,10 +41,23 @@ while getopts "k:p:h" opt; do
|
||||
esac
|
||||
done
|
||||
|
||||
# Exit if brew is not installed
|
||||
# Check if brew is installed, prompt to install if not
|
||||
if ! command -v brew &>/dev/null; then
|
||||
echo "Homebrew is not installed. Please install Homebrew and try again."
|
||||
exit 1
|
||||
read -p "Homebrew is not installed. Would you like to install it now? (y/n): " install_brew
|
||||
if [[ $install_brew =~ ^[Yy]$ ]]; then
|
||||
echo "Installing Homebrew..."
|
||||
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
|
||||
|
||||
# Verify installation was successful
|
||||
if ! command -v brew &>/dev/null; then
|
||||
echo "Homebrew installation failed. Please install manually and try again."
|
||||
exit 1
|
||||
fi
|
||||
echo "Homebrew installed successfully."
|
||||
else
|
||||
echo "Homebrew is required. Please install Homebrew and try again."
|
||||
exit 1
|
||||
fi
|
||||
fi
|
||||
|
||||
if [ -z "$KEY" ]; then
|
||||
@@ -54,7 +67,7 @@ fi
|
||||
mkdir -p ~/.config/beszel ~/.cache/beszel
|
||||
|
||||
echo "KEY=\"$KEY\"" >~/.config/beszel/beszel-agent.env
|
||||
echo "PORT=$PORT" >>~/.config/beszel/beszel-agent.env
|
||||
echo "LISTEN=$PORT" >>~/.config/beszel/beszel-agent.env
|
||||
|
||||
brew tap henrygd/beszel
|
||||
brew install beszel-agent
|
||||
@@ -67,4 +80,4 @@ echo "Restart: brew services restart beszel-agent"
|
||||
echo "Upgrade: brew upgrade beszel-agent"
|
||||
echo "Uninstall: brew uninstall beszel-agent"
|
||||
echo "View logs in ~/.cache/beszel/beszel-agent.log"
|
||||
echo "Change environment variables in ~/.config/beszel/beszel-agent.env"
|
||||
printf "Change environment variables in ~/.config/beszel/beszel-agent.env\n"
|
||||
|
||||
Reference in New Issue
Block a user