ClusterCockpit/cc-metric-collector
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-metric-collector.git synced 2024-12-25 23:19:06 +01:00
Code Issues Packages Projects Releases Wiki Activity

Fixed topology detection

Browse Source
This commit is contained in:
Holger Obermaier 2022-01-24 20:22:08 +01:00
parent 7953629940
commit 2026c3acd9
1 changed files with 101 additions and 74 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

175
collectors/cpufreqMetric.go
View File

@ -32,10 +32,19 @@ func readOneLine(filename string) (text string, ok bool) {
return return
} }
type CPUFreqCollectorCPU struct { type CPUFreqCollectorTopology struct {
// coreID, packageID, num_cores, num_package processor string // logical processor number (continuous, starting at 0)
tagSet map[string]string coreID string // socket local core ID
coreID_int int
physicalID string // socket / package ID
physicalID_int int
numPhysicalID string // number of sockets / packages
numPhysicalID_int int
isHT bool
numNonHT string // number of non hyperthreading processors
numNonHT_int int
scalingCurFreqFile string scalingCurFreqFile string
tagSet map[string]string
} }
// //
@ -48,10 +57,10 @@ type CPUFreqCollectorCPU struct {
// //
type CPUFreqCollector struct { type CPUFreqCollector struct {
MetricCollector MetricCollector
config struct { topology []CPUFreqCollectorTopology
config struct {
ExcludeMetrics []string `json:"exclude_metrics,omitempty"` ExcludeMetrics []string `json:"exclude_metrics,omitempty"`
} }
cpus []CPUFreqCollectorCPU
} }
func (m *CPUFreqCollector) Init(config []byte) error { func (m *CPUFreqCollector) Init(config []byte) error {
@ -64,9 +73,6 @@ func (m *CPUFreqCollector) Init(config []byte) error {
} }
} }
// Initialize CPU list
m.cpus = make([]CPUFreqCollectorCPU, 0)
// Loop for all CPU directories // Loop for all CPU directories
baseDir := "/sys/devices/system/cpu" baseDir := "/sys/devices/system/cpu"
globPattern := filepath.Join(baseDir, "cpu[0-9]*") globPattern := filepath.Join(baseDir, "cpu[0-9]*")
@ -78,82 +84,98 @@ func (m *CPUFreqCollector) Init(config []byte) error {
return fmt.Errorf("CPUFreqCollector.Init() unable to find any files with pattern %s", globPattern) return fmt.Errorf("CPUFreqCollector.Init() unable to find any files with pattern %s", globPattern)
} }
maxPackageID := 0 // Initialize CPU topology
maxCoreID := 0 m.topology = make([]CPUFreqCollectorTopology, len(cpuDirs))
for _, cpuDir := range cpuDirs { for _, cpuDir := range cpuDirs {
cpuID := strings.TrimPrefix(cpuDir, "/sys/devices/system/cpu/cpu") processor := strings.TrimPrefix(cpuDir, "/sys/devices/system/cpu/cpu")
processor_int, err := strconv.Atoi(processor)
if err != nil {
return fmt.Errorf("CPUFreqCollector.Init() unable to convert cpuID to int: %v", err)
}
// Read thread sibling list // Read package ID
threadSiblingListFile := filepath.Join(cpuDir, "topology", "thread_siblings_list") packageIDFile := filepath.Join(cpuDir, "topology", "physical_package_id")
threadSiblingList, ok := readOneLine(threadSiblingListFile) packageID, ok := readOneLine(packageIDFile)
if !ok { if !ok {
return fmt.Errorf("CPUFreqCollector.Init() unable to read thread siblings list from %s", threadSiblingListFile) return fmt.Errorf("CPUFreqCollector.Init() unable to read physical package ID from %s", packageIDFile)
}
packageID_int, err := strconv.Atoi(packageID)
if err != nil {
return fmt.Errorf("CPUFreqCollector.Init() unable to convert packageID to int: %v", err)
} }
// Read frequency only from first hardware thread // Read core ID
// Ignore Simultaneous Multithreading (SMT) / Hyper-Threading coreIDFile := filepath.Join(cpuDir, "topology", "core_id")
if strings.Split(threadSiblingList, ",")[0] == cpuID { coreID, ok := readOneLine(coreIDFile)
// Read package ID if !ok {
packageIDFile := filepath.Join(cpuDir, "topology", "physical_package_id") return fmt.Errorf("CPUFreqCollector.Init() unable to read core ID from %s", coreIDFile)
packageID, ok := readOneLine(packageIDFile) }
if !ok { coreID_int, err := strconv.Atoi(coreID)
return fmt.Errorf("CPUFreqCollector.Init() unable to read physical package ID from %s", packageIDFile) if err != nil {
} return fmt.Errorf("CPUFreqCollector.Init() unable to convert coreID to int: %v", err)
packageID_int, err := strconv.Atoi(packageID) }
if err != nil {
return fmt.Errorf("CPUFreqCollector.Init() unable to convert packageID to int: %v", err)
}
// Update maxPackageID // Check access to current frequency file
if packageID_int > maxPackageID { scalingCurFreqFile := filepath.Join(cpuDir, "cpufreq", "scaling_cur_freq")
maxPackageID = packageID_int err = unix.Access(scalingCurFreqFile, unix.R_OK)
} if err != nil {
return fmt.Errorf("CPUFreqCollector.Init() unable to access %s: %v", scalingCurFreqFile, err)
}
// Read core ID t := &m.topology[processor_int]
coreIDFile := filepath.Join(cpuDir, "topology", "core_id") t.processor = processor
coreID, ok := readOneLine(coreIDFile) t.physicalID = packageID
if !ok { t.physicalID_int = packageID_int
return fmt.Errorf("CPUFreqCollector.Init() unable to read core ID from %s", coreIDFile) t.coreID = coreID
} t.coreID_int = coreID_int
coreID_int, err := strconv.Atoi(coreID) t.scalingCurFreqFile = scalingCurFreqFile
if err != nil { }
return fmt.Errorf("CPUFreqCollector.Init() unable to convert coreID to int: %v", err)
}
// Update maxCoreID // is processor a hyperthread?
if coreID_int > maxCoreID { coreSeenBefore := make(map[string]bool)
maxCoreID = coreID_int for i := range m.topology {
} t := &m.topology[i]
// Check access to current frequency file globalID := t.physicalID + ":" + t.coreID
scalingCurFreqFile := filepath.Join(cpuDir, "cpufreq", "scaling_cur_freq") t.isHT = coreSeenBefore[globalID]
err = unix.Access(scalingCurFreqFile, unix.R_OK) coreSeenBefore[globalID] = true
if err != nil { }
return fmt.Errorf("CPUFreqCollector.Init() unable to access %s: %v", scalingCurFreqFile, err)
}
m.cpus = append( // number of non hyper thread cores and packages / sockets
m.cpus, numNonHT_int := 0
CPUFreqCollectorCPU{ maxPhysicalID := 0
tagSet: map[string]string{ for i := range m.topology {
"type": "cpu", t := &m.topology[i]
"type-id": strings.TrimSpace(coreID),
"packageID": strings.TrimSpace(packageID), // Update maxPackageID
}, if t.physicalID_int > maxPhysicalID {
scalingCurFreqFile: scalingCurFreqFile, maxPhysicalID = t.physicalID_int
}) }
if !t.isHT {
numNonHT_int++
} }
} }
// Add num packages and num cores as tags numPhysicalID_int := maxPhysicalID + 1
numPackages := strconv.Itoa(maxPackageID + 1) numPhysicalID := fmt.Sprint(numPhysicalID_int)
numCores := strconv.Itoa(maxCoreID + 1) numNonHT := fmt.Sprint(numNonHT_int)
for i := range m.cpus { for i := range m.topology {
c := &m.cpus[i] t := &m.topology[i]
c.tagSet["num_core"] = numCores t.numPhysicalID = numPhysicalID
c.tagSet["num_package"] = numPackages t.numPhysicalID_int = numPhysicalID_int
t.numNonHT = numNonHT
t.numNonHT_int = numNonHT_int
t.tagSet = map[string]string{
"type": "cpu",
"type-id": t.processor,
"num_core": t.numNonHT,
"package_id": t.physicalID,
"num_package": t.numPhysicalID,
}
} }
fmt.Printf("%+v\n", m.topology)
m.init = true m.init = true
return nil return nil
} }
@ -164,13 +186,18 @@ func (m *CPUFreqCollector) Read(interval time.Duration, out *[]lp.MutableMetric)
} }
now := time.Now() now := time.Now()
for i := range m.cpus { for i := range m.topology {
cpu := &m.cpus[i] t := &m.topology[i]
// skip hyperthreads
if t.isHT {
continue
}
// Read current frequency // Read current frequency
line, ok := readOneLine(cpu.scalingCurFreqFile) line, ok := readOneLine(t.scalingCurFreqFile)
if !ok { if !ok {
log.Printf("CPUFreqCollector.Read(): Failed to read one line from file '%s'", cpu.scalingCurFreqFile) log.Printf("CPUFreqCollector.Read(): Failed to read one line from file '%s'", t.scalingCurFreqFile)
continue continue
} }
cpuFreq, err := strconv.Atoi(line) cpuFreq, err := strconv.Atoi(line)
@ -179,7 +206,7 @@ func (m *CPUFreqCollector) Read(interval time.Duration, out *[]lp.MutableMetric)
continue continue
} }
y, err := lp.New("cpufreq", cpu.tagSet, map[string]interface{}{"value": cpuFreq}, now) y, err := lp.New("cpufreq", t.tagSet, map[string]interface{}{"value": cpuFreq}, now)
if err == nil { if err == nil {
*out = append(*out, y) *out = append(*out, y)
} }