ClusterCockpit/cc-metric-collector
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-metric-collector.git synced 2025-04-06 13:35:55 +02:00
Code Issues Packages Projects Releases Wiki Activity

LIKWID collector: write log owner change only once

Browse Source
This commit is contained in:
Thomas Roehl 2024-12-19 14:29:49 +01:00
parent 8270d93b67
commit 2f6f8c846a
1 changed files with 22 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
collectors/likwidMetric.go
View File

@ -43,7 +43,7 @@ const (
type LikwidCollectorMetricConfig struct { type LikwidCollectorMetricConfig struct {
Name string `json:"name"` // Name of the metric Name string `json:"name"` // Name of the metric
Calc string `json:"calc"` // Calculation for the metric using Calc string `json:"calc"` // Calculation for the metric using
Type string `json:"type"` // Metric type (aka node, socket, cpu, ...) Type string `json:"type"` // Metric type (aka node, socket, hwthread, ...)
Publish bool `json:"publish"` Publish bool `json:"publish"`
SendCoreTotalVal bool `json:"send_core_total_values,omitempty"` SendCoreTotalVal bool `json:"send_core_total_values,omitempty"`
SendSocketTotalVal bool `json:"send_socket_total_values,omitempty"` SendSocketTotalVal bool `json:"send_socket_total_values,omitempty"`
@ -91,6 +91,8 @@ type LikwidCollector struct {
running bool running bool
initialized bool initialized bool
needs_reinit bool needs_reinit bool
myuid int
lock_err_once bool
likwidGroups map[C.int]LikwidEventsetConfig likwidGroups map[C.int]LikwidEventsetConfig
lock sync.Mutex lock sync.Mutex
measureThread thread.Thread measureThread thread.Thread
@ -204,6 +206,7 @@ func (m *LikwidCollector) Init(config json.RawMessage) error {
m.initialized = false m.initialized = false
m.needs_reinit = true m.needs_reinit = true
m.running = false m.running = false
m.myuid = os.Getuid()
m.config.AccessMode = LIKWID_DEF_ACCESSMODE m.config.AccessMode = LIKWID_DEF_ACCESSMODE
m.config.LibraryPath = LIKWID_LIB_NAME m.config.LibraryPath = LIKWID_LIB_NAME
m.config.LockfilePath = LIKWID_DEF_LOCKFILE m.config.LockfilePath = LIKWID_DEF_LOCKFILE
@ -390,14 +393,24 @@ func (m *LikwidCollector) takeMeasurement(evidx int, evset LikwidEventsetConfig,
} }
// Check file ownership // Check file ownership
uid := info.Sys().(*syscall.Stat_t).Uid uid := info.Sys().(*syscall.Stat_t).Uid
if uid != uint32(os.Getuid()) { if uid != uint32(m.myuid) {
usr, err := user.LookupId(fmt.Sprint(uid)) usr, err := user.LookupId(fmt.Sprint(uid))
if err == nil { if err == nil {
return true, fmt.Errorf("access to performance counters locked by %s", usr.Username) err = fmt.Errorf("access to performance counters locked by %s", usr.Username)
} else { } else {
return true, fmt.Errorf("access to performance counters locked by %d", uid) err = fmt.Errorf("access to performance counters locked by %d", uid)
} }
// delete error if we already returned the error once.
if !m.lock_err_once {
m.lock_err_once = true
} else {
err = nil
}
return true, err
} }
// reset lock_err_once
m.lock_err_once = false
// Add the lock file to the watcher // Add the lock file to the watcher
err = watcher.Add(m.config.LockfilePath) err = watcher.Add(m.config.LockfilePath)
if err != nil { if err != nil {
@ -798,11 +811,13 @@ func (m *LikwidCollector) ReadThread(interval time.Duration, output chan lp.CCMe
if !skip { if !skip {
// read measurements and derive event set metrics // read measurements and derive event set metrics
m.calcEventsetMetrics(e, interval, output) m.calcEventsetMetrics(e, interval, output)
groups = append(groups, e)
} }
groups = append(groups, e)
} }
// calculate global metrics if len(groups) > 0 {
m.calcGlobalMetrics(groups, interval, output) // calculate global metrics
m.calcGlobalMetrics(groups, interval, output)
}
} }
// main read function taking multiple measurement rounds, each 'interval' seconds long // main read function taking multiple measurement rounds, each 'interval' seconds long