Merge latest development changes to main branch (#79)

* Cleanup: Remove unused code

* Use Golang duration parser for 'interval' and 'duration'
 in main config

* Update handling of LIKWID headers. Download only if not already present in the system. Fixes #73

* Units with cc-units (#64)

* Add option to normalize units with cc-unit

* Add unit conversion to router

* Add option to change unit prefix in the router

* Add to MetricRouter README

* Add order of operations in router to README

* Use second add_tags/del_tags only if metric gets renamed

* Skip disks in DiskstatCollector that have size=0

* Check readability of sensor files in TempCollector

* Fix for --once option

* Rename `cpu` type to `hwthread` (#69)

* Rename 'cpu' type to 'hwthread' to avoid naming clashes with MetricStore and CC-Webfrontend

* Collectors in parallel (#74)

* Provide info to CollectorManager whether the collector can be executed in parallel with others

* Split serial and parallel collectors. Read in parallel first

* Update NvidiaCollector with new metrics, MIG and NvLink support (#75)

* CC topology module update (#76)

* Rename CPU to hardware thread, write some comments

* Do renaming in other parts

* Remove CpuList and SocketList function from metricCollector. Available in ccTopology

* Option to use MIG UUID as subtype-id in NvidiaCollector

* Option to use MIG slice name as subtype-id in NvidiaCollector

* MetricRouter: Fix JSON in README

* Fix for Github Action to really use the selected version

* Remove Ganglia installation in runonce Action and add Go 1.18

* Fix daemon options in init script

* Add separate go.mod files to use it with deprecated 1.16

* Minor updates for Makefiles

* fix string comparison

* AMD ROCm SMI collector (#77)

* Add collector for AMD ROCm SMI metrics

* Fix import path

* Fix imports

* Remove Board Number

* store GPU index explicitly

* Remove board number from description

* Use http instead of ftp to download likwid

* Fix serial number in rocmCollector

* Improved http sink (#78)

* automatic flush in NatsSink

* tweak default options of HttpSink

* shorter cirt. section and retries for HttpSink

* fix error handling

* Remove file added by mistake.

* Use http instead of ftp to download likwid

* Fix serial number in rocmCollector

Co-authored-by: Thomas Roehl <thomas.roehl@fau.de>

Co-authored-by: Holger Obermaier <40787752+ho-ob@users.noreply.github.com>
Co-authored-by: Lou <lou.knauer@gmx.de>
This commit is contained in:
Thomas Gruber
2022-06-08 15:25:40 +02:00
committed by GitHub
parent 186a62a86b
commit 8d85bd53f1
51 changed files with 2097 additions and 705 deletions

View File

@@ -19,9 +19,9 @@ type HttpSinkConfig struct {
URL string `json:"url,omitempty"`
JWT string `json:"jwt,omitempty"`
Timeout string `json:"timeout,omitempty"`
MaxIdleConns int `json:"max_idle_connections,omitempty"`
IdleConnTimeout string `json:"idle_connection_timeout,omitempty"`
FlushDelay string `json:"flush_delay,omitempty"`
MaxRetries int `json:"max_retries,omitempty"`
}
type HttpSink struct {
@@ -32,83 +32,85 @@ type HttpSink struct {
buffer *bytes.Buffer
flushTimer *time.Timer
config HttpSinkConfig
maxIdleConns int
idleConnTimeout time.Duration
timeout time.Duration
flushDelay time.Duration
}
func (s *HttpSink) Write(m lp.CCMetric) error {
if s.buffer.Len() == 0 && s.flushDelay != 0 {
// This is the first write since the last flush, start the flushTimer!
if s.flushTimer != nil && s.flushTimer.Stop() {
cclog.ComponentDebug(s.name, "unexpected: the flushTimer was already running?")
}
// Run a batched flush for all lines that have arrived in the last second
s.flushTimer = time.AfterFunc(s.flushDelay, func() {
if err := s.Flush(); err != nil {
cclog.ComponentError(s.name, "flush failed:", err.Error())
}
})
}
p := m.ToPoint(s.meta_as_tags)
s.lock.Lock()
firstWriteOfBatch := s.buffer.Len() == 0
_, err := s.encoder.Encode(p)
s.lock.Unlock() // defer does not work here as Flush() takes the lock as well
s.lock.Unlock()
if err != nil {
cclog.ComponentError(s.name, "encoding failed:", err.Error())
return err
}
// Flush synchronously if "flush_delay" is zero
if s.flushDelay == 0 {
return s.Flush()
}
return err
if firstWriteOfBatch {
if s.flushTimer == nil {
s.flushTimer = time.AfterFunc(s.flushDelay, func() {
if err := s.Flush(); err != nil {
cclog.ComponentError(s.name, "flush failed:", err.Error())
}
})
} else {
s.flushTimer.Reset(s.flushDelay)
}
}
return nil
}
func (s *HttpSink) Flush() error {
// buffer is read by client.Do, prevent concurrent modifications
// Own lock for as short as possible: the time it takes to copy the buffer.
s.lock.Lock()
defer s.lock.Unlock()
// Do not flush empty buffer
if s.buffer.Len() == 0 {
buf := make([]byte, s.buffer.Len())
copy(buf, s.buffer.Bytes())
s.buffer.Reset()
s.lock.Unlock()
if len(buf) == 0 {
return nil
}
// Create new request to send buffer
req, err := http.NewRequest(http.MethodPost, s.config.URL, s.buffer)
if err != nil {
cclog.ComponentError(s.name, "failed to create request:", err.Error())
return err
var res *http.Response
for i := 0; i < s.config.MaxRetries; i++ {
// Create new request to send buffer
req, err := http.NewRequest(http.MethodPost, s.config.URL, bytes.NewReader(buf))
if err != nil {
cclog.ComponentError(s.name, "failed to create request:", err.Error())
return err
}
// Set authorization header
if len(s.config.JWT) != 0 {
req.Header.Set("Authorization", fmt.Sprintf("Bearer %s", s.config.JWT))
}
// Do request
res, err = s.client.Do(req)
if err != nil {
cclog.ComponentError(s.name, "transport/tcp error:", err.Error())
// Wait between retries
time.Sleep(time.Duration(i+1) * (time.Second / 2))
continue
}
break
}
// Set authorization header
if len(s.config.JWT) != 0 {
req.Header.Set("Authorization", fmt.Sprintf("Bearer %s", s.config.JWT))
}
// Send
res, err := s.client.Do(req)
// Clear buffer
s.buffer.Reset()
// Handle transport/tcp errors
if err != nil {
cclog.ComponentError(s.name, "transport/tcp error:", err.Error())
return err
if res == nil {
return errors.New("flush failed due to repeated errors")
}
// Handle application errors
if res.StatusCode != http.StatusOK {
err = errors.New(res.Status)
err := errors.New(res.Status)
cclog.ComponentError(s.name, "application error:", err.Error())
return err
}
@@ -128,10 +130,10 @@ func NewHttpSink(name string, config json.RawMessage) (Sink, error) {
s := new(HttpSink)
// Set default values
s.name = fmt.Sprintf("HttpSink(%s)", name)
s.config.MaxIdleConns = 10
s.config.IdleConnTimeout = "5s"
s.config.IdleConnTimeout = "120s" // should be larger than the measurement interval.
s.config.Timeout = "5s"
s.config.FlushDelay = "1s"
s.config.FlushDelay = "5s"
s.config.MaxRetries = 3
// Read config
if len(config) > 0 {
@@ -143,9 +145,6 @@ func NewHttpSink(name string, config json.RawMessage) (Sink, error) {
if len(s.config.URL) == 0 {
return nil, errors.New("`url` config option is required for HTTP sink")
}
if s.config.MaxIdleConns > 0 {
s.maxIdleConns = s.config.MaxIdleConns
}
if len(s.config.IdleConnTimeout) > 0 {
t, err := time.ParseDuration(s.config.IdleConnTimeout)
if err == nil {
@@ -170,7 +169,7 @@ func NewHttpSink(name string, config json.RawMessage) (Sink, error) {
s.meta_as_tags[k] = true
}
tr := &http.Transport{
MaxIdleConns: s.maxIdleConns,
MaxIdleConns: 1, // We will only ever talk to one host.
IdleConnTimeout: s.idleConnTimeout,
}
s.client = &http.Client{Transport: tr, Timeout: s.timeout}

View File

@@ -25,7 +25,6 @@ type InfluxAsyncSinkConfig struct {
Password string `json:"password,omitempty"`
Organization string `json:"organization,omitempty"`
SSL bool `json:"ssl,omitempty"`
RetentionPol string `json:"retention_policy,omitempty"`
// Maximum number of points sent to server in single request. Default 5000
BatchSize uint `json:"batch_size,omitempty"`
// Interval, in ms, in which is buffer flushed if it has not been already written (by reaching batch size) . Default 1000ms
@@ -186,12 +185,17 @@ func NewInfluxAsyncSink(name string, config json.RawMessage) (Sink, error) {
return nil, err
}
}
if len(s.config.Host) == 0 ||
len(s.config.Port) == 0 ||
len(s.config.Database) == 0 ||
len(s.config.Organization) == 0 ||
len(s.config.Password) == 0 {
return nil, errors.New("not all configuration variables set required by InfluxAsyncSink")
if len(s.config.Port) == 0 {
return nil, errors.New("Missing port configuration required by InfluxSink")
}
if len(s.config.Database) == 0 {
return nil, errors.New("Missing database configuration required by InfluxSink")
}
if len(s.config.Organization) == 0 {
return nil, errors.New("Missing organization configuration required by InfluxSink")
}
if len(s.config.Password) == 0 {
return nil, errors.New("Missing password configuration required by InfluxSink")
}
// Create lookup map to use meta infos as tags in the output metric
s.meta_as_tags = make(map[string]bool)

View File

@@ -16,37 +16,28 @@ import (
"github.com/influxdata/influxdb-client-go/v2/api/write"
)
type InfluxSinkConfig struct {
defaultSinkConfig
Host string `json:"host,omitempty"`
Port string `json:"port,omitempty"`
Database string `json:"database,omitempty"`
User string `json:"user,omitempty"`
Password string `json:"password,omitempty"`
Organization string `json:"organization,omitempty"`
SSL bool `json:"ssl,omitempty"`
FlushDelay string `json:"flush_delay,omitempty"`
BatchSize int `json:"batch_size,omitempty"`
RetentionPol string `json:"retention_policy,omitempty"`
// InfluxRetryInterval string `json:"retry_interval"`
// InfluxExponentialBase uint `json:"retry_exponential_base"`
// InfluxMaxRetries uint `json:"max_retries"`
// InfluxMaxRetryTime string `json:"max_retry_time"`
//InfluxMaxRetryDelay string `json:"max_retry_delay"` // It is mentioned in the docs but there is no way to set it
}
type InfluxSink struct {
sink
client influxdb2.Client
writeApi influxdb2Api.WriteAPIBlocking
config InfluxSinkConfig
influxRetryInterval uint
influxMaxRetryTime uint
batch []*write.Point
flushTimer *time.Timer
flushDelay time.Duration
lock sync.Mutex // Flush() runs in another goroutine, so this lock has to protect the buffer
//influxMaxRetryDelay uint
client influxdb2.Client
writeApi influxdb2Api.WriteAPIBlocking
config struct {
defaultSinkConfig
Host string `json:"host,omitempty"`
Port string `json:"port,omitempty"`
Database string `json:"database,omitempty"`
User string `json:"user,omitempty"`
Password string `json:"password,omitempty"`
Organization string `json:"organization,omitempty"`
SSL bool `json:"ssl,omitempty"`
// Maximum number of points sent to server in single request. Default 100
BatchSize int `json:"batch_size,omitempty"`
// Interval, in which is buffer flushed if it has not been already written (by reaching batch size). Default 1s
FlushInterval string `json:"flush_delay,omitempty"`
}
batch []*write.Point
flushTimer *time.Timer
flushDelay time.Duration
lock sync.Mutex // Flush() runs in another goroutine, so this lock has to protect the buffer
}
// connect connects to the InfluxDB server
@@ -76,23 +67,6 @@ func (s *InfluxSink) connect() error {
// Set influxDB client options
clientOptions := influxdb2.DefaultOptions()
// if s.influxRetryInterval != 0 {
// cclog.ComponentDebug(s.name, "MaxRetryInterval", s.influxRetryInterval)
// clientOptions.SetMaxRetryInterval(s.influxRetryInterval)
// }
// if s.influxMaxRetryTime != 0 {
// cclog.ComponentDebug(s.name, "MaxRetryTime", s.influxMaxRetryTime)
// clientOptions.SetMaxRetryTime(s.influxMaxRetryTime)
// }
// if s.config.InfluxExponentialBase != 0 {
// cclog.ComponentDebug(s.name, "Exponential Base", s.config.InfluxExponentialBase)
// clientOptions.SetExponentialBase(s.config.InfluxExponentialBase)
// }
// if s.config.InfluxMaxRetries != 0 {
// cclog.ComponentDebug(s.name, "Max Retries", s.config.InfluxMaxRetries)
// clientOptions.SetMaxRetries(s.config.InfluxMaxRetries)
// }
// Do not check InfluxDB certificate
clientOptions.SetTLSConfig(
&tls.Config{
@@ -126,11 +100,13 @@ func (s *InfluxSink) Write(m lp.CCMetric) error {
}
// Run a batched flush for all lines that have arrived in the last flush delay interval
s.flushTimer = time.AfterFunc(s.flushDelay, func() {
if err := s.Flush(); err != nil {
cclog.ComponentError(s.name, "flush failed:", err.Error())
}
})
s.flushTimer = time.AfterFunc(
s.flushDelay,
func() {
if err := s.Flush(); err != nil {
cclog.ComponentError(s.name, "flush failed:", err.Error())
}
})
}
// Append metric to batch slice
@@ -194,7 +170,7 @@ func NewInfluxSink(name string, config json.RawMessage) (Sink, error) {
// Set config default values
s.config.BatchSize = 100
s.config.FlushDelay = "1s"
s.config.FlushInterval = "1s"
// Read config
if len(config) > 0 {
@@ -203,12 +179,6 @@ func NewInfluxSink(name string, config json.RawMessage) (Sink, error) {
return nil, err
}
}
s.influxRetryInterval = 0
s.influxMaxRetryTime = 0
// s.config.InfluxRetryInterval = ""
// s.config.InfluxMaxRetryTime = ""
// s.config.InfluxMaxRetries = 0
// s.config.InfluxExponentialBase = 0
if len(s.config.Host) == 0 {
return nil, errors.New("Missing host configuration required by InfluxSink")
@@ -232,21 +202,9 @@ func NewInfluxSink(name string, config json.RawMessage) (Sink, error) {
s.meta_as_tags[k] = true
}
// toUint := func(duration string, def uint) uint {
// if len(duration) > 0 {
// t, err := time.ParseDuration(duration)
// if err == nil {
// return uint(t.Milliseconds())
// }
// }
// return def
// }
// s.influxRetryInterval = toUint(s.config.InfluxRetryInterval, s.influxRetryInterval)
// s.influxMaxRetryTime = toUint(s.config.InfluxMaxRetryTime, s.influxMaxRetryTime)
// Configure flush delay duration
if len(s.config.FlushDelay) > 0 {
t, err := time.ParseDuration(s.config.FlushDelay)
if len(s.config.FlushInterval) > 0 {
t, err := time.ParseDuration(s.config.FlushInterval)
if err == nil {
s.flushDelay = t
}

View File

@@ -5,6 +5,7 @@ import (
"encoding/json"
"errors"
"fmt"
"sync"
"time"
cclog "github.com/ClusterCockpit/cc-metric-collector/internal/ccLogger"
@@ -15,11 +16,12 @@ import (
type NatsSinkConfig struct {
defaultSinkConfig
Host string `json:"host,omitempty"`
Port string `json:"port,omitempty"`
Database string `json:"database,omitempty"`
User string `json:"user,omitempty"`
Password string `json:"password,omitempty"`
Host string `json:"host,omitempty"`
Port string `json:"port,omitempty"`
Subject string `json:"subject,omitempty"`
User string `json:"user,omitempty"`
Password string `json:"password,omitempty"`
FlushDelay string `json:"flush_delay,omitempty"`
}
type NatsSink struct {
@@ -28,6 +30,10 @@ type NatsSink struct {
encoder *influx.Encoder
buffer *bytes.Buffer
config NatsSinkConfig
lock sync.Mutex
flushDelay time.Duration
flushTimer *time.Timer
}
func (s *NatsSink) connect() error {
@@ -54,37 +60,53 @@ func (s *NatsSink) connect() error {
}
func (s *NatsSink) Write(m lp.CCMetric) error {
if s.client != nil {
_, err := s.encoder.Encode(m.ToPoint(s.meta_as_tags))
if err != nil {
cclog.ComponentError(s.name, "Write:", err.Error())
return err
}
s.lock.Lock()
_, err := s.encoder.Encode(m.ToPoint(s.meta_as_tags))
s.lock.Unlock()
if err != nil {
cclog.ComponentError(s.name, "Write:", err.Error())
return err
}
if s.flushDelay == 0 {
s.Flush()
} else if s.flushTimer == nil {
s.flushTimer = time.AfterFunc(s.flushDelay, func() {
s.Flush()
})
} else {
s.flushTimer.Reset(s.flushDelay)
}
return nil
}
func (s *NatsSink) Flush() error {
if s.client != nil {
if err := s.client.Publish(s.config.Database, s.buffer.Bytes()); err != nil {
cclog.ComponentError(s.name, "Flush:", err.Error())
return err
}
s.buffer.Reset()
s.lock.Lock()
buf := append([]byte{}, s.buffer.Bytes()...) // copy bytes
s.buffer.Reset()
s.lock.Unlock()
if len(buf) == 0 {
return nil
}
if err := s.client.Publish(s.config.Subject, buf); err != nil {
cclog.ComponentError(s.name, "Flush:", err.Error())
return err
}
return nil
}
func (s *NatsSink) Close() {
if s.client != nil {
cclog.ComponentDebug(s.name, "Close")
s.client.Close()
}
cclog.ComponentDebug(s.name, "Close")
s.client.Close()
}
func NewNatsSink(name string, config json.RawMessage) (Sink, error) {
s := new(NatsSink)
s.name = fmt.Sprintf("NatsSink(%s)", name)
s.flushDelay = 10 * time.Second
if len(config) > 0 {
err := json.Unmarshal(config, &s.config)
if err != nil {
@@ -94,7 +116,7 @@ func NewNatsSink(name string, config json.RawMessage) (Sink, error) {
}
if len(s.config.Host) == 0 ||
len(s.config.Port) == 0 ||
len(s.config.Database) == 0 {
len(s.config.Subject) == 0 {
return nil, errors.New("not all configuration variables set required by NatsSink")
}
// Create lookup map to use meta infos as tags in the output metric
@@ -112,5 +134,15 @@ func NewNatsSink(name string, config json.RawMessage) (Sink, error) {
if err := s.connect(); err != nil {
return nil, fmt.Errorf("unable to connect: %v", err)
}
s.flushTimer = nil
if len(s.config.FlushDelay) != 0 {
var err error
s.flushDelay, err = time.ParseDuration(s.config.FlushDelay)
if err != nil {
return nil, err
}
}
return s, nil
}