cc-metric-collector/sinks/natsSink.go
Thomas Gruber 7840de7b82
Merge develop branch into main (#123)
* Add cpu_used (all-cpu_idle) to CpustatCollector

* Update cc-metric-collector.init

* Allow selection of timestamp precision in HttpSink

* Add comment about precision requirement for cc-metric-store

* Fix for API changes in gofish@v0.15.0

* Update requirements to latest version

* Read sensors through redfish

* Update golang toolchain to 1.21

* Remove stray error check

* Update main config in configuration.md

* Update Release action to use golang 1.22 stable release, no golang RPMs anymore

* Update runonce action to use golang 1.22 stable release, no golang RPMs anymore

* Update README.md

Use right JSON type in configuration

* Update sink's README

* Test whether ipmitool or ipmi-sensors can be executed without errors

* Little fixes to the prometheus sink (#115)

* Add uint64 to float64 cast option

* Add prometheus sink to the list of available sinks

* Add aggregated counters by gpu for nvlink errors

---------

Co-authored-by: Michael Schwarz <schwarz@uni-paderborn.de>

* Ccmessage migration (#119)

* Add cpu_used (all-cpu_idle) to CpustatCollector

* Update cc-metric-collector.init

* Allow selection of timestamp precision in HttpSink

* Add comment about precision requirement for cc-metric-store

* Fix for API changes in gofish@v0.15.0

* Update requirements to latest version

* Read sensors through redfish

* Update golang toolchain to 1.21

* Remove stray error check

* Update main config in configuration.md

* Update Release action to use golang 1.22 stable release, no golang RPMs anymore

* Update runonce action to use golang 1.22 stable release, no golang RPMs anymore

* Switch to CCMessage for all files.

---------

Co-authored-by: Holger Obermaier <Holger.Obermaier@kit.edu>
Co-authored-by: Holger Obermaier <40787752+ho-ob@users.noreply.github.com>

* Switch to ccmessage also for latest additions in nvidiaMetric

* New Message processor (#118)

* Add cpu_used (all-cpu_idle) to CpustatCollector

* Update cc-metric-collector.init

* Allow selection of timestamp precision in HttpSink

* Add comment about precision requirement for cc-metric-store

* Fix for API changes in gofish@v0.15.0

* Update requirements to latest version

* Read sensors through redfish

* Update golang toolchain to 1.21

* Remove stray error check

* Update main config in configuration.md

* Update Release action to use golang 1.22 stable release, no golang RPMs anymore

* Update runonce action to use golang 1.22 stable release, no golang RPMs anymore

* New message processor to check whether a message should be dropped or manipulate it in flight

* Create a copy of message before manipulation

---------

Co-authored-by: Holger Obermaier <Holger.Obermaier@kit.edu>
Co-authored-by: Holger Obermaier <40787752+ho-ob@users.noreply.github.com>

* Update collector's Makefile and go.mod/sum files

* Use message processor in router, all sinks and all receivers

* Add support for credential file (NKEY) to NATS sink and receiver

* Fix JSON keys in message processor configuration

* Update docs for message processor, router and the default router config file

* Add link to expr syntax and fix regex matching docs

* Update sample collectors

* Minor style change in collector manager

* Some helpers for ccTopology

* LIKWID collector: write log owner change only once

* Fix for metrics without units and reduce debugging messages for messageProcessor

* Use shorted hostname for hostname added by router

* Define default port for NATS

* CPUstat collector: only add unit for applicable metrics

* Add precision option to all sinks using Influx's encoder

* Add message processor to all sink documentation

* Add units to documentation of cpustat collector

---------

Co-authored-by: Holger Obermaier <Holger.Obermaier@kit.edu>
Co-authored-by: Holger Obermaier <40787752+ho-ob@users.noreply.github.com>
Co-authored-by: oscarminus <me@oscarminus.de>
Co-authored-by: Michael Schwarz <schwarz@uni-paderborn.de>
2024-12-19 23:00:14 +01:00

240 lines
5.6 KiB
Go

package sinks
import (
"bytes"
"encoding/json"
"errors"
"fmt"
"os"
"sync"
"time"
lp "github.com/ClusterCockpit/cc-energy-manager/pkg/cc-message"
cclog "github.com/ClusterCockpit/cc-metric-collector/pkg/ccLogger"
mp "github.com/ClusterCockpit/cc-metric-collector/pkg/messageProcessor"
influx "github.com/influxdata/line-protocol/v2/lineprotocol"
nats "github.com/nats-io/nats.go"
"golang.org/x/exp/slices"
)
type NatsSinkConfig struct {
defaultSinkConfig
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"`
NkeyFile string `json:"nkey_file,omitempty"`
// Timestamp precision
Precision string `json:"precision,omitempty"`
}
type NatsSink struct {
sink
client *nats.Conn
encoder influx.Encoder
buffer *bytes.Buffer
config NatsSinkConfig
lock sync.Mutex
flushDelay time.Duration
flushTimer *time.Timer
extended_tag_list []key_value_pair
}
func (s *NatsSink) connect() error {
var err error
var uinfo nats.Option = nil
var nc *nats.Conn
if len(s.config.User) > 0 && len(s.config.Password) > 0 {
uinfo = nats.UserInfo(s.config.User, s.config.Password)
} else if len(s.config.NkeyFile) > 0 {
if _, err := os.Stat(s.config.NkeyFile); err == nil {
uinfo = nats.UserCredentials(s.config.NkeyFile)
} else {
cclog.ComponentError(s.name, "NKEY file", s.config.NkeyFile, "does not exist: %v", err.Error())
return err
}
}
uri := fmt.Sprintf("nats://%s:%s", s.config.Host, s.config.Port)
cclog.ComponentDebug(s.name, "Connect to", uri)
s.client = nil
if uinfo != nil {
nc, err = nats.Connect(uri, uinfo)
} else {
nc, err = nats.Connect(uri)
}
if err != nil {
cclog.ComponentError(s.name, "Connect to", uri, "failed:", err.Error())
return err
}
s.client = nc
return nil
}
func (s *NatsSink) Write(m lp.CCMessage) error {
msg, err := s.mp.ProcessMessage(m)
if err == nil && msg != nil {
s.lock.Lock()
// Encode measurement name
s.encoder.StartLine(msg.Name())
// copy tags and meta data which should be used as tags
s.extended_tag_list = s.extended_tag_list[:0]
for key, value := range m.Tags() {
s.extended_tag_list =
append(
s.extended_tag_list,
key_value_pair{
key: key,
value: value,
},
)
}
// Encode tags (they musts be in lexical order)
slices.SortFunc(
s.extended_tag_list,
func(a key_value_pair, b key_value_pair) int {
if a.key < b.key {
return -1
}
if a.key > b.key {
return +1
}
return 0
},
)
for i := range s.extended_tag_list {
s.encoder.AddTag(
s.extended_tag_list[i].key,
s.extended_tag_list[i].value,
)
}
// Encode fields
for key, value := range msg.Fields() {
s.encoder.AddField(key, influx.MustNewValue(value))
}
// Encode time stamp
s.encoder.EndLine(msg.Time())
// Check for encoder errors
err := s.encoder.Err()
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 {
s.lock.Lock()
buf := slices.Clone(s.encoder.Bytes())
s.encoder.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() {
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
s.config.Port = "4222"
s.config.Precision = "s"
if len(config) > 0 {
d := json.NewDecoder(bytes.NewReader(config))
d.DisallowUnknownFields()
if err := d.Decode(&s.config); err != nil {
cclog.ComponentError(s.name, "Error reading config:", err.Error())
return nil, err
}
}
if len(s.config.Host) == 0 ||
len(s.config.Port) == 0 ||
len(s.config.Subject) == 0 {
return nil, errors.New("not all configuration variables set required by NatsSink")
}
p, err := mp.NewMessageProcessor()
if err != nil {
return nil, fmt.Errorf("initialization of message processor failed: %v", err.Error())
}
s.mp = p
if len(s.config.MessageProcessor) > 0 {
err = s.mp.FromConfigJSON(s.config.MessageProcessor)
if err != nil {
return nil, fmt.Errorf("failed parsing JSON for message processor: %v", err.Error())
}
}
// Create lookup map to use meta infos as tags in the output metric
for _, k := range s.config.MetaAsTags {
s.mp.AddMoveMetaToTags("true", k, k)
}
precision := influx.Second
if len(s.config.Precision) > 0 {
switch s.config.Precision {
case "s":
precision = influx.Second
case "ms":
precision = influx.Millisecond
case "us":
precision = influx.Microsecond
case "ns":
precision = influx.Nanosecond
}
}
// s.meta_as_tags = make(map[string]bool)
// for _, k := range s.config.MetaAsTags {
// s.meta_as_tags[k] = true
// }
// Setup Influx line protocol
s.encoder.SetPrecision(precision)
// Setup infos for connection
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
}
}
s.extended_tag_list = make([]key_value_pair, 0)
return s, nil
}