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

Merge branch 'main' of github.com:ClusterCockpit/cc-metric-collector into main

Browse Source
This commit is contained in:
Thomas Roehl 2022-06-08 15:26:05 +02:00
commit e22c3287e9
57 changed files with 2580 additions and 771 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
.github/ci-config.json vendored
View File

@ -3,6 +3,6 @@
"collectors" : ".github/ci-collectors.json",
"receivers" : ".github/ci-receivers.json",
"router" : ".github/ci-router.json",
"interval": 5,
"duration": 1
"interval": "5s",
"duration": "1s"
}

4
.github/ci-sinks.json vendored
View File

@ -1,6 +1,8 @@
{
"testoutput" : {
"type" : "stdout",
"meta_as_tags" : true
"meta_as_tags" : [
"unit"
]
}
}

42
.github/workflows/runonce.yml vendored
View File

@ -7,6 +7,32 @@ name: Run Test
on: push
jobs:
#
# Job build-1-18
# Build on latest Ubuntu using golang version 1.18
#
build-1-18:
runs-on: ubuntu-latest
steps:
# See: https://github.com/marketplace/actions/checkout
# Checkout git repository and submodules
- name: Checkout
uses: actions/checkout@v2
with:
submodules: recursive
# See: https://github.com/marketplace/actions/setup-go-environment
- name: Setup Golang
uses: actions/setup-go@v3
with:
go-version: '1.18.2'
- name: Build MetricCollector
run: make
- name: Run MetricCollector once
run: ./cc-metric-collector --once --config .github/ci-config.json
#
# Job build-1-17
# Build on latest Ubuntu using golang version 1.17
@ -23,13 +49,9 @@ jobs:
# See: https://github.com/marketplace/actions/setup-go-environment
- name: Setup Golang
uses: actions/setup-go@v2
uses: actions/setup-go@v3
with:
go-version: '^1.17.7'
# Install libganglia
- name: Setup Ganglia
run: sudo apt install ganglia-monitor libganglia1
go-version: '1.17.7'
- name: Build MetricCollector
run: make
@ -53,13 +75,9 @@ jobs:
# See: https://github.com/marketplace/actions/setup-go-environment
- name: Setup Golang
uses: actions/setup-go@v2
uses: actions/setup-go@v3
with:
go-version: '^1.16.7' # The version AlmaLinux 8.5 uses
# Install libganglia
- name: Setup Ganglia
run: sudo apt install ganglia-monitor libganglia1
go-version: '1.16.7' # The version AlmaLinux 8.5 uses
- name: Build MetricCollector
run: make

28
Makefile
View File

@ -16,15 +16,23 @@ COMPONENT_DIRS := collectors \
internal/multiChanTicker
BINDIR = bin
GOBIN = $(shell which go)
.PHONY: all
all: $(APP)
$(APP): $(GOSRC)
if [ "$(shell $(GOBIN) version | cut -d' ' -f 3 | cut -d '.' -f1-2)" = "go1.16" ]; then \
echo "1.16"; \
cp go.mod.1.16 go.mod; \
else \
echo "1.17+"; \
cp go.mod.1.17+ go.mod; \
fi
make -C collectors
go get
go build -o $(APP) $(GOSRC_APP)
$(GOBIN) get
$(GOBIN) build -o $(APP) $(GOSRC_APP)
install: $(APP)
@WORKSPACE=$(PREFIX)
@ -51,25 +59,25 @@ clean:
.PHONY: fmt
fmt:
go fmt $(GOSRC_COLLECTORS)
go fmt $(GOSRC_SINKS)
go fmt $(GOSRC_RECEIVERS)
go fmt $(GOSRC_APP)
@for F in $(GOSRC_INTERNAL); do go fmt $$F; done
$(GOBIN) fmt $(GOSRC_COLLECTORS)
$(GOBIN) fmt $(GOSRC_SINKS)
$(GOBIN) fmt $(GOSRC_RECEIVERS)
$(GOBIN) fmt $(GOSRC_APP)
@for F in $(GOSRC_INTERNAL); do $(GOBIN) fmt $$F; done
# Examine Go source code and reports suspicious constructs
.PHONY: vet
vet:
go vet ./...
$(GOBIN) vet ./...
# Run linter for the Go programming language.
# Using static analysis, it finds bugs and performance issues, offers simplifications, and enforces style rules
.PHONY: staticcheck
staticcheck:
go install honnef.co/go/tools/cmd/staticcheck@latest
$$(go env GOPATH)/bin/staticcheck ./...
$(GOBIN) install honnef.co/go/tools/cmd/staticcheck@latest
$$($(GOBIN) env GOPATH)/bin/staticcheck ./...
.ONESHELL:
.PHONY: RPM

34
cc-metric-collector.go
View File

@ -22,8 +22,8 @@ import (
)
type CentralConfigFile struct {
Interval int `json:"interval"`
Duration int `json:"duration"`
Interval string `json:"interval"`
Duration string `json:"duration"`
CollectorConfigFile string `json:"collectors"`
RouterConfigFile string `json:"router"`
SinkConfigFile string `json:"sinks"`
@ -173,16 +173,36 @@ func mainFunc() int {
cclog.Error("Error reading configuration file ", rcfg.CliArgs["configfile"], ": ", err.Error())
return 1
}
if rcfg.ConfigFile.Interval <= 0 || time.Duration(rcfg.ConfigFile.Interval)*time.Second <= 0 {
// Properly use duration parser with inputs like '60s', '5m' or similar
if len(rcfg.ConfigFile.Interval) > 0 {
t, err := time.ParseDuration(rcfg.ConfigFile.Interval)
if err != nil {
cclog.Error("Configuration value 'interval' no valid duration")
}
rcfg.Interval = t
if rcfg.Interval == 0 {
cclog.Error("Configuration value 'interval' must be greater than zero")
return 1
}
rcfg.Interval = time.Duration(rcfg.ConfigFile.Interval) * time.Second
if rcfg.ConfigFile.Duration <= 0 || time.Duration(rcfg.ConfigFile.Duration)*time.Second <= 0 {
}
// Properly use duration parser with inputs like '60s', '5m' or similar
if len(rcfg.ConfigFile.Duration) > 0 {
t, err := time.ParseDuration(rcfg.ConfigFile.Duration)
if err != nil {
cclog.Error("Configuration value 'duration' no valid duration")
}
rcfg.Duration = t
if rcfg.Duration == 0 {
cclog.Error("Configuration value 'duration' must be greater than zero")
return 1
}
rcfg.Duration = time.Duration(rcfg.ConfigFile.Duration) * time.Second
}
if rcfg.Duration > rcfg.Interval {
cclog.Error("The interval should be greater than duration")
return 1
}
if len(rcfg.ConfigFile.RouterConfigFile) == 0 {
cclog.Error("Metric router configuration file must be set")
@ -271,7 +291,7 @@ func mainFunc() int {
// Wait until one tick has passed. This is a workaround
if rcfg.CliArgs["once"] == "true" {
x := 1.2 * float64(rcfg.ConfigFile.Interval)
x := 1.2 * float64(rcfg.Interval.Seconds())
time.Sleep(time.Duration(int(x)) * time.Second)
shutdownSignal <- os.Interrupt
}

36
collectors/Makefile
View File

@ -1,22 +1,28 @@
all: likwid
# LIKWID version
LIKWID_VERSION = 5.2.1
LIKWID_INSTALLED_FOLDER=$(shell dirname $(shell which likwid-topology 2>/dev/null) 2>/dev/null)
LIKWID_FOLDER="$(shell pwd)/likwid"
all: $(LIKWID_FOLDER)/likwid.h
.ONESHELL:
.PHONY: likwid
likwid:
INSTALL_FOLDER="$${PWD}/likwid"
BUILD_FOLDER="$${PWD}/likwidbuild"
if [ -d $${INSTALL_FOLDER} ]; then rm -r $${INSTALL_FOLDER}; fi
mkdir --parents --verbose $${INSTALL_FOLDER} $${BUILD_FOLDER}
wget -P "$${BUILD_FOLDER}" ftp://ftp.rrze.uni-erlangen.de/mirrors/likwid/likwid-$(LIKWID_VERSION).tar.gz
tar -C $${BUILD_FOLDER} -xf $${BUILD_FOLDER}/likwid-$(LIKWID_VERSION).tar.gz
install -Dpm 0644 $${BUILD_FOLDER}/likwid-$(LIKWID_VERSION)/src/includes/likwid*.h $${INSTALL_FOLDER}/
install -Dpm 0644 $${BUILD_FOLDER}/likwid-$(LIKWID_VERSION)/src/includes/bstrlib.h $${INSTALL_FOLDER}/
rm -r $${BUILD_FOLDER}
.PHONY: $(LIKWID_FOLDER)/likwid.h
$(LIKWID_FOLDER)/likwid.h:
if [ "$(LIKWID_INSTALLED_FOLDER)" != "" ]; then \
BASE="$(LIKWID_INSTALLED_FOLDER)/../include"; \
mkdir -p $(LIKWID_FOLDER); \
cp $$BASE/*.h $(LIKWID_FOLDER); \
else \
BUILD_FOLDER="$${PWD}/likwidbuild"; \
if [ -d $(LIKWID_FOLDER) ]; then rm -r $(LIKWID_FOLDER); fi; \
mkdir --parents --verbose $(LIKWID_FOLDER) $${BUILD_FOLDER}; \
wget -P "$${BUILD_FOLDER}" http://ftp.rrze.uni-erlangen.de/mirrors/likwid/likwid-$(LIKWID_VERSION).tar.gz; \
tar -C $${BUILD_FOLDER} -xf $${BUILD_FOLDER}/likwid-$(LIKWID_VERSION).tar.gz; \
install -Dpm 0644 $${BUILD_FOLDER}/likwid-$(LIKWID_VERSION)/src/includes/likwid*.h $(LIKWID_FOLDER)/; \
install -Dpm 0644 $${BUILD_FOLDER}/likwid-$(LIKWID_VERSION)/src/includes/bstrlib.h $(LIKWID_FOLDER)/; \
rm -r $${BUILD_FOLDER}; \
fi
clean:

1
collectors/README.md
View File

@ -39,6 +39,7 @@ In contrast to the configuration files for sinks and receivers, the collectors c
* [`gpfs`](./gpfsMetric.md)
* [`beegfs_meta`](./beegfsmetaMetric.md)
* [`beegfs_storage`](./beegfsstorageMetric.md)
* [`rocm_smi`](./rocmsmiMetric.md)
## Todos

1
collectors/beegfsmetaMetric.go
View File

@ -55,6 +55,7 @@ func (m *BeegfsMetaCollector) Init(config json.RawMessage) error {
m.name = "BeegfsMetaCollector"
m.setup()
m.parallel = true
// Set default beegfs-ctl binary
m.config.Beegfs = DEFAULT_BEEGFS_CMD

1
collectors/beegfsstorageMetric.go
View File

@ -48,6 +48,7 @@ func (m *BeegfsStorageCollector) Init(config json.RawMessage) error {
m.name = "BeegfsStorageCollector"
m.setup()
m.parallel = true
// Set default beegfs-ctl binary
m.config.Beegfs = DEFAULT_BEEGFS_CMD

34
collectors/collectorManager.go
View File

@ -36,17 +36,21 @@ var AvailableCollectors = map[string]MetricCollector{
"numastats": new(NUMAStatsCollector),
"beegfs_meta": new(BeegfsMetaCollector),
"beegfs_storage": new(BeegfsStorageCollector),
"rocm_smi": new(RocmSmiCollector),
}
// Metric collector manager data structure
type collectorManager struct {
collectors []MetricCollector // List of metric collectors to use
collectors []MetricCollector // List of metric collectors to read in parallel
serial []MetricCollector // List of metric collectors to read serially
output chan lp.CCMetric // Output channels
done chan bool // channel to finish / stop metric collector manager
ticker mct.MultiChanTicker // periodically ticking once each interval
duration time.Duration // duration (for metrics that measure over a given duration)
wg *sync.WaitGroup // wait group for all goroutines in cc-metric-collector
config map[string]json.RawMessage // json encoded config for collector manager
collector_wg sync.WaitGroup // internally used wait group for the parallel reading of collector
parallel_run bool // Flag whether the collectors are currently read in parallel
}
// Metric collector manager access functions
@ -66,6 +70,7 @@ type CollectorManager interface {
// Initialization is done for all configured collectors
func (cm *collectorManager) Init(ticker mct.MultiChanTicker, duration time.Duration, wg *sync.WaitGroup, collectConfigFile string) error {
cm.collectors = make([]MetricCollector, 0)
cm.serial = make([]MetricCollector, 0)
cm.output = nil
cm.done = make(chan bool)
cm.wg = wg
@ -100,7 +105,11 @@ func (cm *collectorManager) Init(ticker mct.MultiChanTicker, duration time.Durat
continue
}
cclog.ComponentDebug("CollectorManager", "ADD COLLECTOR", collector.Name())
if collector.Parallel() {
cm.collectors = append(cm.collectors, collector)
} else {
cm.serial = append(cm.serial, collector)
}
}
return nil
}
@ -116,6 +125,10 @@ func (cm *collectorManager) Start() {
// Collector manager is done
done := func() {
// close all metric collectors
if cm.parallel_run {
cm.collector_wg.Wait()
cm.parallel_run = false
}
for _, c := range cm.collectors {
c.Close()
}
@ -130,7 +143,26 @@ func (cm *collectorManager) Start() {
done()
return
case t := <-tick:
cm.parallel_run = true
for _, c := range cm.collectors {
// Wait for done signal or execute the collector
select {
case <-cm.done:
done()
return
default:
// Read metrics from collector c via goroutine
cclog.ComponentDebug("CollectorManager", c.Name(), t)
cm.collector_wg.Add(1)
go func(myc MetricCollector) {
myc.Read(cm.duration, cm.output)
cm.collector_wg.Done()
}(c)
}
}
cm.collector_wg.Wait()
cm.parallel_run = false
for _, c := range cm.serial {
// Wait for done signal or execute the collector
select {
case <-cm.done:

3
collectors/cpufreqCpuinfoMetric.go
View File

@ -48,6 +48,7 @@ func (m *CPUFreqCpuInfoCollector) Init(config json.RawMessage) error {
m.setup()
m.name = "CPUFreqCpuInfoCollector"
m.parallel = true
m.meta = map[string]string{
"source": m.name,
"group": "CPU",
@ -150,7 +151,7 @@ func (m *CPUFreqCpuInfoCollector) Init(config json.RawMessage) error {
t.numNonHT = numNonHT
t.numNonHT_int = numNonHT_int
t.tagSet = map[string]string{
"type": "cpu",
"type": "hwthread",
"type-id": t.processor,
"package_id": t.physicalPackageID,
}

2
collectors/cpufreqCpuinfoMetric.md
View File

@ -4,7 +4,7 @@
"cpufreq_cpuinfo": {}
```
The `cpufreq_cpuinfo` collector reads the clock frequency from `/proc/cpuinfo` and outputs a handful **cpu** metrics.
The `cpufreq_cpuinfo` collector reads the clock frequency from `/proc/cpuinfo` and outputs a handful **hwthread** metrics.
Metrics:
* `cpufreq`

3
collectors/cpufreqMetric.go
View File

@ -53,6 +53,7 @@ func (m *CPUFreqCollector) Init(config json.RawMessage) error {
m.name = "CPUFreqCollector"
m.setup()
m.parallel = true
if len(config) > 0 {
err := json.Unmarshal(config, &m.config)
if err != nil {
@ -161,7 +162,7 @@ func (m *CPUFreqCollector) Init(config json.RawMessage) error {
t.numNonHT = numNonHT
t.numNonHT_int = numNonHT_int
t.tagSet = map[string]string{
"type": "cpu",
"type": "hwthread",
"type-id": t.processor,
"package_id": t.physicalPackageID,
}

2
collectors/cpufreqMetric.md
View File

@ -5,7 +5,7 @@
}
```
The `cpufreq` collector reads the clock frequency from `/sys/devices/system/cpu/cpu*/cpufreq` and outputs a handful **cpu** metrics.
The `cpufreq` collector reads the clock frequency from `/sys/devices/system/cpu/cpu*/cpufreq` and outputs a handful **hwthread** metrics.
Metrics:
* `cpufreq`

3
collectors/cpustatMetric.go
View File

@ -30,6 +30,7 @@ type CpustatCollector struct {
func (m *CpustatCollector) Init(config json.RawMessage) error {
m.name = "CpustatCollector"
m.setup()
m.parallel = true
m.meta = map[string]string{"source": m.name, "group": "CPU", "unit": "Percent"}
m.nodetags = map[string]string{"type": "node"}
if len(config) > 0 {
@ -82,7 +83,7 @@ func (m *CpustatCollector) Init(config json.RawMessage) error {
if strings.HasPrefix(linefields[0], "cpu") && strings.Compare(linefields[0], "cpu") != 0 {
cpustr := strings.TrimLeft(linefields[0], "cpu")
cpu, _ := strconv.Atoi(cpustr)
m.cputags[linefields[0]] = map[string]string{"type": "cpu", "type-id": fmt.Sprintf("%d", cpu)}
m.cputags[linefields[0]] = map[string]string{"type": "hwthread", "type-id": fmt.Sprintf("%d", cpu)}
num_cpus++
}
}

1
collectors/customCmdMetric.go
View File

@ -33,6 +33,7 @@ type CustomCmdCollector struct {
func (m *CustomCmdCollector) Init(config json.RawMessage) error {
var err error
m.name = "CustomCmdCollector"
m.parallel = true
m.meta = map[string]string{"source": m.name, "group": "Custom"}
if len(config) > 0 {
err = json.Unmarshal(config, &m.config)

16
collectors/diskstatMetric.go
View File

@ -3,7 +3,6 @@ package collectors
import (
"bufio"
"encoding/json"
"fmt"
"os"
"strings"
"syscall"
@ -30,6 +29,7 @@ type DiskstatCollector struct {
func (m *DiskstatCollector) Init(config json.RawMessage) error {
m.name = "DiskstatCollector"
m.parallel = true
m.meta = map[string]string{"source": m.name, "group": "Disk"}
m.setup()
if len(config) > 0 {
@ -78,11 +78,17 @@ func (m *DiskstatCollector) Read(interval time.Duration, output chan lp.CCMetric
continue
}
path := strings.Replace(linefields[1], `\040`, " ", -1)
stat := syscall.Statfs_t{}
stat := syscall.Statfs_t{
Blocks: 0,
Bsize: 0,
Bfree: 0,
}
err := syscall.Statfs(path, &stat)
if err != nil {
fmt.Println(err.Error())
return
continue
}
if stat.Blocks == 0 || stat.Bsize == 0 {
continue
}
tags := map[string]string{"type": "node", "device": linefields[0]}
total := (stat.Blocks * uint64(stat.Bsize)) / uint64(1000000000)
@ -97,11 +103,13 @@ func (m *DiskstatCollector) Read(interval time.Duration, output chan lp.CCMetric
y.AddMeta("unit", "GBytes")
output <- y
}
if total > 0 {
perc := (100 * (total - free)) / total
if perc > part_max_used {
part_max_used = perc
}
}
}
y, err := lp.New("part_max_used", map[string]string{"type": "node"}, m.meta, map[string]interface{}{"value": int(part_max_used)}, time.Now())
if err == nil {
y.AddMeta("unit", "percent")

7
collectors/gpfsMetric.go
View File

@ -46,6 +46,7 @@ func (m *GpfsCollector) Init(config json.RawMessage) error {
var err error
m.name = "GpfsCollector"
m.setup()
m.parallel = true
// Set default mmpmon binary
m.config.Mmpmon = DEFAULT_GPFS_CMD
@ -70,6 +71,7 @@ func (m *GpfsCollector) Init(config json.RawMessage) error {
for _, fs := range m.config.ExcludeFilesystem {
m.skipFS[fs] = struct{}{}
}
m.lastState = make(map[string]GpfsCollectorLastState)
// GPFS / IBM Spectrum Scale file system statistics can only be queried by user root
user, err := user.Current()
@ -162,13 +164,18 @@ func (m *GpfsCollector) Read(interval time.Duration, output chan lp.CCMetric) {
continue
}
// Add filesystem tag
m.tags["filesystem"] = filesystem
// Create initial last state
if m.config.SendBandwidths {
if _, ok := m.lastState[filesystem]; !ok {
m.lastState[filesystem] = GpfsCollectorLastState{
bytesRead: -1,
bytesWritten: -1,
}
}
}
// return code
rc, err := strconv.Atoi(key_value["_rc_"])

1
collectors/infinibandMetric.go
View File

@ -54,6 +54,7 @@ func (m *InfinibandCollector) Init(config json.RawMessage) error {
var err error
m.name = "InfinibandCollector"
m.setup()
m.parallel = true
m.meta = map[string]string{
"source": m.name,
"group": "Network",

1
collectors/iostatMetric.go
View File

@ -37,6 +37,7 @@ type IOstatCollector struct {
func (m *IOstatCollector) Init(config json.RawMessage) error {
var err error
m.name = "IOstatCollector"
m.parallel = true
m.meta = map[string]string{"source": m.name, "group": "Disk"}
m.setup()
if len(config) > 0 {

1
collectors/ipmiMetric.go
View File

@ -34,6 +34,7 @@ type IpmiCollector struct {
func (m *IpmiCollector) Init(config json.RawMessage) error {
m.name = "IpmiCollector"
m.setup()
m.parallel = true
m.meta = map[string]string{"source": m.name, "group": "IPMI"}
m.config.IpmitoolPath = string(IPMITOOL_PATH)
m.config.IpmisensorsPath = string(IPMISENSORS_PATH)

3
collectors/likwidMetric.go
View File

@ -177,6 +177,7 @@ func getBaseFreq() float64 {
func (m *LikwidCollector) Init(config json.RawMessage) error {
m.name = "LikwidCollector"
m.parallel = false
m.initialized = false
m.running = false
m.config.AccessMode = LIKWID_DEF_ACCESSMODE
@ -204,7 +205,7 @@ func (m *LikwidCollector) Init(config json.RawMessage) error {
m.meta = map[string]string{"group": "PerfCounter"}
cclog.ComponentDebug(m.name, "Get cpulist and init maps and lists")
cpulist := topo.CpuList()
cpulist := topo.HwthreadList()
m.cpulist = make([]C.int, len(cpulist))
m.cpu2tid = make(map[int]int)
for i, c := range cpulist {

85
collectors/likwidMetric.md
View File

@ -3,32 +3,63 @@
The `likwid` collector is probably the most complicated collector. The LIKWID library is included as static library with *direct* access mode. The *direct* access mode is suitable if the daemon is executed by a root user. The static library does not contain the performance groups, so all information needs to be provided in the configuration.
The `likwid` configuration consists of two parts, the "eventsets" and "globalmetrics":
- An event set list itself has two parts, the "events" and a set of derivable "metrics". Each of the "events" is a counter:event pair in LIKWID's syntax. The "metrics" are a list of formulas to derive the metric value from the measurements of the "events". Each metric has a name, the formula, a scope and a publish flag. Counter names can be used like variables in the formulas, so `PMC0+PMC1` sums the measurements for the both events configured in the counters `PMC0` and `PMC1`. The scope tells the Collector whether it is a metric for each hardware thread (`cpu`) or each CPU socket (`socket`). You may specify a unit for the metric with `unit`. The last one is the publishing flag. It tells the collector whether a metric should be sent to the router.
- The global metrics are metrics which require data from all event set measurements to be derived. The inputs are the metrics in the event sets. Similar to the metrics in the event sets, the global metrics are defined by a name, a formula, a scope and a publish flag. See event set metrics for details. The only difference is that there is no access to the raw event measurements anymore but only to the metrics. So, the idea is to derive a metric in the "eventsets" section and reuse it in the "globalmetrics" part. If you need a metric only for deriving the global metrics, disable forwarding of the event set metrics (`publish=false`). **Be aware** that the combination might be misleading because the "behavior" of a metric changes over time and the multiple measurements might count different computing phases. Similar to the metrics in the eventset, you can specify a metric unit with the `unit` field.
```json
"likwid": {
"force_overwrite" : false,
"invalid_to_zero" : false,
"eventsets": [
{
"events" : {
"COUNTER0": "EVENT0",
"COUNTER1": "EVENT1",
},
"metrics" : [
{
"name": "sum_01",
"calc": "COUNTER0 + COUNTER1",
"publish": false,
"unit": "myunit",
"type": "hwthread"
}
]
}
]
"globalmetrics" : [
{
"name": "global_sum",
"calc": "sum_01",
"publish": true,
"unit": "myunit",
"type": "hwthread"
}
]
}
```
The `likwid` configuration consists of two parts, the `eventsets` and `globalmetrics`:
- An event set list itself has two parts, the `events` and a set of derivable `metrics`. Each of the `events` is a `counter:event` pair in LIKWID's syntax. The `metrics` are a list of formulas to derive the metric value from the measurements of the `events`' values. Each metric has a name, the formula, a type and a publish flag. There is an optional `unit` field. Counter names can be used like variables in the formulas, so `PMC0+PMC1` sums the measurements for the both events configured in the counters `PMC0` and `PMC1`. You can optionally use `time` for the measurement time and `inverseClock` for `1.0/baseCpuFrequency`. The type tells the LikwidCollector whether it is a metric for each hardware thread (`cpu`) or each CPU socket (`socket`). You may specify a unit for the metric with `unit`. The last one is the publishing flag. It tells the LikwidCollector whether a metric should be sent to the router or is only used internally to compute a global metric.
- The `globalmetrics` are metrics which require data from multiple event set measurements to be derived. The inputs are the metrics in the event sets. Similar to the metrics in the event sets, the global metrics are defined by a name, a formula, a scope and a publish flag. See event set metrics for details. The only difference is that there is no access to the raw event measurements anymore but only to the metrics. Also `time` and `inverseClock` cannot be used anymore. So, the idea is to derive a metric in the `eventsets` section and reuse it in the `globalmetrics` part. If you need a metric only for deriving the global metrics, disable forwarding of the event set metrics (`"publish": false`). **Be aware** that the combination might be misleading because the "behavior" of a metric changes over time and the multiple measurements might count different computing phases. Similar to the metrics in the eventset, you can specify a metric unit with the `unit` field.
Additional options:
- `access_mode` : Method to use for hardware performance monitoring (`direct` access as root user, `accessdaemon` for the daemon mode)
- `accessdaemon_path`: Folder with the access daemon `likwid-accessD`, commonly `$LIKWID_INSTALL_LOC/sbin`
- `force_overwrite`: Same as setting `LIKWID_FORCE=1`. In case counters are already in-use, LIKWID overwrites their configuration to do its measurements
- `invalid_to_zero`: In some cases, the calculations result in `NaN` or `Inf`. With this option, all `NaN` and `Inf` values are replaces with `0.0`.
- `access_mode`: Specify LIKWID access mode: `direct` for direct register access as root user or `accessdaemon`
- `accessdaemon_path`: Folder of the accessDaemon `likwid-accessD`
- `liblikwid_path`: Location of `liblikwid.so`
- `invalid_to_zero`: In some cases, the calculations result in `NaN` or `Inf`. With this option, all `NaN` and `Inf` values are replaces with `0.0`. See below in [seperate section](./likwidMetric.md#invalid_to_zero-option)
- `access_mode`: Specify LIKWID access mode: `direct` for direct register access as root user or `accessdaemon`. The access mode `perf_event` is current untested.
- `accessdaemon_path`: Folder of the accessDaemon `likwid-accessD` (like `/usr/local/sbin`)
- `liblikwid_path`: Location of `liblikwid.so` including file name like `/usr/local/lib/liblikwid.so`
### Available metric scopes
Hardware performance counters are scattered all over the system nowadays. A counter coveres a specific part of the system. While there are hardware thread specific counter for CPU cycles, instructions and so on, some others are specific for a whole CPU socket/package. To address that, the collector provides the specification of a 'scope' for each metric.
Hardware performance counters are scattered all over the system nowadays. A counter coveres a specific part of the system. While there are hardware thread specific counter for CPU cycles, instructions and so on, some others are specific for a whole CPU socket/package. To address that, the LikwidCollector provides the specification of a `type` for each metric.
- `cpu` : One metric per CPU hardware thread with the tags `"type" : "cpu"` and `"type-id" : "$cpu_id"`
- `hwthread` : One metric per CPU hardware thread with the tags `"type" : "hwthread"` and `"type-id" : "$hwthread_id"`
- `socket` : One metric per CPU socket/package with the tags `"type" : "socket"` and `"type-id" : "$socket_id"`
**Note:** You cannot specify `socket` scope for a metric that is measured at `cpu` scope, so some kind of expert knowledge or lookup work in the [Likwid Wiki](https://github.com/RRZE-HPC/likwid/wiki) is required. Get the scope of each counter from the *Architecture* pages and as soon as one counter in a metric is socket-specific, the whole metric is socket-specific.
**Note:** You cannot specify `socket` scope for a metric that is measured at `hwthread` scope, so some kind of expert knowledge or lookup work in the [Likwid Wiki](https://github.com/RRZE-HPC/likwid/wiki) is required. Get the scope of each counter from the *Architecture* pages and as soon as one counter in a metric is socket-specific, the whole metric is socket-specific.
As a guideline:
- All counters `FIXCx`, `PMCy` and `TMAz` have the scope `cpu`
- All counters `FIXCx`, `PMCy` and `TMAz` have the scope `hwthread`
- All counters names containing `BOX` have the scope `socket`
- All `PWRx` counters have scope `socket`, except `"PWR1" : "RAPL_CORE_ENERGY"` has `cpu` scope
- All `PWRx` counters have scope `socket`, except `"PWR1" : "RAPL_CORE_ENERGY"` has `hwthread` scope
- All `DFCx` counters have scope `socket`
### Help with the configuration
@ -50,6 +81,7 @@ $ scripts/likwid_perfgroup_to_cc_config.py ICX MEM_DP
{
"events": {
"FIXC0": "INSTR_RETIRED_ANY",
"FIXC1": "CPU_CLK_UNHALTED_CORE",
"..." : "..."
},
"metrics" : [
@ -58,7 +90,7 @@ $ scripts/likwid_perfgroup_to_cc_config.py ICX MEM_DP
"name": "Runtime (RDTSC) [s]",
"publish": true,
"unit": "seconds"
"scope": "cpu"
"scope": "hwthread"
},
{
"..." : "..."
@ -75,21 +107,28 @@ LIKWID checks the file `/var/run/likwid.lock` before performing any interfering
Before (SLURM prolog, ...)
```
$ chwon $JOBUSER /var/run/likwid.lock
$ chown $JOBUSER /var/run/likwid.lock
```
After (SLURM epilog, ...)
```
$ chwon $CCUSER /var/run/likwid.lock
$ chown $CCUSER /var/run/likwid.lock
```
### `invalid_to_zero` option
In some cases LIKWID returns `0.0` for some events that are further used in processing and maybe used as divisor in a calculation. After evaluation of a metric, the result might be `NaN` or `+-Inf`. These resulting metrics are commonly not created and forwarded to the router because the [InfluxDB line protocol](https://docs.influxdata.com/influxdb/cloud/reference/syntax/line-protocol/#float) does not support these special floating-point values. If you want to have them sent, this option forces these metric values to be `0.0` instead.
One might think this does not happen often but often used metrics in the world of performance engineering like Instructions-per-Cycle (IPC) or more frequently the actual CPU clock are derived with events like `CPU_CLK_UNHALTED_CORE` (Intel) which do not increment in halted state (as the name implies). In there are different power management systems in a chip which can cause a hardware thread to go in such a state. Moreover, if no cycles are executed by the core, also many other events are not incremented as well (like `INSTR_RETIRED_ANY` for retired instructions and part of IPC).
### Example configuration
#### AMD Zen3
```json
"likwid": {
"force_overwrite" : false,
"nan_to_zero" : false,
"invalid_to_zero" : false,
"eventsets": [
{
"events": {
@ -108,20 +147,20 @@ $ chwon $CCUSER /var/run/likwid.lock
{
"name": "ipc",
"calc": "PMC0/PMC1",
"type": "cpu",
"type": "hwthread",
"publish": true
},
{
"name": "flops_any",
"calc": "0.000001*PMC2/time",
"unit": "MFlops/s",
"type": "cpu",
"type": "hwthread",
"publish": true
},
{
"name": "clock",
"calc": "0.000001*(FIXC1/FIXC2)/inverseClock",
"type": "cpu",
"type": "hwthread",
"unit": "MHz",
"publish": true
},
@ -182,7 +221,7 @@ $ chwon $CCUSER /var/run/likwid.lock
### How to get the eventsets and metrics from LIKWID
The `likwid` collector reads hardware performance counters at a **cpu** and **socket** level. The configuration looks quite complicated but it is basically copy&paste from [LIKWID's performance groups](https://github.com/RRZE-HPC/likwid/tree/master/groups). The collector made multiple iterations and tried to use the performance groups but it lacked flexibility. The current way of configuration provides most flexibility.
The `likwid` collector reads hardware performance counters at a **hwthread** and **socket** level. The configuration looks quite complicated but it is basically copy&paste from [LIKWID's performance groups](https://github.com/RRZE-HPC/likwid/tree/master/groups). The collector made multiple iterations and tried to use the performance groups but it lacked flexibility. The current way of configuration provides most flexibility.
The logic is as following: There are multiple eventsets, each consisting of a list of counters+events and a list of metrics. If you compare a common performance group with the example setting above, there is not much difference:
```
@ -203,7 +242,7 @@ METRICS -> "metrics": [
IPC PMC0/PMC1 -> {
-> "name" : "IPC",
-> "calc" : "PMC0/PMC1",
-> "scope": "cpu",
-> "scope": "hwthread",
-> "publish": true
-> }
-> ]

1
collectors/loadavgMetric.go
View File

@ -36,6 +36,7 @@ type LoadavgCollector struct {
func (m *LoadavgCollector) Init(config json.RawMessage) error {
m.name = "LoadavgCollector"
m.parallel = true
m.setup()
if len(config) > 0 {
err := json.Unmarshal(config, &m.config)

1
collectors/lustreMetric.go
View File

@ -288,6 +288,7 @@ var LustreDeriveMetrics = []LustreMetricDefinition{
func (m *LustreCollector) Init(config json.RawMessage) error {
var err error
m.name = "LustreCollector"
m.parallel = true
if len(config) > 0 {
err = json.Unmarshal(config, &m.config)
if err != nil {

2
collectors/memstatMetric.go
View File

@ -81,6 +81,7 @@ func getStats(filename string) map[string]MemstatStats {
func (m *MemstatCollector) Init(config json.RawMessage) error {
var err error
m.name = "MemstatCollector"
m.parallel = true
m.config.NodeStats = true
m.config.NumaStats = false
if len(config) > 0 {
@ -159,6 +160,7 @@ func (m *MemstatCollector) Init(config json.RawMessage) error {
func (m *MemstatCollector) Read(interval time.Duration, output chan lp.CCMetric) {
if !m.init {
cclog.ComponentPrint(m.name, "Here")
return
}

63
collectors/metricCollector.go
View File

@ -3,10 +3,6 @@ package collectors
import (
"encoding/json"
"fmt"
"io/ioutil"
"log"
"strconv"
"strings"
"time"
lp "github.com/ClusterCockpit/cc-metric-collector/internal/ccMetric"
@ -16,6 +12,7 @@ type MetricCollector interface {
Name() string // Name of the metric collector
Init(config json.RawMessage) error // Initialize metric collector
Initialized() bool // Is metric collector initialized?
Parallel() bool
Read(duration time.Duration, output chan lp.CCMetric) // Read metrics from metric collector
Close() // Close / finish metric collector
}
@ -23,6 +20,7 @@ type MetricCollector interface {
type metricCollector struct {
name string // name of the metric
init bool // is metric collector initialized?
parallel bool // can the metric collector be executed in parallel with others
meta map[string]string // static meta data tags
}
@ -31,6 +29,11 @@ func (c *metricCollector) Name() string {
return c.name
}
// Name returns the name of the metric collector
func (c *metricCollector) Parallel() bool {
return c.parallel
}
// Setup is for future use
func (c *metricCollector) setup() error {
return nil
@ -65,58 +68,6 @@ func stringArrayContains(array []string, str string) (int, bool) {
return -1, false
}
// SocketList returns the list of physical sockets as read from /proc/cpuinfo
func SocketList() []int {
buffer, err := ioutil.ReadFile("/proc/cpuinfo")
if err != nil {
log.Print(err)
return nil
}
ll := strings.Split(string(buffer), "\n")
var packs []int
for _, line := range ll {
if strings.HasPrefix(line, "physical id") {
lv := strings.Fields(line)
id, err := strconv.ParseInt(lv[3], 10, 32)
if err != nil {
log.Print(err)
return packs
}
_, found := intArrayContains(packs, int(id))
if !found {
packs = append(packs, int(id))
}
}
}
return packs
}
// CpuList returns the list of physical CPUs (in contrast to logical CPUs) as read from /proc/cpuinfo
func CpuList() []int {
buffer, err := ioutil.ReadFile("/proc/cpuinfo")
if err != nil {
log.Print(err)
return nil
}
ll := strings.Split(string(buffer), "\n")
var cpulist []int
for _, line := range ll {
if strings.HasPrefix(line, "processor") {
lv := strings.Fields(line)
id, err := strconv.ParseInt(lv[2], 10, 32)
if err != nil {
log.Print(err)
return cpulist
}
_, found := intArrayContains(cpulist, int(id))
if !found {
cpulist = append(cpulist, int(id))
}
}
}
return cpulist
}
// RemoveFromStringList removes the string r from the array of strings s
// If r is not contained in the array an error is returned
func RemoveFromStringList(s []string, r string) ([]string, error) {

1
collectors/netstatMetric.go
View File

@ -39,6 +39,7 @@ type NetstatCollector struct {
func (m *NetstatCollector) Init(config json.RawMessage) error {
m.name = "NetstatCollector"
m.parallel = true
m.setup()
m.lastTimestamp = time.Now()

1
collectors/nfsMetric.go
View File

@ -114,6 +114,7 @@ func (m *nfsCollector) MainInit(config json.RawMessage) error {
m.data = make(map[string]NfsCollectorData)
m.initStats()
m.init = true
m.parallel = true
return nil
}

1
collectors/numastatsMetric.go
View File

@ -54,6 +54,7 @@ func (m *NUMAStatsCollector) Init(config json.RawMessage) error {
}
m.name = "NUMAStatsCollector"
m.parallel = true
m.setup()
m.meta = map[string]string{
"source": m.name,

908
collectors/nvidiaMetric.go
View File

File diff suppressed because it is too large Load Diff

66
collectors/nvidiaMetric.md
View File

@ -4,37 +4,73 @@
```json
"nvidia": {
"exclude_devices": [
"0","1"
"0","1", "0000000:ff:01.0"
],
"exclude_metrics": [
"nv_fb_memory",
"nv_fb_mem_used",
"nv_fan"
]
],
"process_mig_devices": false,
"use_pci_info_as_type_id": true,
"add_pci_info_tag": false,
"add_uuid_meta": false,
"add_board_number_meta": false,
"add_serial_meta": false,
"use_uuid_for_mig_device": false,
"use_slice_for_mig_device": false
}
```
The `nvidia` collector can be configured to leave out specific devices with the `exclude_devices` option. It takes IDs as supplied to the NVML with `nvmlDeviceGetHandleByIndex()` or the PCI address in NVML format (`%08X:%02X:%02X.0`). Metrics (listed below) that should not be sent to the MetricRouter can be excluded with the `exclude_metrics` option. Commonly only the physical GPUs are monitored. If MIG devices should be analyzed as well, set `process_mig_devices` (adds `stype=mig,stype-id=<mig_index>`). With the options `use_uuid_for_mig_device` and `use_slice_for_mig_device`, the `<mig_index>` can be replaced with the UUID (e.g. `MIG-6a9f7cc8-6d5b-5ce0-92de-750edc4d8849`) or the MIG slice name (e.g. `1g.5gb`).
The metrics sent by the `nvidia` collector use `accelerator` as `type` tag. For the `type-id`, it uses the device handle index by default. With the `use_pci_info_as_type_id` option, the PCI ID is used instead. If both values should be added as tags, activate the `add_pci_info_tag` option. It uses the device handle index as `type-id` and adds the PCI ID as separate `pci_identifier` tag.
Optionally, it is possible to add the UUID, the board part number and the serial to the meta informations. They are not sent to the sinks (if not configured otherwise).
Metrics:
* `nv_util`
* `nv_mem_util`
* `nv_mem_total`
* `nv_fb_memory`
* `nv_fb_mem_total`
* `nv_fb_mem_used`
* `nv_bar1_mem_total`
* `nv_bar1_mem_used`
* `nv_temp`
* `nv_fan`
* `nv_ecc_mode`
* `nv_perf_state`
* `nv_power_usage_report`
* `nv_graphics_clock_report`
* `nv_sm_clock_report`
* `nv_mem_clock_report`
* `nv_power_usage`
* `nv_graphics_clock`
* `nv_sm_clock`
* `nv_mem_clock`
* `nv_video_clock`
* `nv_max_graphics_clock`
* `nv_max_sm_clock`
* `nv_max_mem_clock`
* `nv_ecc_db_error`
* `nv_ecc_sb_error`
* `nv_power_man_limit`
* `nv_max_video_clock`
* `nv_ecc_uncorrected_error`
* `nv_ecc_corrected_error`
* `nv_power_max_limit`
* `nv_encoder_util`
* `nv_decoder_util`
* `nv_remapped_rows_corrected`
* `nv_remapped_rows_uncorrected`
* `nv_remapped_rows_pending`
* `nv_remapped_rows_failure`
* `nv_compute_processes`
* `nv_graphics_processes`
* `nv_violation_power`
* `nv_violation_thermal`
* `nv_violation_sync_boost`
* `nv_violation_board_limit`
* `nv_violation_low_util`
* `nv_violation_reliability`
* `nv_violation_below_app_clock`
* `nv_violation_below_base_clock`
* `nv_nvlink_crc_flit_errors`
* `nv_nvlink_crc_errors`
* `nv_nvlink_ecc_errors`
* `nv_nvlink_replay_errors`
* `nv_nvlink_recovery_errors`
It uses a separate `type` in the metrics. The output metric looks like this:
`<name>,type=accelerator,type-id=<nvidia-gpu-id> value=<metric value> <timestamp>`
Some metrics add the additional sub type tag (`stype`) like the `nv_nvlink_*` metrics set `stype=nvlink,stype-id=<link_number>`.

319
collectors/rocmsmiMetric.go Normal file
View File

@ -0,0 +1,319 @@
package collectors
import (
"encoding/json"
"errors"
"fmt"
"time"
cclog "github.com/ClusterCockpit/cc-metric-collector/internal/ccLogger"
lp "github.com/ClusterCockpit/cc-metric-collector/internal/ccMetric"
"github.com/ClusterCockpit/go-rocm-smi/pkg/rocm_smi"
)
type RocmSmiCollectorConfig struct {
ExcludeMetrics []string `json:"exclude_metrics,omitempty"`
ExcludeDevices []string `json:"exclude_devices,omitempty"`
AddPciInfoTag bool `json:"add_pci_info_tag,omitempty"`
UsePciInfoAsTypeId bool `json:"use_pci_info_as_type_id,omitempty"`
AddSerialMeta bool `json:"add_serial_meta,omitempty"`
}
type RocmSmiCollectorDevice struct {
device rocm_smi.DeviceHandle
index int
tags map[string]string // default tags
meta map[string]string // default meta information
excludeMetrics map[string]bool // copy of exclude metrics from config
}
type RocmSmiCollector struct {
metricCollector
config RocmSmiCollectorConfig // the configuration structure
devices []RocmSmiCollectorDevice
}
// Functions to implement MetricCollector interface
// Init(...), Read(...), Close()
// See: metricCollector.go
// Init initializes the sample collector
// Called once by the collector manager
// All tags, meta data tags and metrics that do not change over the runtime should be set here
func (m *RocmSmiCollector) Init(config json.RawMessage) error {
var err error = nil
// Always set the name early in Init() to use it in cclog.Component* functions
m.name = "RocmSmiCollector"
// This is for later use, also call it early
m.setup()
// Define meta information sent with each metric
// (Can also be dynamic or this is the basic set with extension through AddMeta())
//m.meta = map[string]string{"source": m.name, "group": "AMD"}
// Define tags sent with each metric
// The 'type' tag is always needed, it defines the granulatity of the metric
// node -> whole system
// socket -> CPU socket (requires socket ID as 'type-id' tag)
// cpu -> single CPU hardware thread (requires cpu ID as 'type-id' tag)
//m.tags = map[string]string{"type": "node"}
// Read in the JSON configuration
if len(config) > 0 {
err = json.Unmarshal(config, &m.config)
if err != nil {
cclog.ComponentError(m.name, "Error reading config:", err.Error())
return err
}
}
ret := rocm_smi.Init()
if ret != rocm_smi.STATUS_SUCCESS {
err = errors.New("Failed to initialize ROCm SMI library")
cclog.ComponentError(m.name, err.Error())
return err
}
numDevs, ret := rocm_smi.NumMonitorDevices()
if ret != rocm_smi.STATUS_SUCCESS {
err = errors.New("Failed to get number of GPUs from ROCm SMI library")
cclog.ComponentError(m.name, err.Error())
return err
}
exclDev := func(s string) bool {
skip_device := false
for _, excl := range m.config.ExcludeDevices {
if excl == s {
skip_device = true
break
}
}
return skip_device
}
m.devices = make([]RocmSmiCollectorDevice, 0)
for i := 0; i < numDevs; i++ {
str_i := fmt.Sprintf("%d", i)
if exclDev(str_i) {
continue
}
device, ret := rocm_smi.DeviceGetHandleByIndex(i)
if ret != rocm_smi.STATUS_SUCCESS {
err = fmt.Errorf("Failed to get handle for GPU %d", i)
cclog.ComponentError(m.name, err.Error())
return err
}
pciInfo, ret := rocm_smi.DeviceGetPciInfo(device)
if ret != rocm_smi.STATUS_SUCCESS {
err = fmt.Errorf("Failed to get PCI information for GPU %d", i)
cclog.ComponentError(m.name, err.Error())
return err
}
pciId := fmt.Sprintf(
"%08X:%02X:%02X.%X",
pciInfo.Domain,
pciInfo.Bus,
pciInfo.Device,
pciInfo.Function)
if exclDev(pciId) {
continue
}
dev := RocmSmiCollectorDevice{
device: device,
tags: map[string]string{
"type": "accelerator",
"type-id": str_i,
},
meta: map[string]string{
"source": m.name,
"group": "AMD",
},
}
if m.config.UsePciInfoAsTypeId {
dev.tags["type-id"] = pciId
} else if m.config.AddPciInfoTag {
dev.tags["pci_identifier"] = pciId
}
if m.config.AddSerialMeta {
serial, ret := rocm_smi.DeviceGetSerialNumber(device)
if ret != rocm_smi.STATUS_SUCCESS {
cclog.ComponentError(m.name, "Unable to get serial number for device at index", i, ":", rocm_smi.StatusStringNoError(ret))
} else {
dev.meta["serial"] = serial
}
}
// Add excluded metrics
dev.excludeMetrics = map[string]bool{}
for _, e := range m.config.ExcludeMetrics {
dev.excludeMetrics[e] = true
}
dev.index = i
m.devices = append(m.devices, dev)
}
// Set this flag only if everything is initialized properly, all required files exist, ...
m.init = true
return err
}
// Read collects all metrics belonging to the sample collector
// and sends them through the output channel to the collector manager
func (m *RocmSmiCollector) Read(interval time.Duration, output chan lp.CCMetric) {
// Create a sample metric
timestamp := time.Now()
for _, dev := range m.devices {
metrics, ret := rocm_smi.DeviceGetMetrics(dev.device)
if ret != rocm_smi.STATUS_SUCCESS {
cclog.ComponentError(m.name, "Unable to get metrics for device at index", dev.index, ":", rocm_smi.StatusStringNoError(ret))
continue
}
if !dev.excludeMetrics["rocm_gfx_util"] {
value := metrics.Average_gfx_activity
y, err := lp.New("rocm_gfx_util", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_umc_util"] {
value := metrics.Average_umc_activity
y, err := lp.New("rocm_umc_util", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_mm_util"] {
value := metrics.Average_mm_activity
y, err := lp.New("rocm_mm_util", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_avg_power"] {
value := metrics.Average_socket_power
y, err := lp.New("rocm_avg_power", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_temp_mem"] {
value := metrics.Temperature_mem
y, err := lp.New("rocm_temp_mem", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_temp_hotspot"] {
value := metrics.Temperature_hotspot
y, err := lp.New("rocm_temp_hotspot", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_temp_edge"] {
value := metrics.Temperature_edge
y, err := lp.New("rocm_temp_edge", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_temp_vrgfx"] {
value := metrics.Temperature_vrgfx
y, err := lp.New("rocm_temp_vrgfx", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_temp_vrsoc"] {
value := metrics.Temperature_vrsoc
y, err := lp.New("rocm_temp_vrsoc", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_temp_vrmem"] {
value := metrics.Temperature_vrmem
y, err := lp.New("rocm_temp_vrmem", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_gfx_clock"] {
value := metrics.Average_gfxclk_frequency
y, err := lp.New("rocm_gfx_clock", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_soc_clock"] {
value := metrics.Average_socclk_frequency
y, err := lp.New("rocm_soc_clock", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_u_clock"] {
value := metrics.Average_uclk_frequency
y, err := lp.New("rocm_u_clock", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_v0_clock"] {
value := metrics.Average_vclk0_frequency
y, err := lp.New("rocm_v0_clock", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_v1_clock"] {
value := metrics.Average_vclk1_frequency
y, err := lp.New("rocm_v1_clock", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_d0_clock"] {
value := metrics.Average_dclk0_frequency
y, err := lp.New("rocm_d0_clock", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_d1_clock"] {
value := metrics.Average_dclk1_frequency
y, err := lp.New("rocm_d1_clock", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
output <- y
}
}
if !dev.excludeMetrics["rocm_temp_hbm"] {
for i := 0; i < rocm_smi.NUM_HBM_INSTANCES; i++ {
value := metrics.Temperature_hbm[i]
y, err := lp.New("rocm_temp_hbm", dev.tags, dev.meta, map[string]interface{}{"value": value}, timestamp)
if err == nil {
y.AddTag("stype", "device")
y.AddTag("stype-id", fmt.Sprintf("%d", i))
output <- y
}
}
}
}
}
// Close metric collector: close network connection, close files, close libraries, ...
// Called once by the collector manager
func (m *RocmSmiCollector) Close() {
// Unset flag
ret := rocm_smi.Shutdown()
if ret != rocm_smi.STATUS_SUCCESS {
cclog.ComponentError(m.name, "Failed to shutdown ROCm SMI library")
}
m.init = false
}

47
collectors/rocmsmiMetric.md Normal file
View File

@ -0,0 +1,47 @@
## `rocm_smi` collector
```json
"rocm_smi": {
"exclude_devices": [
"0","1", "0000000:ff:01.0"
],
"exclude_metrics": [
"rocm_mm_util",
"rocm_temp_vrsoc"
],
"use_pci_info_as_type_id": true,
"add_pci_info_tag": false,
"add_serial_meta": false,
}
```
The `rocm_smi` collector can be configured to leave out specific devices with the `exclude_devices` option. It takes logical IDs in the list of available devices or the PCI address similar to NVML format (`%08X:%02X:%02X.0`). Metrics (listed below) that should not be sent to the MetricRouter can be excluded with the `exclude_metrics` option.
The metrics sent by the `rocm_smi` collector use `accelerator` as `type` tag. For the `type-id`, it uses the device handle index by default. With the `use_pci_info_as_type_id` option, the PCI ID is used instead. If both values should be added as tags, activate the `add_pci_info_tag` option. It uses the device handle index as `type-id` and adds the PCI ID as separate `pci_identifier` tag.
Optionally, it is possible to add the serial to the meta informations. They are not sent to the sinks (if not configured otherwise).
Metrics:
* `rocm_gfx_util`
* `rocm_umc_util`
* `rocm_mm_util`
* `rocm_avg_power`
* `rocm_temp_mem`
* `rocm_temp_hotspot`
* `rocm_temp_edge`
* `rocm_temp_vrgfx`
* `rocm_temp_vrsoc`
* `rocm_temp_vrmem`
* `rocm_gfx_clock`
* `rocm_soc_clock`
* `rocm_u_clock`
* `rocm_v0_clock`
* `rocm_v1_clock`
* `rocm_d0_clock`
* `rocm_d1_clock`
* `rocm_temp_hbm`
Some metrics add the additional sub type tag (`stype`) like the `rocm_temp_hbm` metrics set `stype=device,stype-id=<HBM_slice_number>`.

11
collectors/sampleMetric.go
View File

@ -35,6 +35,10 @@ func (m *SampleCollector) Init(config json.RawMessage) error {
m.name = "InternalCollector"
// This is for later use, also call it early
m.setup()
// Tell whether the collector should be run in parallel with others (reading files, ...)
// or it should be run serially, mostly for collectors acutally doing measurements
// because they should not measure the execution of the other collectors
m.parallel = true
// Define meta information sent with each metric
// (Can also be dynamic or this is the basic set with extension through AddMeta())
m.meta = map[string]string{"source": m.name, "group": "SAMPLE"}
@ -42,7 +46,12 @@ func (m *SampleCollector) Init(config json.RawMessage) error {
// The 'type' tag is always needed, it defines the granulatity of the metric
// node -> whole system
// socket -> CPU socket (requires socket ID as 'type-id' tag)
// cpu -> single CPU hardware thread (requires cpu ID as 'type-id' tag)
// die -> CPU die (requires CPU die ID as 'type-id' tag)
// memoryDomain -> NUMA domain (requires NUMA domain ID as 'type-id' tag)
// llc -> Last level cache (requires last level cache ID as 'type-id' tag)
// core -> single CPU core that may consist of multiple hardware threads (SMT) (requires core ID as 'type-id' tag)
// hwthtread -> single CPU hardware thread (requires hardware thread ID as 'type-id' tag)
// accelerator -> A accelerator device like GPU or FPGA (requires an accelerator ID as 'type-id' tag)
m.tags = map[string]string{"type": "node"}
// Read in the JSON configuration
if len(config) > 0 {

5
collectors/tempMetric.go
View File

@ -50,6 +50,7 @@ func (m *TempCollector) Init(config json.RawMessage) error {
}
m.name = "TempCollector"
m.parallel = true
m.setup()
if len(config) > 0 {
err := json.Unmarshal(config, &m.config)
@ -116,6 +117,10 @@ func (m *TempCollector) Init(config json.RawMessage) error {
}
// Sensor file
_, err = ioutil.ReadFile(file)
if err != nil {
continue
}
sensor.file = file
// Sensor tags

1
collectors/topprocsMetric.go
View File

@ -28,6 +28,7 @@ type TopProcsCollector struct {
func (m *TopProcsCollector) Init(config json.RawMessage) error {
var err error
m.name = "TopProcsCollector"
m.parallel = true
m.tags = map[string]string{"type": "node"}
m.meta = map[string]string{"source": m.name, "group": "TopProcs"}
if len(config) > 0 {

12
config.json
View File

@ -1,8 +1,8 @@
{
"sinks": "sinks.json",
"collectors" : "collectors.json",
"receivers" : "receivers.json",
"router" : "router.json",
"interval": 10,
"duration": 1
"sinks": "./sinks.json",
"collectors" : "./collectors.json",
"receivers" : "./receivers.json",
"router" : "./router.json",
"interval": "10s",
"duration": "1s"
}

42
go.mod
View File

@ -1,19 +1,35 @@
module github.com/ClusterCockpit/cc-metric-collector
go 1.16
require (
github.com/NVIDIA/go-nvml v0.11.1-0
github.com/influxdata/influxdb-client-go/v2 v2.7.0
github.com/influxdata/line-protocol v0.0.0-20210922203350-b1ad95c89adf
github.com/nats-io/nats.go v1.13.1-0.20211122170419-d7c1d78a50fc
golang.org/x/sys v0.0.0-20220114195835-da31bd327af9
gopkg.in/Knetic/govaluate.v2 v2.3.0
)
go 1.17
require (
github.com/ClusterCockpit/cc-units v0.0.0-20220318130935-92a0c6442220
github.com/NVIDIA/go-nvml v0.11.6-0
github.com/PaesslerAG/gval v1.1.2
github.com/golang/protobuf v1.5.2 // indirect
github.com/nats-io/nats-server/v2 v2.7.0 // indirect
google.golang.org/protobuf v1.27.1 // indirect
github.com/gorilla/mux v1.8.0
github.com/influxdata/influxdb-client-go/v2 v2.8.1
github.com/influxdata/line-protocol v0.0.0-20210922203350-b1ad95c89adf
github.com/nats-io/nats.go v1.14.0
github.com/prometheus/client_golang v1.12.1
github.com/stmcginnis/gofish v0.13.0
golang.org/x/sys v0.0.0-20220412211240-33da011f77ad
)
require (
github.com/beorn7/perks v1.0.1 // indirect
github.com/cespare/xxhash/v2 v2.1.2 // indirect
github.com/deepmap/oapi-codegen v1.8.2 // indirect
github.com/golang/protobuf v1.5.2 // indirect
github.com/matttproud/golang_protobuf_extensions v1.0.1 // indirect
github.com/nats-io/nats-server/v2 v2.8.0 // indirect
github.com/nats-io/nkeys v0.3.0 // indirect
github.com/nats-io/nuid v1.0.1 // indirect
github.com/pkg/errors v0.9.1 // indirect
github.com/prometheus/client_model v0.2.0 // indirect
github.com/prometheus/common v0.32.1 // indirect
github.com/prometheus/procfs v0.7.3 // indirect
golang.org/x/crypto v0.0.0-20220112180741-5e0467b6c7ce // indirect
golang.org/x/net v0.0.0-20211112202133-69e39bad7dc2 // indirect
google.golang.org/protobuf v1.26.0 // indirect
gopkg.in/yaml.v2 v2.4.0 // indirect
)

16
go.mod.1.16 Normal file
View File

@ -0,0 +1,16 @@
module github.com/ClusterCockpit/cc-metric-collector
go 1.16
require (
github.com/NVIDIA/go-nvml v0.11.6-0
github.com/PaesslerAG/gval v1.1.2
github.com/gorilla/mux v1.8.0
github.com/influxdata/influxdb-client-go/v2 v2.7.0
github.com/influxdata/line-protocol v0.0.0-20210922203350-b1ad95c89adf
github.com/nats-io/nats-server/v2 v2.8.0 // indirect
github.com/nats-io/nats.go v1.14.0
github.com/prometheus/client_golang v1.12.1
github.com/stmcginnis/gofish v0.13.0
golang.org/x/sys v0.0.0-20220412211240-33da011f77ad
)

16
go.mod.1.17+ Normal file
View File

@ -0,0 +1,16 @@
module github.com/ClusterCockpit/cc-metric-collector
go 1.17
require (
github.com/NVIDIA/go-nvml v0.11.6-0
github.com/PaesslerAG/gval v1.1.2
github.com/gorilla/mux v1.8.0
github.com/influxdata/influxdb-client-go/v2 v2.8.1
github.com/influxdata/line-protocol v0.0.0-20210922203350-b1ad95c89adf
github.com/nats-io/nats-server/v2 v2.8.0 // indirect
github.com/nats-io/nats.go v1.14.0
github.com/prometheus/client_golang v1.12.1
github.com/stmcginnis/gofish v0.13.0
golang.org/x/sys v0.0.0-20220412211240-33da011f77ad
)

108
internal/ccTopology/ccTopology.go
View File

@ -29,6 +29,7 @@ func intArrayContains(array []int, str int) (int, bool) {
return -1, false
}
// Used internally for sysfs file reads
func fileToInt(path string) int {
buffer, err := ioutil.ReadFile(path)
if err != nil {
@ -47,6 +48,7 @@ func fileToInt(path string) int {
return int(id)
}
// Get list of CPU socket IDs
func SocketList() []int {
buffer, err := ioutil.ReadFile(string(PROCFS_CPUINFO))
if err != nil {
@ -54,7 +56,7 @@ func SocketList() []int {
return nil
}
ll := strings.Split(string(buffer), "\n")
var packs []int
packs := make([]int, 0)
for _, line := range ll {
if strings.HasPrefix(line, "physical id") {
lv := strings.Fields(line)
@ -72,7 +74,8 @@ func SocketList() []int {
return packs
}
func CpuList() []int {
// Get list of hardware thread IDs in the order of listing in /proc/cpuinfo
func HwthreadList() []int {
buffer, err := ioutil.ReadFile(string(PROCFS_CPUINFO))
if err != nil {
log.Print(err)
@ -97,6 +100,13 @@ func CpuList() []int {
return cpulist
}
// Get list of hardware thread IDs in the order of listing in /proc/cpuinfo
// Deprecated! Use HwthreadList()
func CpuList() []int {
return HwthreadList()
}
// Get list of CPU core IDs in the order of listing in /proc/cpuinfo
func CoreList() []int {
buffer, err := ioutil.ReadFile(string(PROCFS_CPUINFO))
if err != nil {
@ -122,6 +132,7 @@ func CoreList() []int {
return corelist
}
// Get list of NUMA node IDs
func NumaNodeList() []int {
numaList := make([]int, 0)
globPath := filepath.Join(string(SYSFS_NUMABASE), "node*")
@ -156,8 +167,9 @@ func NumaNodeList() []int {
return numaList
}
// Get list of CPU die IDs
func DieList() []int {
cpulist := CpuList()
cpulist := HwthreadList()
dielist := make([]int, 0)
for _, c := range cpulist {
diepath := filepath.Join(string(SYSFS_CPUBASE), fmt.Sprintf("cpu%d", c), "topology/die_id")
@ -175,7 +187,27 @@ func DieList() []int {
return SocketList()
}
type CpuEntry struct {
// Get list of specified type using the naming format inside ClusterCockpit
func GetTypeList(topology_type string) []int {
switch topology_type {
case "node":
return []int{0}
case "socket":
return SocketList()
case "die":
return DieList()
case "memoryDomain":
return NumaNodeList()
case "core":
return CoreList()
case "hwthread":
return HwthreadList()
}
return []int{}
}
// Structure holding all information about a hardware thread
type HwthreadEntry struct {
Cpuid int
SMT int
Core int
@ -184,25 +216,25 @@ type CpuEntry struct {
Die int
}
func CpuData() []CpuEntry {
func CpuData() []HwthreadEntry {
fileToInt := func(path string) int {
buffer, err := ioutil.ReadFile(path)
if err != nil {
log.Print(err)
//cclogger.ComponentError("ccTopology", "Reading", path, ":", err.Error())
return -1
}
sbuffer := strings.Replace(string(buffer), "\n", "", -1)
var id int64
//_, err = fmt.Scanf("%d", sbuffer, &id)
id, err = strconv.ParseInt(sbuffer, 10, 32)
if err != nil {
cclogger.ComponentError("ccTopology", "Parsing", path, ":", sbuffer, err.Error())
return -1
}
return int(id)
}
// fileToInt := func(path string) int {
// buffer, err := ioutil.ReadFile(path)
// if err != nil {
// log.Print(err)
// //cclogger.ComponentError("ccTopology", "Reading", path, ":", err.Error())
// return -1
// }
// sbuffer := strings.Replace(string(buffer), "\n", "", -1)
// var id int64
// //_, err = fmt.Scanf("%d", sbuffer, &id)
// id, err = strconv.ParseInt(sbuffer, 10, 32)
// if err != nil {
// cclogger.ComponentError("ccTopology", "Parsing", path, ":", sbuffer, err.Error())
// return -1
// }
// return int(id)
// }
getCore := func(basepath string) int {
return fileToInt(fmt.Sprintf("%s/core_id", basepath))
}
@ -260,9 +292,9 @@ func CpuData() []CpuEntry {
return 0
}
clist := make([]CpuEntry, 0)
for _, c := range CpuList() {
clist = append(clist, CpuEntry{Cpuid: c})
clist := make([]HwthreadEntry, 0)
for _, c := range HwthreadList() {
clist = append(clist, HwthreadEntry{Cpuid: c})
}
for i, centry := range clist {
centry.Socket = -1
@ -298,6 +330,7 @@ func CpuData() []CpuEntry {
return clist
}
// Structure holding basic information about a CPU
type CpuInformation struct {
NumHWthreads int
SMTWidth int
@ -307,6 +340,7 @@ type CpuInformation struct {
NumNumaDomains int
}
// Get basic information about the CPU
func CpuInfo() CpuInformation {
var c CpuInformation
@ -342,7 +376,8 @@ func CpuInfo() CpuInformation {
return c
}
func GetCpuSocket(cpuid int) int {
// Get the CPU socket ID for a given hardware thread ID
func GetHwthreadSocket(cpuid int) int {
cdata := CpuData()
for _, d := range cdata {
if d.Cpuid == cpuid {
@ -352,7 +387,8 @@ func GetCpuSocket(cpuid int) int {
return -1
}
func GetCpuNumaDomain(cpuid int) int {
// Get the NUMA node ID for a given hardware thread ID
func GetHwthreadNumaDomain(cpuid int) int {
cdata := CpuData()
for _, d := range cdata {
if d.Cpuid == cpuid {
@ -362,7 +398,8 @@ func GetCpuNumaDomain(cpuid int) int {
return -1
}
func GetCpuDie(cpuid int) int {
// Get the CPU die ID for a given hardware thread ID
func GetHwthreadDie(cpuid int) int {
cdata := CpuData()
for _, d := range cdata {
if d.Cpuid == cpuid {
@ -372,7 +409,8 @@ func GetCpuDie(cpuid int) int {
return -1
}
func GetCpuCore(cpuid int) int {
// Get the CPU core ID for a given hardware thread ID
func GetHwthreadCore(cpuid int) int {
cdata := CpuData()
for _, d := range cdata {
if d.Cpuid == cpuid {
@ -382,7 +420,8 @@ func GetCpuCore(cpuid int) int {
return -1
}
func GetSocketCpus(socket int) []int {
// Get the all hardware thread ID associated with a CPU socket
func GetSocketHwthreads(socket int) []int {
all := CpuData()
cpulist := make([]int, 0)
for _, d := range all {
@ -393,7 +432,8 @@ func GetSocketCpus(socket int) []int {
return cpulist
}
func GetNumaDomainCpus(domain int) []int {
// Get the all hardware thread ID associated with a NUMA node
func GetNumaDomainHwthreads(domain int) []int {
all := CpuData()
cpulist := make([]int, 0)
for _, d := range all {
@ -404,7 +444,8 @@ func GetNumaDomainCpus(domain int) []int {
return cpulist
}
func GetDieCpus(die int) []int {
// Get the all hardware thread ID associated with a CPU die
func GetDieHwthreads(die int) []int {
all := CpuData()
cpulist := make([]int, 0)
for _, d := range all {
@ -415,7 +456,8 @@ func GetDieCpus(die int) []int {
return cpulist
}
func GetCoreCpus(core int) []int {
// Get the all hardware thread ID associated with a CPU core
func GetCoreHwthreads(core int) []int {
all := CpuData()
cpulist := make([]int, 0)
for _, d := range all {

14
internal/metricAggregator/metricAggregatorFunctions.go
View File

@ -246,7 +246,7 @@ func matchfunc(args ...interface{}) (interface{}, error) {
func getCpuCoreFunc(args ...interface{}) (interface{}, error) {
switch cpuid := args[0].(type) {
case int:
return topo.GetCpuCore(cpuid), nil
return topo.GetHwthreadCore(cpuid), nil
}
return -1, errors.New("function 'getCpuCore' accepts only an 'int' cpuid")
}
@ -255,7 +255,7 @@ func getCpuCoreFunc(args ...interface{}) (interface{}, error) {
func getCpuSocketFunc(args ...interface{}) (interface{}, error) {
switch cpuid := args[0].(type) {
case int:
return topo.GetCpuSocket(cpuid), nil
return topo.GetHwthreadSocket(cpuid), nil
}
return -1, errors.New("function 'getCpuCore' accepts only an 'int' cpuid")
}
@ -264,7 +264,7 @@ func getCpuSocketFunc(args ...interface{}) (interface{}, error) {
func getCpuNumaDomainFunc(args ...interface{}) (interface{}, error) {
switch cpuid := args[0].(type) {
case int:
return topo.GetCpuNumaDomain(cpuid), nil
return topo.GetHwthreadNumaDomain(cpuid), nil
}
return -1, errors.New("function 'getCpuNuma' accepts only an 'int' cpuid")
}
@ -273,7 +273,7 @@ func getCpuNumaDomainFunc(args ...interface{}) (interface{}, error) {
func getCpuDieFunc(args ...interface{}) (interface{}, error) {
switch cpuid := args[0].(type) {
case int:
return topo.GetCpuDie(cpuid), nil
return topo.GetHwthreadDie(cpuid), nil
}
return -1, errors.New("function 'getCpuDie' accepts only an 'int' cpuid")
}
@ -336,7 +336,7 @@ func getCpuListOfDieFunc(args ...interface{}) (interface{}, error) {
// wrapper function to get a list of all cpuids of the node
func getCpuListOfNode(args ...interface{}) (interface{}, error) {
return topo.CpuList(), nil
return topo.HwthreadList(), nil
}
// helper function to get the cpuid list for a CCMetric type tag set (type and type-id)
@ -348,14 +348,14 @@ func getCpuListOfType(args ...interface{}) (interface{}, error) {
case string:
switch typ {
case "node":
return topo.CpuList(), nil
return topo.HwthreadList(), nil
case "socket":
return getCpuListOfSocketFunc(args[1])
case "numadomain":
return getCpuListOfNumaDomainFunc(args[1])
case "core":
return getCpuListOfCoreFunc(args[1])
case "cpu":
case "hwthread":
var cpu int
switch id := args[1].(type) {

32
internal/metricRouter/README.md
View File

@ -52,6 +52,11 @@ The CCMetric router sits in between the collectors and the sinks and can be used
],
"rename_metrics" : {
"metric_12345" : "mymetric"
},
"normalize_units" : true,
"change_unit_prefix" : {
"mem_used" : "G",
"mem_total" : "G"
}
}
```
@ -192,6 +197,14 @@ This option takes a list of evaluable conditions and performs them one after the
```
The first line is comparable with the example in `drop_metrics`, it drops all metrics starting with `drop_metric_` and ending with a number. The second line drops all metrics of the first hardware thread (**not** recommended)
# Manipulating the metric units
## The `normalize_units` option
The cc-metric-collector tries to read the data from the system as it is reported. If available, it tries to read the metric unit from the system as well (e.g. from `/proc/meminfo`). The problem is that, depending on the source, the metric units are named differently. Just think about `byte`, `Byte`, `B`, `bytes`, ...
The [cc-units](https://github.com/ClusterCockpit/cc-units) package provides us a normalization option to use the same metric unit name for all metrics. It this option is set to true, all `unit` meta tags are normalized.
## The `change_unit_prefix` section
It is often the case that metrics are reported by the system using a rather outdated unit prefix (like `/proc/meminfo` still uses kByte despite current memory sizes are in the GByte range). If you want to change the prefix of a unit, you can do that with the help of [cc-units](https://github.com/ClusterCockpit/cc-units). The setting works on the metric name and requires the new prefix for the metric. The cc-units package determines the scaling factor.
# Aggregate metric values of the current interval with the `interval_aggregates` option
@ -239,3 +252,22 @@ Use cases for `interval_aggregates`:
}
}
```
# Order of operations
The router performs the above mentioned options in a specific order. In order to get the logic you want for a specific metric, it is crucial to know the processing order:
- Add the `hostname` tag (c)
- Manipulate the timestamp to the interval timestamp (c,r)
- Drop metrics based on `drop_metrics` and `drop_metrics_if` (c,r)
- Add tags based on `add_tags` (c,r)
- Delete tags based on `del_tags` (c,r)
- Rename metric based on `rename_metric` (c,r)
- Add tags based on `add_tags` to still work if the configuration uses the new name (c,r)
- Delete tags based on `del_tags` to still work if the configuration uses the new name (c,r)
- Normalize units when `normalize_units` is set (c,r)
- Convert unit prefix based on `change_unit_prefix` (c,r)
Legend:
- 'c' if metric is coming from a collector
- 'r' if metric is coming from a receiver

39
internal/metricRouter/metricRouter.go
View File

@ -12,6 +12,7 @@ import (
lp "github.com/ClusterCockpit/cc-metric-collector/internal/ccMetric"
agg "github.com/ClusterCockpit/cc-metric-collector/internal/metricAggregator"
mct "github.com/ClusterCockpit/cc-metric-collector/internal/multiChanTicker"
units "github.com/ClusterCockpit/cc-units"
)
const ROUTER_MAX_FORWARD = 50
@ -35,6 +36,8 @@ type metricRouterConfig struct {
IntervalStamp bool `json:"interval_timestamp"` // Update timestamp periodically by ticker each interval?
NumCacheIntervals int `json:"num_cache_intervals"` // Number of intervals of cached metrics for evaluation
MaxForward int `json:"max_forward"` // Number of maximal forwarded metrics at one select
NormalizeUnits bool `json:"normalize_units"` // Check unit meta flag and normalize it using cc-units
ChangeUnitPrefix map[string]string `json:"change_unit_prefix"` // Add prefix that should be applied to the metrics
dropMetrics map[string]bool // Internal map for O(1) lookup
}
@ -207,6 +210,38 @@ func (r *metricRouter) dropMetric(point lp.CCMetric) bool {
return false
}
func (r *metricRouter) prepareUnit(point lp.CCMetric) bool {
if r.config.NormalizeUnits {
if in_unit, ok := point.GetMeta("unit"); ok {
u := units.NewUnit(in_unit)
if u.Valid() {
point.AddMeta("unit", u.Short())
}
}
}
if newP, ok := r.config.ChangeUnitPrefix[point.Name()]; ok {
newPrefix := units.NewPrefix(newP)
if in_unit, ok := point.GetMeta("unit"); ok && newPrefix != units.InvalidPrefix {
u := units.NewUnit(in_unit)
if u.Valid() {
cclog.ComponentDebug("MetricRouter", "Change prefix to", newP, "for metric", point.Name())
conv, out_unit := units.GetUnitPrefixFactor(u, newPrefix)
if conv != nil && out_unit.Valid() {
if val, ok := point.GetField("value"); ok {
point.AddField("value", conv(val))
point.AddMeta("unit", out_unit.Short())
}
}
}
}
}
return true
}
// Start starts the metric router
func (r *metricRouter) Start() {
// start timer if configured
@ -232,9 +267,11 @@ func (r *metricRouter) Start() {
if new, ok := r.config.RenameMetrics[name]; ok {
point.SetName(new)
point.AddMeta("oldname", name)
}
r.DoAddTags(point)
r.DoDelTags(point)
}
r.prepareUnit(point)
for _, o := range r.outputs {
o <- point

17
receivers.json
View File

@ -4,5 +4,22 @@
"address": "nats://my-url",
"port" : "4222",
"database": "testcluster"
},
"redfish_recv": {
"type": "redfish",
"client_config": [
{
"hostname": "my-host-1",
"username": "username-1",
"password": "password-1",
"endpoint": "https://my-endpoint-1"
},
{
"hostname": "my-host-2",
"username": "username-2",
"password": "password-2",
"endpoint": "https://my-endpoint-2"
}
]
}
}

16
receivers/receiveManager.go
View File

@ -11,13 +11,12 @@ import (
var AvailableReceivers = map[string]func(name string, config json.RawMessage) (Receiver, error){
"nats": NewNatsReceiver,
"redfish": NewRedfishReceiver,
}
type receiveManager struct {
inputs []Receiver
output chan lp.CCMetric
done chan bool
wg *sync.WaitGroup
config []json.RawMessage
}
@ -33,8 +32,6 @@ func (rm *receiveManager) Init(wg *sync.WaitGroup, receiverConfigFile string) er
// Initialize struct fields
rm.inputs = make([]Receiver, 0)
rm.output = nil
rm.done = make(chan bool)
rm.wg = wg
rm.config = make([]json.RawMessage, 0)
configFile, err := os.Open(receiverConfigFile)
@ -58,7 +55,7 @@ func (rm *receiveManager) Init(wg *sync.WaitGroup, receiverConfigFile string) er
}
func (rm *receiveManager) Start() {
rm.wg.Add(1)
cclog.ComponentDebug("ReceiveManager", "START")
for _, r := range rm.inputs {
cclog.ComponentDebug("ReceiveManager", "START", r.Name())
@ -97,16 +94,19 @@ func (rm *receiveManager) AddOutput(output chan lp.CCMetric) {
}
func (rm *receiveManager) Close() {
cclog.ComponentDebug("ReceiveManager", "CLOSE")
// Close all receivers
for _, r := range rm.inputs {
cclog.ComponentDebug("ReceiveManager", "CLOSE", r.Name())
r.Close()
}
rm.wg.Done()
cclog.ComponentDebug("ReceiveManager", "CLOSE")
cclog.ComponentDebug("ReceiveManager", "DONE")
}
func New(wg *sync.WaitGroup, receiverConfigFile string) (ReceiveManager, error) {
r := &receiveManager{}
r := new(receiveManager)
err := r.Init(wg, receiverConfigFile)
if err != nil {
return nil, err

324
receivers/redfishReceiver.go Normal file
View File

@ -0,0 +1,324 @@
package receivers
import (
"encoding/json"
"fmt"
"strconv"
"sync"
"time"
cclog "github.com/ClusterCockpit/cc-metric-collector/internal/ccLogger"
lp "github.com/ClusterCockpit/cc-metric-collector/internal/ccMetric"
// See: https://pkg.go.dev/github.com/stmcginnis/gofish
"github.com/stmcginnis/gofish"
)
// RedfishReceiver configuration:
type RedfishReceiver struct {
receiver
config struct {
Type string `json:"type"`
Fanout int `json:"fanout,omitempty"` // Default fanout: 64
Interval int `json:"interval,omitempty"` // Default interval: 30s
// Client config for each redfish service
ClientConfigs []struct {
Hostname *string `json:"hostname"`
Username *string `json:"username"`
Password *string `json:"password"`
Endpoint *string `json:"endpoint"`
Insecure *bool `json:"insecure,omitempty"`
ExcludeMetrics []string `json:"exclude_metrics,omitempty"`
gofish gofish.ClientConfig
} `json:"client_config"`
}
done chan bool // channel to finish / stop redfish receiver
wg sync.WaitGroup // wait group for redfish receiver
}
// Start starts the redfish receiver
func (r *RedfishReceiver) Start() {
cclog.ComponentDebug(r.name, "START")
// readPowerMetric reads readfish power metric from the endpoint configured in conf
readPowerMetric := func(clientConfigIndex int) error {
clientConfig := &r.config.ClientConfigs[clientConfigIndex]
// Connect to redfish service
c, err := gofish.Connect(clientConfig.gofish)
if err != nil {
c := struct {
Username string
Endpoint string
BasicAuth bool
Insecure bool
}{
Username: clientConfig.gofish.Username,
Endpoint: clientConfig.gofish.Endpoint,
BasicAuth: clientConfig.gofish.BasicAuth,
Insecure: clientConfig.gofish.Insecure,
}
return fmt.Errorf("readPowerMetric: gofish.Connect(%+v) failed: %v", c, err)
}
defer c.Logout()
// Get all chassis managed by this service
chassis_list, err := c.Service.Chassis()
if err != nil {
return fmt.Errorf("readPowerMetric: c.Service.Chassis() failed: %v", err)
}
for _, chassis := range chassis_list {
timestamp := time.Now()
// Get power information for each chassis
power, err := chassis.Power()
if err != nil {
return fmt.Errorf("readPowerMetric: chassis.Power() failed: %v", err)
}
if power == nil {
continue
}
// Read min, max and average consumed watts for each power control
for _, pc := range power.PowerControl {
// Map of collected metrics
metrics := map[string]float32{
// PowerConsumedWatts shall represent the actual power being consumed (in
// Watts) by the chassis
"consumed_watts": pc.PowerConsumedWatts,
// AverageConsumedWatts shall represent the
// average power level that occurred averaged over the last IntervalInMin
// minutes.
"average_consumed_watts": pc.PowerMetrics.AverageConsumedWatts,
// MinConsumedWatts shall represent the
// minimum power level in watts that occurred within the last
// IntervalInMin minutes.
"min_consumed_watts": pc.PowerMetrics.MinConsumedWatts,
// MaxConsumedWatts shall represent the
// maximum power level in watts that occurred within the last
// IntervalInMin minutes
"max_consumed_watts": pc.PowerMetrics.MaxConsumedWatts,
}
intervalInMin := strconv.FormatFloat(float64(pc.PowerMetrics.IntervalInMin), 'f', -1, 32)
// Metrics to exclude
for _, key := range clientConfig.ExcludeMetrics {
delete(metrics, key)
}
// Set tags
tags := map[string]string{
"hostname": *clientConfig.Hostname,
"type": "node",
// ID uniquely identifies the resource
"id": pc.ID,
// MemberID shall uniquely identify the member within the collection. For
// services supporting Redfish v1.6 or higher, this value shall be the
// zero-based array index.
"member_id": pc.MemberID,
// PhysicalContext shall be a description of the affected device(s) or region
// within the chassis to which this power control applies.
"physical_context": string(pc.PhysicalContext),
// Name
"power_control_name": pc.Name,
}
// Delete empty tags
for key, value := range tags {
if value == "" {
delete(tags, key)
}
}
// Set meta data tags
meta := map[string]string{
"source": r.name,
"group": "Energy",
"interval_in_minutes": intervalInMin,
"unit": "watts",
}
// Delete empty meta data tags
for key, value := range meta {
if value == "" {
delete(meta, key)
}
}
for name, value := range metrics {
y, err := lp.New(name, tags, meta,
map[string]interface{}{
"value": value,
},
timestamp)
if err == nil {
r.sink <- y
}
}
}
}
return nil
}
// doReadPowerMetric read power metrics for all configure redfish services.
// To compensate latencies of the Redfish services a fanout is used.
doReadPowerMetric := func() {
// Compute fanout to use
realFanout := r.config.Fanout
if len(r.config.ClientConfigs) < realFanout {
realFanout = len(r.config.ClientConfigs)
}
// Create wait group and input channel for workers
var workerWaitGroup sync.WaitGroup
workerInput := make(chan int, realFanout)
// Create worker go routines
for i := 0; i < realFanout; i++ {
// Increment worker wait group counter
workerWaitGroup.Add(1)
go func() {
// Decrement worker wait group counter
defer workerWaitGroup.Done()
// Read power metrics for each client config
for clientConfigIndex := range workerInput {
err := readPowerMetric(clientConfigIndex)
if err != nil {
cclog.ComponentError(r.name, err)
}
}
}()
}
// Distribute client configs to workers
for i := range r.config.ClientConfigs {
// Check done channel status
select {
case workerInput <- i:
case <-r.done:
// process done event
// Stop workers, clear channel and wait for all workers to finish
close(workerInput)
for range workerInput {
}
workerWaitGroup.Wait()
return
}
}
// Stop workers and wait for all workers to finish
close(workerInput)
workerWaitGroup.Wait()
}
// Start redfish receiver
r.wg.Add(1)
go func() {
defer r.wg.Done()
// Create ticker
ticker := time.NewTicker(time.Duration(r.config.Interval) * time.Second)
defer ticker.Stop()
for {
doReadPowerMetric()
select {
case <-ticker.C:
// process ticker event -> continue
continue
case <-r.done:
// process done event
return
}
}
}()
cclog.ComponentDebug(r.name, "STARTED")
}
// Close redfish receiver
func (r *RedfishReceiver) Close() {
cclog.ComponentDebug(r.name, "CLOSE")
// Send the signal and wait
close(r.done)
r.wg.Wait()
cclog.ComponentDebug(r.name, "DONE")
}
// New function to create a new instance of the receiver
// Initialize the receiver by giving it a name and reading in the config JSON
func NewRedfishReceiver(name string, config json.RawMessage) (Receiver, error) {
r := new(RedfishReceiver)
// Set name
r.name = fmt.Sprintf("RedfishReceiver(%s)", name)
// Create done channel
r.done = make(chan bool)
// Set defaults in r.config
// Allow overwriting these defaults by reading config JSON
r.config.Fanout = 64
r.config.Interval = 30
// Read the redfish receiver specific JSON config
if len(config) > 0 {
err := json.Unmarshal(config, &r.config)
if err != nil {
cclog.ComponentError(r.name, "Error reading config:", err.Error())
return nil, err
}
}
// Create gofish client config
for i := range r.config.ClientConfigs {
clientConfig := &r.config.ClientConfigs[i]
gofishConfig := &clientConfig.gofish
if clientConfig.Hostname == nil {
err := fmt.Errorf("client config number %v requires hostname", i)
cclog.ComponentError(r.name, err)
return nil, err
}
if clientConfig.Endpoint == nil {
err := fmt.Errorf("client config number %v requires endpoint", i)
cclog.ComponentError(r.name, err)
return nil, err
}
gofishConfig.Endpoint = *clientConfig.Endpoint
if clientConfig.Username == nil {
err := fmt.Errorf("client config number %v requires username", i)
cclog.ComponentError(r.name, err)
return nil, err
}
gofishConfig.Username = *clientConfig.Username
if clientConfig.Password == nil {
err := fmt.Errorf("client config number %v requires password", i)
cclog.ComponentError(r.name, err)
return nil, err
}
gofishConfig.Password = *clientConfig.Password
gofishConfig.Insecure = true
if clientConfig.Insecure != nil {
gofishConfig.Insecure = *clientConfig.Insecure
}
}
return r, nil
}

14
receivers/sampleReceiver.go
View File

@ -36,16 +36,26 @@ func (r *SampleReceiver) Start() {
// or use own go routine but always make sure it exits
// as soon as it gets the signal of the r.done channel
//
// r.done = make(chan bool)
// r.wg.Add(1)
// go func() {
// defer r.wg.Done()
//
// // Create ticker
// ticker := time.NewTicker(30 * time.Second)
// defer ticker.Stop()
//
// for {
// readMetric()
// select {
// case <-ticker.C:
// // process ticker event -> continue
// continue
// case <-r.done:
// r.wg.Done()
// return
// }
// }
// r.wg.Done()
// }()
}

6
scripts/cc-metric-collector.config
View File

@ -15,3 +15,9 @@ CONF_DIR=/etc/cc-metric-collector
CONF_FILE=/etc/cc-metric-collector/cc-metric-collector.json
RESTART_ON_UPGRADE=true
# Golang runtime debugging. (see: https://pkg.go.dev/runtime)
# GODEBUG=gctrace=1
# Golang garbage collection target percentage
# GOGC=100

2
scripts/cc-metric-collector.init
View File

@ -75,7 +75,7 @@ case "$1" in
fi
# Start Daemon
start-stop-daemon --start -b --chdir "$WORK_DIR" --user "$CC_USER" -c "$CC_USER" --pidfile "$PID_FILE" --exec $DAEMON -- $DAEMON_OPTS
start-stop-daemon --start -b --chdir "$WORK_DIR" --user "$CC_USER" -c "$CC_USER" --pidfile "$PID_FILE" --exec $DAEMON -- $CC_OPTS
return=$?
if [ $return -eq 0 ]
then

2
scripts/likwid_perfgroup_to_cc_config.py
View File

@ -39,7 +39,7 @@ def group_to_json(groupfile):
llist = re.split("\s+", line)
calc = llist[-1]
metric = " ".join(llist[:-1])
scope = "cpu"
scope = "hwthread"
if "BOX" in calc:
scope = "socket"
if "PWR" in calc:

4
sinks.json
View File

@ -1,6 +1,8 @@
{
"mystdout": {
"type": "stdout",
"meta_as_tags" : true
"meta_as_tags": [
"unit"
]
}
}

89
sinks/httpSink.go
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 {
@ -32,58 +32,56 @@ 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)
}
}
func (s *HttpSink) Flush() error {
// buffer is read by client.Do, prevent concurrent modifications
s.lock.Lock()
defer s.lock.Unlock()
// Do not flush empty buffer
if s.buffer.Len() == 0 {
return nil
}
func (s *HttpSink) Flush() error {
// Own lock for as short as possible: the time it takes to copy the buffer.
s.lock.Lock()
buf := make([]byte, s.buffer.Len())
copy(buf, s.buffer.Bytes())
s.buffer.Reset()
s.lock.Unlock()
if len(buf) == 0 {
return nil
}
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, s.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
@ -94,21 +92,25 @@ func (s *HttpSink) Flush() error {
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
// Do request
res, err = s.client.Do(req)
if err != nil {
cclog.ComponentError(s.name, "transport/tcp error:", err.Error())
return err
// Wait between retries
time.Sleep(time.Duration(i+1) * (time.Second / 2))
continue
}
break
}
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}

18
sinks/influxAsyncSink.go
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)

144
sinks/influxSink.go
View File

@ -16,7 +16,11 @@ import (
"github.com/influxdata/influxdb-client-go/v2/api/write"
)
type InfluxSinkConfig struct {
type InfluxSink struct {
sink
client influxdb2.Client
writeApi influxdb2Api.WriteAPIBlocking
config struct {
defaultSinkConfig
Host string `json:"host,omitempty"`
Port string `json:"port,omitempty"`
@ -25,63 +29,45 @@ type InfluxSinkConfig struct {
Password string `json:"password,omitempty"`
Organization string `json:"organization,omitempty"`
SSL bool `json:"ssl,omitempty"`
FlushDelay string `json:"flush_delay,omitempty"`
// Maximum number of points sent to server in single request. Default 100
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
// 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"`
}
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
}
// connect connects to the InfluxDB server
func (s *InfluxSink) connect() error {
var auth string
// URI options:
// * http://host:port
// * https://host:port
var uri string
if s.config.SSL {
uri = fmt.Sprintf("https://%s:%s", s.config.Host, s.config.Port)
} else {
uri = fmt.Sprintf("http://%s:%s", s.config.Host, s.config.Port)
}
// Authentication options:
// * token
// * username:password
var auth string
if len(s.config.User) == 0 {
auth = s.config.Password
} else {
auth = fmt.Sprintf("%s:%s", s.config.User, s.config.Password)
}
cclog.ComponentDebug(s.name, "Using URI", uri, "Org", s.config.Organization, "Bucket", s.config.Database)
// 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{
InsecureSkipVerify: true,
@ -90,8 +76,11 @@ func (s *InfluxSink) connect() error {
clientOptions.SetPrecision(time.Second)
// Create new writeAPI
s.client = influxdb2.NewClientWithOptions(uri, auth, clientOptions)
s.writeApi = s.client.WriteAPIBlocking(s.config.Organization, s.config.Database)
// Check InfluxDB server accessibility
ok, err := s.client.Ping(context.Background())
if err != nil {
return err
@ -103,24 +92,24 @@ func (s *InfluxSink) connect() error {
}
func (s *InfluxSink) Write(m lp.CCMetric) error {
// err :=
// s.writeApi.WritePoint(
// context.Background(),
// m.ToPoint(s.meta_as_tags),
// )
if len(s.batch) == 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() {
// 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())
}
})
}
// Append metric to batch slice
p := m.ToPoint(s.meta_as_tags)
s.lock.Lock()
s.batch = append(s.batch, p)
@ -131,21 +120,39 @@ func (s *InfluxSink) Write(m lp.CCMetric) error {
return s.Flush()
}
// Flush if batch size is reached
if len(s.batch) == s.config.BatchSize {
return s.Flush()
}
return nil
}
// Flush sends all metrics buffered in batch slice to InfluxDB server
func (s *InfluxSink) Flush() error {
// Lock access to batch slice
s.lock.Lock()
defer s.lock.Unlock()
// Nothing to do, batch slice is empty
if len(s.batch) == 0 {
return nil
}
// Send metrics from batch slice
err := s.writeApi.WritePoint(context.Background(), s.batch...)
if err != nil {
cclog.ComponentError(s.name, "flush failed:", err.Error())
return err
}
// Clear batch slice
for i := range s.batch {
s.batch[i] = nil
}
s.batch = s.batch[:0]
return nil
}
@ -156,55 +163,54 @@ func (s *InfluxSink) Close() {
s.client.Close()
}
// NewInfluxSink create a new InfluxDB sink
func NewInfluxSink(name string, config json.RawMessage) (Sink, error) {
s := new(InfluxSink)
s.name = fmt.Sprintf("InfluxSink(%s)", name)
// Set config default values
s.config.BatchSize = 100
s.config.FlushDelay = "1s"
s.config.FlushInterval = "1s"
// Read config
if len(config) > 0 {
err := json.Unmarshal(config, &s.config)
if err != nil {
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 ||
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 InfluxSink")
if len(s.config.Host) == 0 {
return nil, errors.New("Missing host configuration required by InfluxSink")
}
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)
for _, k := range s.config.MetaAsTags {
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)
if len(s.config.FlushDelay) > 0 {
t, err := time.ParseDuration(s.config.FlushDelay)
// Configure flush delay duration
if len(s.config.FlushInterval) > 0 {
t, err := time.ParseDuration(s.config.FlushInterval)
if err == nil {
s.flushDelay = t
}
}
// allocate batch slice
s.batch = make([]*write.Point, 0, s.config.BatchSize)
// Connect to InfluxDB server

50
sinks/natsSink.go
View File

@ -5,6 +5,7 @@ import (
"encoding/json"
"errors"
"fmt"
"sync"
"time"
cclog "github.com/ClusterCockpit/cc-metric-collector/internal/ccLogger"
@ -17,9 +18,10 @@ type NatsSinkConfig struct {
defaultSinkConfig
Host string `json:"host,omitempty"`
Port string `json:"port,omitempty"`
Database string `json:"database,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 {
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 {
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
}
s.buffer.Reset()
}
return nil
}
func (s *NatsSink) Close() {
if s.client != nil {
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
}