ClusterCockpit/cc-backend
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-backend synced 2025-07-22 20:41:40 +02:00
Code Issues Packages Releases Activity

Merge branch 'master' into import-data-sanitation

Browse Source
This commit is contained in:
Jan Eitzinger
2023-04-07 08:57:42 +02:00
parent 52a07b51d2 13c31155fc
commit f8ba79e9e7
110 changed files with 8191 additions and 1464 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
internal/metricdata/cc-metric-store.go
View File

@@ -16,6 +16,7 @@ import (
"time"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
)
@@ -78,6 +79,7 @@ func (ccms *CCMetricStore) Init(rawConfig json.RawMessage) error {
var config CCMetricStoreConfig
if err := json.Unmarshal(rawConfig, &config); err != nil {
log.Warn("Error while unmarshaling raw json config")
return err
}
@@ -124,11 +126,13 @@ func (ccms *CCMetricStore) doRequest(
buf := &bytes.Buffer{}
if err := json.NewEncoder(buf).Encode(body); err != nil {
log.Warn("Error while encoding request body")
return nil, err
}
req, err := http.NewRequestWithContext(ctx, http.MethodPost, ccms.queryEndpoint, buf)
if err != nil {
log.Warn("Error while building request body")
return nil, err
}
if ccms.jwt != "" {
@@ -137,6 +141,7 @@ func (ccms *CCMetricStore) doRequest(
res, err := ccms.client.Do(req)
if err != nil {
log.Error("Error while performing request")
return nil, err
}
@@ -146,6 +151,7 @@ func (ccms *CCMetricStore) doRequest(
var resBody ApiQueryResponse
if err := json.NewDecoder(bufio.NewReader(res.Body)).Decode(&resBody); err != nil {
log.Warn("Error while decoding result body")
return nil, err
}
@@ -160,6 +166,7 @@ func (ccms *CCMetricStore) LoadData(
queries, assignedScope, err := ccms.buildQueries(job, metrics, scopes)
if err != nil {
log.Warn("Error while building queries")
return nil, err
}
@@ -174,6 +181,7 @@ func (ccms *CCMetricStore) LoadData(
resBody, err := ccms.doRequest(ctx, &req)
if err != nil {
log.Error("Error while performing request")
return nil, err
}
@@ -200,6 +208,7 @@ func (ccms *CCMetricStore) LoadData(
for _, res := range row {
if res.Error != nil {
/* Build list for "partial errors", if any */
errors = append(errors, fmt.Sprintf("failed to fetch '%s' from host '%s': %s", query.Metric, query.Hostname, *res.Error))
continue
}
@@ -240,7 +249,8 @@ func (ccms *CCMetricStore) LoadData(
}
if len(errors) != 0 {
return jobData, fmt.Errorf("cc-metric-store: %s", strings.Join(errors, ", "))
/* Returns list for "partial errors" */
return jobData, fmt.Errorf("METRICDATA/CCMS > Errors: %s", strings.Join(errors, ", "))
}
return jobData, nil
@@ -272,8 +282,8 @@ func (ccms *CCMetricStore) buildQueries(
remoteName := ccms.toRemoteName(metric)
mc := archive.GetMetricConfig(job.Cluster, metric)
if mc == nil {
// return nil, fmt.Errorf("metric '%s' is not specified for cluster '%s'", metric, job.Cluster)
// log.Printf("metric '%s' is not specified for cluster '%s'", metric, job.Cluster)
// return nil, fmt.Errorf("METRICDATA/CCMS > metric '%s' is not specified for cluster '%s'", metric, job.Cluster)
log.Infof("metric '%s' is not specified for cluster '%s'", metric, job.Cluster)
continue
}
@@ -483,7 +493,7 @@ func (ccms *CCMetricStore) buildQueries(
continue
}
return nil, nil, fmt.Errorf("TODO: unhandled case: native-scope=%s, requested-scope=%s", nativeScope, requestedScope)
return nil, nil, fmt.Errorf("METRICDATA/CCMS > TODO: unhandled case: native-scope=%s, requested-scope=%s", nativeScope, requestedScope)
}
}
}
@@ -498,6 +508,7 @@ func (ccms *CCMetricStore) LoadStats(
queries, _, err := ccms.buildQueries(job, metrics, []schema.MetricScope{schema.MetricScopeNode})
if err != nil {
log.Warn("Error while building query")
return nil, err
}
@@ -512,6 +523,7 @@ func (ccms *CCMetricStore) LoadStats(
resBody, err := ccms.doRequest(ctx, &req)
if err != nil {
log.Error("Error while performing request")
return nil, err
}
@@ -521,7 +533,7 @@ func (ccms *CCMetricStore) LoadStats(
metric := ccms.toLocalName(query.Metric)
data := res[0]
if data.Error != nil {
return nil, fmt.Errorf("fetching %s for node %s failed: %s", metric, query.Hostname, *data.Error)
return nil, fmt.Errorf("METRICDATA/CCMS > fetching %s for node %s failed: %s", metric, query.Hostname, *data.Error)
}
metricdata, ok := stats[metric]
@@ -531,7 +543,7 @@ func (ccms *CCMetricStore) LoadStats(
}
if data.Avg.IsNaN() || data.Min.IsNaN() || data.Max.IsNaN() {
return nil, fmt.Errorf("fetching %s for node %s failed: %s", metric, query.Hostname, "avg/min/max is NaN")
return nil, fmt.Errorf("METRICDATA/CCMS > fetching %s for node %s failed: %s", metric, query.Hostname, "avg/min/max is NaN")
}
metricdata[query.Hostname] = schema.MetricStatistics{
@@ -577,6 +589,7 @@ func (ccms *CCMetricStore) LoadNodeData(
resBody, err := ccms.doRequest(ctx, &req)
if err != nil {
log.Error("Error while performing request")
return nil, err
}
@@ -593,11 +606,12 @@ func (ccms *CCMetricStore) LoadNodeData(
metric := ccms.toLocalName(query.Metric)
qdata := res[0]
if qdata.Error != nil {
/* Build list for "partial errors", if any */
errors = append(errors, fmt.Sprintf("fetching %s for node %s failed: %s", metric, query.Hostname, *qdata.Error))
}
if qdata.Avg.IsNaN() || qdata.Min.IsNaN() || qdata.Max.IsNaN() {
// return nil, fmt.Errorf("fetching %s for node %s failed: %s", metric, query.Hostname, "avg/min/max is NaN")
// return nil, fmt.Errorf("METRICDATA/CCMS > fetching %s for node %s failed: %s", metric, query.Hostname, "avg/min/max is NaN")
qdata.Avg, qdata.Min, qdata.Max = 0., 0., 0.
}
@@ -626,7 +640,8 @@ func (ccms *CCMetricStore) LoadNodeData(
}
if len(errors) != 0 {
return data, fmt.Errorf("cc-metric-store: %s", strings.Join(errors, ", "))
/* Returns list of "partial errors" */
return data, fmt.Errorf("METRICDATA/CCMS > Errors: %s", strings.Join(errors, ", "))
}
return data, nil

49
internal/metricdata/influxdb-v2.go
View File

@@ -10,11 +10,11 @@ import (
"encoding/json"
"errors"
"fmt"
"log"
"strings"
"time"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
influxdb2 "github.com/influxdata/influxdb-client-go/v2"
influxdb2Api "github.com/influxdata/influxdb-client-go/v2/api"
@@ -37,6 +37,7 @@ type InfluxDBv2DataRepository struct {
func (idb *InfluxDBv2DataRepository) Init(rawConfig json.RawMessage) error {
var config InfluxDBv2DataRepositoryConfig
if err := json.Unmarshal(rawConfig, &config); err != nil {
log.Warn("Error while unmarshaling raw json config")
return err
}
@@ -71,7 +72,7 @@ func (idb *InfluxDBv2DataRepository) LoadData(
for _, h := range job.Resources {
if h.HWThreads != nil || h.Accelerators != nil {
// TODO
return nil, errors.New("the InfluxDB metric data repository does not yet support HWThreads or Accelerators")
return nil, errors.New("METRICDATA/INFLUXV2 > the InfluxDB metric data repository does not yet support HWThreads or Accelerators")
}
hostsConds = append(hostsConds, fmt.Sprintf(`r["hostname"] == "%s"`, h.Hostname))
}
@@ -84,7 +85,7 @@ func (idb *InfluxDBv2DataRepository) LoadData(
switch scope {
case "node":
// Get Finest Granularity, Groupy By Measurement and Hostname (== Metric / Node), Calculate Mean for 60s windows
// log.Println("Note: Scope 'node' requested. ")
// log.Info("Scope 'node' requested. ")
query = fmt.Sprintf(`
from(bucket: "%s")
|> range(start: %s, stop: %s)
@@ -97,10 +98,10 @@ func (idb *InfluxDBv2DataRepository) LoadData(
idb.formatTime(job.StartTime), idb.formatTime(idb.epochToTime(job.StartTimeUnix+int64(job.Duration)+int64(1))),
measurementsCond, hostsCond)
case "socket":
log.Println("Note: Scope 'socket' requested, but not yet supported: Will return 'node' scope only. ")
log.Info("Scope 'socket' requested, but not yet supported: Will return 'node' scope only. ")
continue
case "core":
log.Println("Note: Scope 'core' requested, but not yet supported: Will return 'node' scope only. ")
log.Info(" Scope 'core' requested, but not yet supported: Will return 'node' scope only. ")
continue
// Get Finest Granularity only, Set NULL to 0.0
// query = fmt.Sprintf(`
@@ -114,13 +115,14 @@ func (idb *InfluxDBv2DataRepository) LoadData(
// idb.formatTime(job.StartTime), idb.formatTime(idb.epochToTime(job.StartTimeUnix + int64(job.Duration) + int64(1) )),
// measurementsCond, hostsCond)
default:
log.Println("Note: Unknown Scope requested: Will return 'node' scope. ")
log.Infof("Unknown scope '%s' requested: Will return 'node' scope.", scope)
continue
// return nil, errors.New("the InfluxDB metric data repository does not yet support other scopes than 'node'")
// return nil, errors.New("METRICDATA/INFLUXV2 > the InfluxDB metric data repository does not yet support other scopes than 'node'")
}
rows, err := idb.queryClient.Query(ctx, query)
if err != nil {
log.Error("Error while performing query")
return nil, err
}
@@ -191,6 +193,7 @@ func (idb *InfluxDBv2DataRepository) LoadData(
// hostSeries.Data = append(hostSeries.Data, schema.Float(val))
// }
default:
log.Infof("Unknown scope '%s' requested: Will return 'node' scope.", scope)
continue
// return nil, errors.New("the InfluxDB metric data repository does not yet support other scopes than 'node, core'")
}
@@ -201,21 +204,22 @@ func (idb *InfluxDBv2DataRepository) LoadData(
// Get Stats
stats, err := idb.LoadStats(job, metrics, ctx)
if err != nil {
log.Warn("Error while loading statistics")
return nil, err
}
for _, scope := range scopes {
if scope == "node" { // No 'socket/core' support yet
for metric, nodes := range stats {
// log.Println(fmt.Sprintf("<< Add Stats for : Field %s >>", metric))
// log.Debugf("<< Add Stats for : Field %s >>", metric)
for node, stats := range nodes {
// log.Println(fmt.Sprintf("<< Add Stats for : Host %s : Min %.2f, Max %.2f, Avg %.2f >>", node, stats.Min, stats.Max, stats.Avg ))
// log.Debugf("<< Add Stats for : Host %s : Min %.2f, Max %.2f, Avg %.2f >>", node, stats.Min, stats.Max, stats.Avg )
for index, _ := range jobData[metric][scope].Series {
// log.Println(fmt.Sprintf("<< Try to add Stats to Series in Position %d >>", index))
// log.Debugf("<< Try to add Stats to Series in Position %d >>", index)
if jobData[metric][scope].Series[index].Hostname == node {
// log.Println(fmt.Sprintf("<< Match for Series in Position %d : Host %s >>", index, jobData[metric][scope].Series[index].Hostname))
jobData[metric][scope].Series[index].Statistics = schema.MetricStatistics{Avg: stats.Avg, Min: stats.Min, Max: stats.Max}
// log.Println(fmt.Sprintf("<< Result Inner: Min %.2f, Max %.2f, Avg %.2f >>", jobData[metric][scope].Series[index].Statistics.Min, jobData[metric][scope].Series[index].Statistics.Max, jobData[metric][scope].Series[index].Statistics.Avg))
// log.Debugf("<< Match for Series in Position %d : Host %s >>", index, jobData[metric][scope].Series[index].Hostname)
jobData[metric][scope].Series[index].Statistics = &schema.MetricStatistics{Avg: stats.Avg, Min: stats.Min, Max: stats.Max}
// log.Debugf("<< Result Inner: Min %.2f, Max %.2f, Avg %.2f >>", jobData[metric][scope].Series[index].Statistics.Min, jobData[metric][scope].Series[index].Statistics.Max, jobData[metric][scope].Series[index].Statistics.Avg)
}
}
}
@@ -227,9 +231,9 @@ func (idb *InfluxDBv2DataRepository) LoadData(
// for _, scope := range scopes {
// for _, met := range metrics {
// for _, series := range jobData[met][scope].Series {
// log.Println(fmt.Sprintf("<< Result: %d data points for metric %s on %s with scope %s, Stats: Min %.2f, Max %.2f, Avg %.2f >>",
// log.Debugf("<< Result: %d data points for metric %s on %s with scope %s, Stats: Min %.2f, Max %.2f, Avg %.2f >>",
// len(series.Data), met, series.Hostname, scope,
// series.Statistics.Min, series.Statistics.Max, series.Statistics.Avg))
// series.Statistics.Min, series.Statistics.Max, series.Statistics.Avg)
// }
// }
// }
@@ -248,7 +252,7 @@ func (idb *InfluxDBv2DataRepository) LoadStats(
for _, h := range job.Resources {
if h.HWThreads != nil || h.Accelerators != nil {
// TODO
return nil, errors.New("the InfluxDB metric data repository does not yet support HWThreads or Accelerators")
return nil, errors.New("METRICDATA/INFLUXV2 > the InfluxDB metric data repository does not yet support HWThreads or Accelerators")
}
hostsConds = append(hostsConds, fmt.Sprintf(`r["hostname"] == "%s"`, h.Hostname))
}
@@ -257,7 +261,7 @@ func (idb *InfluxDBv2DataRepository) LoadStats(
// lenMet := len(metrics)
for _, metric := range metrics {
// log.Println(fmt.Sprintf("<< You are here: %s (Index %d of %d metrics)", metric, index, lenMet))
// log.Debugf("<< You are here: %s (Index %d of %d metrics)", metric, index, lenMet)
query := fmt.Sprintf(`
data = from(bucket: "%s")
@@ -274,6 +278,7 @@ func (idb *InfluxDBv2DataRepository) LoadStats(
rows, err := idb.queryClient.Query(ctx, query)
if err != nil {
log.Error("Error while performing query")
return nil, err
}
@@ -284,17 +289,17 @@ func (idb *InfluxDBv2DataRepository) LoadStats(
avg, avgok := row.ValueByKey("avg").(float64)
if !avgok {
// log.Println(fmt.Sprintf(">> Assertion error for metric %s, statistic AVG. Expected 'float64', got %v", metric, avg))
// log.Debugf(">> Assertion error for metric %s, statistic AVG. Expected 'float64', got %v", metric, avg)
avg = 0.0
}
min, minok := row.ValueByKey("min").(float64)
if !minok {
// log.Println(fmt.Sprintf(">> Assertion error for metric %s, statistic MIN. Expected 'float64', got %v", metric, min))
// log.Debugf(">> Assertion error for metric %s, statistic MIN. Expected 'float64', got %v", metric, min)
min = 0.0
}
max, maxok := row.ValueByKey("max").(float64)
if !maxok {
// log.Println(fmt.Sprintf(">> Assertion error for metric %s, statistic MAX. Expected 'float64', got %v", metric, max))
// log.Debugf(">> Assertion error for metric %s, statistic MAX. Expected 'float64', got %v", metric, max)
max = 0.0
}
@@ -318,7 +323,7 @@ func (idb *InfluxDBv2DataRepository) LoadNodeData(
ctx context.Context) (map[string]map[string][]*schema.JobMetric, error) {
// TODO : Implement to be used in Analysis- und System/Node-View
log.Println(fmt.Sprintf("LoadNodeData unimplemented for InfluxDBv2DataRepository, Args: cluster %s, metrics %v, nodes %v, scopes %v", cluster, metrics, nodes, scopes))
log.Infof("LoadNodeData unimplemented for InfluxDBv2DataRepository, Args: cluster %s, metrics %v, nodes %v, scopes %v", cluster, metrics, nodes, scopes)
return nil, errors.New("unimplemented for InfluxDBv2DataRepository")
return nil, errors.New("METRICDATA/INFLUXV2 > unimplemented for InfluxDBv2DataRepository")
}

25
internal/metricdata/metricdata.go
View File

@@ -46,6 +46,7 @@ func Init(disableArchive bool) error {
Kind string `json:"kind"`
}
if err := json.Unmarshal(cluster.MetricDataRepository, &kind); err != nil {
log.Warn("Error while unmarshaling raw json MetricDataRepository")
return err
}
@@ -55,13 +56,16 @@ func Init(disableArchive bool) error {
mdr = &CCMetricStore{}
case "influxdb":
mdr = &InfluxDBv2DataRepository{}
case "prometheus":
mdr = &PrometheusDataRepository{}
case "test":
mdr = &TestMetricDataRepository{}
default:
return fmt.Errorf("unkown metric data repository '%s' for cluster '%s'", kind.Kind, cluster.Name)
return fmt.Errorf("METRICDATA/METRICDATA > Unknown MetricDataRepository %v for cluster %v", kind.Kind, cluster.Name)
}
if err := mdr.Init(cluster.MetricDataRepository); err != nil {
log.Errorf("Error initializing MetricDataRepository %v for cluster %v", kind.Kind, cluster.Name)
return err
}
metricDataRepos[cluster.Name] = mdr
@@ -88,7 +92,7 @@ func LoadData(job *schema.Job,
repo, ok := metricDataRepos[job.Cluster]
if !ok {
return fmt.Errorf("no metric data repository configured for '%s'", job.Cluster), 0, 0
return fmt.Errorf("METRICDATA/METRICDATA > no metric data repository configured for '%s'", job.Cluster), 0, 0
}
if scopes == nil {
@@ -105,8 +109,9 @@ func LoadData(job *schema.Job,
jd, err = repo.LoadData(job, metrics, scopes, ctx)
if err != nil {
if len(jd) != 0 {
log.Errorf("partial error: %s", err.Error())
log.Warnf("partial error: %s", err.Error())
} else {
log.Error("Error while loading job data from metric repository")
return err, 0, 0
}
}
@@ -114,6 +119,7 @@ func LoadData(job *schema.Job,
} else {
jd, err = archive.GetHandle().LoadJobData(job)
if err != nil {
log.Error("Error while loading job data from archive")
return err, 0, 0
}
@@ -156,10 +162,12 @@ func LoadData(job *schema.Job,
}
prepareJobData(job, jd, scopes)
return jd, ttl, size
})
if err, ok := data.(error); ok {
log.Error("Error in returned dataset")
return nil, err
}
@@ -179,11 +187,12 @@ func LoadAverages(
repo, ok := metricDataRepos[job.Cluster]
if !ok {
return fmt.Errorf("no metric data repository configured for '%s'", job.Cluster)
return fmt.Errorf("METRICDATA/METRICDATA > no metric data repository configured for '%s'", job.Cluster)
}
stats, err := repo.LoadStats(job, metrics, ctx)
if err != nil {
log.Errorf("Error while loading statistics for job %v (User %v, Project %v)", job.JobID, job.User, job.Project)
return err
}
@@ -214,7 +223,7 @@ func LoadNodeData(
repo, ok := metricDataRepos[cluster]
if !ok {
return nil, fmt.Errorf("no metric data repository configured for '%s'", cluster)
return nil, fmt.Errorf("METRICDATA/METRICDATA > no metric data repository configured for '%s'", cluster)
}
if metrics == nil {
@@ -226,14 +235,15 @@ func LoadNodeData(
data, err := repo.LoadNodeData(cluster, metrics, nodes, scopes, from, to, ctx)
if err != nil {
if len(data) != 0 {
log.Errorf("partial error: %s", err.Error())
log.Warnf("partial error: %s", err.Error())
} else {
log.Error("Error while loading node data from metric repository")
return nil, err
}
}
if data == nil {
return nil, fmt.Errorf("the metric data repository for '%s' does not support this query", cluster)
return nil, fmt.Errorf("METRICDATA/METRICDATA > the metric data repository for '%s' does not support this query", cluster)
}
return data, nil
@@ -300,6 +310,7 @@ func ArchiveJob(job *schema.Job, ctx context.Context) (*schema.JobMeta, error) {
jobData, err := LoadData(job, allMetrics, scopes, ctx)
if err != nil {
log.Error("Error wile loading job data for archiving")
return nil, err
}

446
internal/metricdata/prometheus.go Normal file
View File

@@ -0,0 +1,446 @@
// Copyright (C) 2022 DKRZ
// All rights reserved.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package metricdata
import (
"bytes"
"context"
"encoding/json"
"errors"
"fmt"
"math"
"net/http"
"os"
"regexp"
"sort"
"strings"
"sync"
"text/template"
"time"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
promapi "github.com/prometheus/client_golang/api"
promv1 "github.com/prometheus/client_golang/api/prometheus/v1"
promcfg "github.com/prometheus/common/config"
promm "github.com/prometheus/common/model"
)
type PrometheusDataRepositoryConfig struct {
Url string `json:"url"`
Username string `json:"username,omitempty"`
Suffix string `json:"suffix,omitempty"`
Templates map[string]string `json:"query-templates"`
}
type PrometheusDataRepository struct {
client promapi.Client
queryClient promv1.API
suffix string
templates map[string]*template.Template
}
type PromQLArgs struct {
Nodes string
}
type Trie map[rune]Trie
var logOnce sync.Once
func contains(s []schema.MetricScope, str schema.MetricScope) bool {
for _, v := range s {
if v == str {
return true
}
}
return false
}
func MinMaxMean(data []schema.Float) (float64, float64, float64) {
if len(data) == 0 {
return 0.0, 0.0, 0.0
}
min := math.MaxFloat64
max := -math.MaxFloat64
var sum float64
var n float64
for _, val := range data {
if val.IsNaN() {
continue
}
sum += float64(val)
n += 1
if float64(val) > max {
max = float64(val)
}
if float64(val) < min {
min = float64(val)
}
}
return min, max, sum / n
}
// Rewritten from
// https://github.com/ermanh/trieregex/blob/master/trieregex/trieregex.py
func nodeRegex(nodes []string) string {
root := Trie{}
// add runes of each compute node to trie
for _, node := range nodes {
_trie := root
for _, c := range node {
if _, ok := _trie[c]; !ok {
_trie[c] = Trie{}
}
_trie = _trie[c]
}
_trie['*'] = Trie{}
}
// recursively build regex from rune trie
var trieRegex func(trie Trie, reset bool) string
trieRegex = func(trie Trie, reset bool) string {
if reset == true {
trie = root
}
if len(trie) == 0 {
return ""
}
if len(trie) == 1 {
for key, _trie := range trie {
if key == '*' {
return ""
}
return regexp.QuoteMeta(string(key)) + trieRegex(_trie, false)
}
} else {
sequences := []string{}
for key, _trie := range trie {
if key != '*' {
sequences = append(sequences, regexp.QuoteMeta(string(key))+trieRegex(_trie, false))
}
}
sort.Slice(sequences, func(i, j int) bool {
return (-len(sequences[i]) < -len(sequences[j])) || (sequences[i] < sequences[j])
})
var result string
// single edge from this tree node
if len(sequences) == 1 {
result = sequences[0]
if len(result) > 1 {
result = "(?:" + result + ")"
}
// multiple edges, each length 1
} else if s := strings.Join(sequences, ""); len(s) == len(sequences) {
// char or numeric range
if len(s)-1 == int(s[len(s)-1])-int(s[0]) {
result = fmt.Sprintf("[%c-%c]", s[0], s[len(s)-1])
// char or numeric set
} else {
result = "[" + s + "]"
}
// multiple edges of different lengths
} else {
result = "(?:" + strings.Join(sequences, "|") + ")"
}
if _, ok := trie['*']; ok {
result += "?"
}
return result
}
return ""
}
return trieRegex(root, true)
}
func (pdb *PrometheusDataRepository) Init(rawConfig json.RawMessage) error {
var config PrometheusDataRepositoryConfig
// parse config
if err := json.Unmarshal(rawConfig, &config); err != nil {
log.Warn("Error while unmarshaling raw json config")
return err
}
// support basic authentication
var rt http.RoundTripper = nil
if prom_pw := os.Getenv("PROMETHEUS_PASSWORD"); prom_pw != "" && config.Username != "" {
prom_pw := promcfg.Secret(prom_pw)
rt = promcfg.NewBasicAuthRoundTripper(config.Username, prom_pw, "", promapi.DefaultRoundTripper)
} else {
if config.Username != "" {
return errors.New("METRICDATA/PROMETHEUS > Prometheus username provided, but PROMETHEUS_PASSWORD not set.")
}
}
// init client
client, err := promapi.NewClient(promapi.Config{
Address: config.Url,
RoundTripper: rt,
})
if err != nil {
log.Error("Error while initializing new prometheus client")
return err
}
// init query client
pdb.client = client
pdb.queryClient = promv1.NewAPI(pdb.client)
// site config
pdb.suffix = config.Suffix
// init query templates
pdb.templates = make(map[string]*template.Template)
for metric, templ := range config.Templates {
pdb.templates[metric], err = template.New(metric).Parse(templ)
if err == nil {
log.Debugf("Added PromQL template for %s: %s", metric, templ)
} else {
log.Warnf("Failed to parse PromQL template %s for metric %s", templ, metric)
}
}
return nil
}
// TODO: respect scope argument
func (pdb *PrometheusDataRepository) FormatQuery(
metric string,
scope schema.MetricScope,
nodes []string,
cluster string) (string, error) {
args := PromQLArgs{}
if len(nodes) > 0 {
args.Nodes = fmt.Sprintf("(%s)%s", nodeRegex(nodes), pdb.suffix)
} else {
args.Nodes = fmt.Sprintf(".*%s", pdb.suffix)
}
buf := &bytes.Buffer{}
if templ, ok := pdb.templates[metric]; ok {
err := templ.Execute(buf, args)
if err != nil {
return "", errors.New(fmt.Sprintf("METRICDATA/PROMETHEUS > Error compiling template %v", templ))
} else {
query := buf.String()
log.Debugf("PromQL: %s", query)
return query, nil
}
} else {
return "", errors.New(fmt.Sprintf("METRICDATA/PROMETHEUS > No PromQL for metric %s configured.", metric))
}
}
// Convert PromAPI row to CC schema.Series
func (pdb *PrometheusDataRepository) RowToSeries(
from time.Time,
step int64,
steps int64,
row *promm.SampleStream) schema.Series {
ts := from.Unix()
hostname := strings.TrimSuffix(string(row.Metric["exported_instance"]), pdb.suffix)
// init array of expected length with NaN
values := make([]schema.Float, steps+1)
for i, _ := range values {
values[i] = schema.NaN
}
// copy recorded values from prom sample pair
for _, v := range row.Values {
idx := (v.Timestamp.Unix() - ts) / step
values[idx] = schema.Float(v.Value)
}
min, max, mean := MinMaxMean(values)
// output struct
return schema.Series{
Hostname: hostname,
Data: values,
Statistics: &schema.MetricStatistics{
Avg: mean,
Min: min,
Max: max,
},
}
}
func (pdb *PrometheusDataRepository) LoadData(
job *schema.Job,
metrics []string,
scopes []schema.MetricScope,
ctx context.Context) (schema.JobData, error) {
// TODO respect requested scope
if len(scopes) == 0 || !contains(scopes, schema.MetricScopeNode) {
scopes = append(scopes, schema.MetricScopeNode)
}
jobData := make(schema.JobData)
// parse job specs
nodes := make([]string, len(job.Resources))
for i, resource := range job.Resources {
nodes[i] = resource.Hostname
}
from := job.StartTime
to := job.StartTime.Add(time.Duration(job.Duration) * time.Second)
for _, scope := range scopes {
if scope != schema.MetricScopeNode {
logOnce.Do(func() {
log.Infof("Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
})
continue
}
for _, metric := range metrics {
metricConfig := archive.GetMetricConfig(job.Cluster, metric)
if metricConfig == nil {
log.Warnf("Error in LoadData: Metric %s for cluster %s not configured", metric, job.Cluster)
return nil, errors.New("Prometheus config error")
}
query, err := pdb.FormatQuery(metric, scope, nodes, job.Cluster)
if err != nil {
log.Warn("Error while formatting prometheus query")
return nil, err
}
// ranged query over all job nodes
r := promv1.Range{
Start: from,
End: to,
Step: time.Duration(metricConfig.Timestep * 1e9),
}
result, warnings, err := pdb.queryClient.QueryRange(ctx, query, r)
if err != nil {
log.Errorf("Prometheus query error in LoadData: %v\nQuery: %s", err, query)
return nil, errors.New("Prometheus query error")
}
if len(warnings) > 0 {
log.Warnf("Warnings: %v\n", warnings)
}
// init data structures
if _, ok := jobData[metric]; !ok {
jobData[metric] = make(map[schema.MetricScope]*schema.JobMetric)
}
jobMetric, ok := jobData[metric][scope]
if !ok {
jobMetric = &schema.JobMetric{
Unit: metricConfig.Unit,
Scope: scope,
Timestep: metricConfig.Timestep,
Series: make([]schema.Series, 0),
}
jobData[metric][scope] = jobMetric
}
step := int64(metricConfig.Timestep)
steps := int64(to.Sub(from).Seconds()) / step
// iter rows of host, metric, values
for _, row := range result.(promm.Matrix) {
jobMetric.Series = append(jobMetric.Series,
pdb.RowToSeries(from, step, steps, row))
}
// sort by hostname to get uniform coloring
sort.Slice(jobMetric.Series, func(i, j int) bool {
return (jobMetric.Series[i].Hostname < jobMetric.Series[j].Hostname)
})
}
}
return jobData, nil
}
// TODO change implementation to precomputed/cached stats
func (pdb *PrometheusDataRepository) LoadStats(
job *schema.Job,
metrics []string,
ctx context.Context) (map[string]map[string]schema.MetricStatistics, error) {
// map of metrics of nodes of stats
stats := map[string]map[string]schema.MetricStatistics{}
data, err := pdb.LoadData(job, metrics, []schema.MetricScope{schema.MetricScopeNode}, ctx)
if err != nil {
log.Warn("Error while loading job for stats")
return nil, err
}
for metric, metricData := range data {
stats[metric] = make(map[string]schema.MetricStatistics)
for _, series := range metricData[schema.MetricScopeNode].Series {
stats[metric][series.Hostname] = *series.Statistics
}
}
return stats, nil
}
func (pdb *PrometheusDataRepository) LoadNodeData(
cluster string,
metrics, nodes []string,
scopes []schema.MetricScope,
from, to time.Time,
ctx context.Context) (map[string]map[string][]*schema.JobMetric, error) {
t0 := time.Now()
// Map of hosts of metrics of value slices
data := make(map[string]map[string][]*schema.JobMetric)
// query db for each metric
// TODO: scopes seems to be always empty
if len(scopes) == 0 || !contains(scopes, schema.MetricScopeNode) {
scopes = append(scopes, schema.MetricScopeNode)
}
for _, scope := range scopes {
if scope != schema.MetricScopeNode {
logOnce.Do(func() {
log.Infof("Note: Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
})
continue
}
for _, metric := range metrics {
metricConfig := archive.GetMetricConfig(cluster, metric)
if metricConfig == nil {
log.Warnf("Error in LoadNodeData: Metric %s for cluster %s not configured", metric, cluster)
return nil, errors.New("Prometheus config error")
}
query, err := pdb.FormatQuery(metric, scope, nodes, cluster)
if err != nil {
log.Warn("Error while formatting prometheus query")
return nil, err
}
// ranged query over all nodes
r := promv1.Range{
Start: from,
End: to,
Step: time.Duration(metricConfig.Timestep * 1e9),
}
result, warnings, err := pdb.queryClient.QueryRange(ctx, query, r)
if err != nil {
log.Errorf("Prometheus query error in LoadNodeData: %v\n", err)
return nil, errors.New("Prometheus query error")
}
if len(warnings) > 0 {
log.Warnf("Warnings: %v\n", warnings)
}
step := int64(metricConfig.Timestep)
steps := int64(to.Sub(from).Seconds()) / step
// iter rows of host, metric, values
for _, row := range result.(promm.Matrix) {
hostname := strings.TrimSuffix(string(row.Metric["exported_instance"]), pdb.suffix)
hostdata, ok := data[hostname]
if !ok {
hostdata = make(map[string][]*schema.JobMetric)
data[hostname] = hostdata
}
// output per host and metric
hostdata[metric] = append(hostdata[metric], &schema.JobMetric{
Unit: metricConfig.Unit,
Scope: scope,
Timestep: metricConfig.Timestep,
Series: []schema.Series{pdb.RowToSeries(from, step, steps, row)},
},
)
}
}
}
t1 := time.Since(t0)
log.Debugf("LoadNodeData of %v nodes took %s", len(data), t1)
return data, nil
}