2020-02-23 12:35:47 +01:00
package promscrape
import (
2020-11-04 16:12:05 +01:00
"flag"
2020-02-23 12:35:47 +01:00
"fmt"
"io"
2020-12-14 12:36:48 +01:00
"net/http"
2022-04-19 17:26:21 +02:00
"regexp"
2020-02-23 12:35:47 +01:00
"sort"
2020-12-14 13:02:57 +01:00
"strconv"
"strings"
2020-02-23 12:35:47 +01:00
"sync"
"time"
2022-02-03 17:57:36 +01:00
"unsafe"
2020-10-20 20:44:59 +02:00
2024-06-12 16:34:18 +02:00
"github.com/VictoriaMetrics/metrics"
"github.com/cespare/xxhash/v2"
2023-12-06 16:31:25 +01:00
"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
2020-10-20 20:44:59 +02:00
"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
2022-11-30 06:22:12 +01:00
"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
2024-07-17 13:52:10 +02:00
"github.com/VictoriaMetrics/VictoriaMetrics/lib/stringsutil"
2020-02-23 12:35:47 +01:00
)
2024-06-12 16:34:18 +02:00
var maxDroppedTargets = flag . Int ( "promscrape.maxDroppedTargets" , 10000 , "The maximum number of droppedTargets to show at /api/v1/targets page. " +
2020-11-04 16:12:05 +01:00
"Increase this value if your setup drops more scrape targets during relabeling and you need investigating labels for all the dropped targets. " +
"Note that the increased number of tracked dropped targets may result in increased memory usage" )
2020-02-23 12:35:47 +01:00
var tsmGlobal = newTargetStatusMap ( )
2022-02-03 17:57:36 +01:00
// WriteTargetResponse serves requests to /target_response?id=<id>
//
// It fetches response for the given target id and returns it.
func WriteTargetResponse ( w http . ResponseWriter , r * http . Request ) error {
targetID := r . FormValue ( "id" )
sw := tsmGlobal . getScrapeWorkByTargetID ( targetID )
if sw == nil {
return fmt . Errorf ( "cannot find target for id=%s" , targetID )
}
data , err := sw . getTargetResponse ( )
if err != nil {
return fmt . Errorf ( "cannot fetch response from id=%s: %w" , targetID , err )
}
w . Header ( ) . Set ( "Content-Type" , "text/plain; charset=utf-8" )
_ , err = w . Write ( data )
return err
}
2020-12-14 13:02:57 +01:00
// WriteHumanReadableTargetsStatus writes human-readable status for all the scrape targets to w according to r.
func WriteHumanReadableTargetsStatus ( w http . ResponseWriter , r * http . Request ) {
2022-06-06 23:57:05 +02:00
filter := getRequestFilter ( r )
tsr := tsmGlobal . getTargetsStatusByJob ( filter )
2020-12-14 13:02:57 +01:00
if accept := r . Header . Get ( "Accept" ) ; strings . Contains ( accept , "text/html" ) {
2020-12-14 12:36:48 +01:00
w . Header ( ) . Set ( "Content-Type" , "text/html; charset=utf-8" )
2022-06-06 23:57:05 +02:00
WriteTargetsResponseHTML ( w , tsr , filter )
2020-12-14 13:02:57 +01:00
} else {
w . Header ( ) . Set ( "Content-Type" , "text/plain; charset=utf-8" )
2022-06-06 23:57:05 +02:00
WriteTargetsResponsePlain ( w , tsr , filter )
2020-12-14 12:36:48 +01:00
}
2020-02-23 12:35:47 +01:00
}
2022-06-06 23:57:05 +02:00
// WriteServiceDiscovery writes /service-discovery response to w similar to http://demo.robustperception.io:9090/service-discovery
func WriteServiceDiscovery ( w http . ResponseWriter , r * http . Request ) {
filter := getRequestFilter ( r )
tsr := tsmGlobal . getTargetsStatusByJob ( filter )
w . Header ( ) . Set ( "Content-Type" , "text/html; charset=utf-8" )
WriteServiceDiscoveryResponse ( w , tsr , filter )
}
2020-10-20 20:44:59 +02:00
// WriteAPIV1Targets writes /api/v1/targets to w according to https://prometheus.io/docs/prometheus/latest/querying/api/#targets
func WriteAPIV1Targets ( w io . Writer , state string ) {
if state == "" {
state = "any"
}
fmt . Fprintf ( w , ` { "status":"success","data": { "activeTargets": ` )
if state == "active" || state == "any" {
tsmGlobal . WriteActiveTargetsJSON ( w )
} else {
fmt . Fprintf ( w , ` [] ` )
}
fmt . Fprintf ( w , ` ,"droppedTargets": ` )
if state == "dropped" || state == "any" {
droppedTargetsMap . WriteDroppedTargetsJSON ( w )
} else {
fmt . Fprintf ( w , ` [] ` )
}
fmt . Fprintf ( w , ` }} ` )
}
2020-02-23 12:35:47 +01:00
type targetStatusMap struct {
2021-06-18 09:53:10 +02:00
mu sync . Mutex
2022-02-03 17:57:36 +01:00
m map [ * scrapeWork ] * targetStatus
2021-06-18 09:53:10 +02:00
jobNames [ ] string
2023-12-06 16:31:25 +01:00
// the current number of `up` targets in the given jobName
upByJob map [ string ] int
// the current number of `down` targets in the given jobName
downByJob map [ string ] int
2020-02-23 12:35:47 +01:00
}
func newTargetStatusMap ( ) * targetStatusMap {
return & targetStatusMap {
2023-12-06 16:31:25 +01:00
m : make ( map [ * scrapeWork ] * targetStatus ) ,
upByJob : make ( map [ string ] int ) ,
downByJob : make ( map [ string ] int ) ,
2020-02-23 12:35:47 +01:00
}
}
2021-06-18 09:53:10 +02:00
func ( tsm * targetStatusMap ) registerJobNames ( jobNames [ ] string ) {
tsm . mu . Lock ( )
2023-12-06 16:31:25 +01:00
tsm . registerJobsMetricsLocked ( tsm . jobNames , jobNames )
2021-06-18 09:53:10 +02:00
tsm . jobNames = append ( tsm . jobNames [ : 0 ] , jobNames ... )
tsm . mu . Unlock ( )
}
2023-12-06 16:31:25 +01:00
// registerJobsMetricsLocked registers metrics for new jobs and unregisters metrics for removed jobs
//
// tsm.mu must be locked when calling this function.
func ( tsm * targetStatusMap ) registerJobsMetricsLocked ( prevJobNames , currentJobNames [ ] string ) {
prevNames := make ( map [ string ] struct { } , len ( prevJobNames ) )
currentNames := make ( map [ string ] struct { } , len ( currentJobNames ) )
for _ , jobName := range currentJobNames {
currentNames [ jobName ] = struct { } { }
}
for _ , jobName := range prevJobNames {
prevNames [ jobName ] = struct { } { }
if _ , ok := currentNames [ jobName ] ; ! ok {
metrics . UnregisterMetric ( fmt . Sprintf ( ` vm_promscrape_scrape_pool_targets { scrape_job=%q, status="up"} ` , jobName ) )
metrics . UnregisterMetric ( fmt . Sprintf ( ` vm_promscrape_scrape_pool_targets { scrape_job=%q, status="down"} ` , jobName ) )
2023-12-06 08:44:39 +01:00
}
}
2023-12-06 16:31:25 +01:00
for _ , jobName := range currentJobNames {
if _ , ok := prevNames [ jobName ] ; ok {
continue
2023-12-06 08:44:39 +01:00
}
2023-12-06 16:31:25 +01:00
jobNameLocal := jobName
_ = metrics . NewGauge ( fmt . Sprintf ( ` vm_promscrape_scrape_pool_targets { scrape_job=%q, status="up"} ` , jobName ) , func ( ) float64 {
tsm . mu . Lock ( )
n := tsm . upByJob [ jobNameLocal ]
tsm . mu . Unlock ( )
return float64 ( n )
} )
_ = metrics . NewGauge ( fmt . Sprintf ( ` vm_promscrape_scrape_pool_targets { scrape_job=%q, status="down"} ` , jobName ) , func ( ) float64 {
tsm . mu . Lock ( )
n := tsm . downByJob [ jobNameLocal ]
tsm . mu . Unlock ( )
return float64 ( n )
} )
2023-12-06 08:44:39 +01:00
}
}
2022-02-03 17:57:36 +01:00
func ( tsm * targetStatusMap ) Register ( sw * scrapeWork ) {
2023-12-06 16:31:25 +01:00
jobName := sw . Config . jobNameOriginal
2020-03-11 02:19:56 +01:00
tsm . mu . Lock ( )
2020-12-17 13:30:33 +01:00
tsm . m [ sw ] = & targetStatus {
sw : sw ,
2020-03-11 02:19:56 +01:00
}
2023-12-06 16:31:25 +01:00
tsm . downByJob [ jobName ] ++
2020-03-11 02:19:56 +01:00
tsm . mu . Unlock ( )
}
2022-02-03 17:57:36 +01:00
func ( tsm * targetStatusMap ) Unregister ( sw * scrapeWork ) {
2023-12-06 16:31:25 +01:00
jobName := sw . Config . jobNameOriginal
2020-03-11 02:19:56 +01:00
tsm . mu . Lock ( )
2023-12-06 16:31:25 +01:00
ts , ok := tsm . m [ sw ]
if ! ok {
logger . Panicf ( "BUG: missing Register() call for the target %q" , jobName )
}
if ts . up {
tsm . upByJob [ jobName ] --
} else {
tsm . downByJob [ jobName ] --
}
2020-12-17 13:30:33 +01:00
delete ( tsm . m , sw )
2020-03-11 02:19:56 +01:00
tsm . mu . Unlock ( )
}
2024-07-16 12:24:14 +02:00
func ( tsm * targetStatusMap ) Update ( sw * scrapeWork , up bool , scrapeTime , scrapeDuration int64 , scrapeResponseSize , samplesScraped int , err error ) {
2023-12-06 16:31:25 +01:00
jobName := sw . Config . jobNameOriginal
2020-02-23 12:35:47 +01:00
tsm . mu . Lock ( )
2023-12-06 16:31:25 +01:00
ts , ok := tsm . m [ sw ]
if ! ok {
logger . Panicf ( "BUG: missing Register() call for the target %q" , jobName )
}
if up && ! ts . up {
tsm . upByJob [ jobName ] ++
tsm . downByJob [ jobName ] --
} else if ! up && ts . up {
tsm . upByJob [ jobName ] --
tsm . downByJob [ jobName ] ++
2020-02-23 12:35:47 +01:00
}
2020-12-17 13:30:33 +01:00
ts . up = up
ts . scrapeTime = scrapeTime
ts . scrapeDuration = scrapeDuration
2021-06-14 13:01:13 +02:00
ts . samplesScraped = samplesScraped
2024-06-20 13:58:42 +02:00
ts . scrapeResponseSize = scrapeResponseSize
2022-02-03 19:22:35 +01:00
ts . scrapesTotal ++
if ! up {
ts . scrapesFailed ++
}
2020-12-17 13:30:33 +01:00
ts . err = err
2020-02-23 12:35:47 +01:00
tsm . mu . Unlock ( )
}
2022-02-03 17:57:36 +01:00
func ( tsm * targetStatusMap ) getScrapeWorkByTargetID ( targetID string ) * scrapeWork {
tsm . mu . Lock ( )
defer tsm . mu . Unlock ( )
for sw := range tsm . m {
2022-12-10 11:09:21 +01:00
// The target is uniquely identified by a pointer to its original labels.
if getLabelsID ( sw . Config . OriginalLabels ) == targetID {
2022-02-03 17:57:36 +01:00
return sw
}
}
return nil
}
2022-12-10 11:09:21 +01:00
func getLabelsID ( labels * promutils . Labels ) string {
return fmt . Sprintf ( "%016x" , uintptr ( unsafe . Pointer ( labels ) ) )
2022-02-03 17:57:36 +01:00
}
2020-07-13 20:52:03 +02:00
// StatusByGroup returns the number of targets with status==up
// for the given group name
func ( tsm * targetStatusMap ) StatusByGroup ( group string , up bool ) int {
var count int
tsm . mu . Lock ( )
2022-06-06 23:57:05 +02:00
for _ , ts := range tsm . m {
if ts . sw . ScrapeGroup == group && ts . up == up {
2020-07-13 20:52:03 +02:00
count ++
}
}
tsm . mu . Unlock ( )
return count
}
2022-06-06 23:57:05 +02:00
func ( tsm * targetStatusMap ) getActiveTargetStatuses ( ) [ ] targetStatus {
2020-10-20 20:44:59 +02:00
tsm . mu . Lock ( )
2022-06-06 23:57:05 +02:00
tss := make ( [ ] targetStatus , 0 , len ( tsm . m ) )
for _ , ts := range tsm . m {
tss = append ( tss , * ts )
2020-10-20 20:44:59 +02:00
}
tsm . mu . Unlock ( )
2022-06-06 23:57:05 +02:00
// Sort discovered targets by __address__ label, so they stay in consistent order across calls
sort . Slice ( tss , func ( i , j int ) bool {
2022-11-30 06:22:12 +01:00
addr1 := tss [ i ] . sw . Config . OriginalLabels . Get ( "__address__" )
addr2 := tss [ j ] . sw . Config . OriginalLabels . Get ( "__address__" )
2022-06-06 23:57:05 +02:00
return addr1 < addr2
2020-10-20 20:44:59 +02:00
} )
2022-06-06 23:57:05 +02:00
return tss
2022-06-03 15:38:45 +02:00
}
// WriteActiveTargetsJSON writes `activeTargets` contents to w according to https://prometheus.io/docs/prometheus/latest/querying/api/#targets
func ( tsm * targetStatusMap ) WriteActiveTargetsJSON ( w io . Writer ) {
2022-06-06 23:57:05 +02:00
tss := tsm . getActiveTargetStatuses ( )
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` [ ` )
2022-06-06 23:57:05 +02:00
for i , ts := range tss {
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` { "discoveredLabels": ` )
2022-06-06 23:57:05 +02:00
writeLabelsJSON ( w , ts . sw . Config . OriginalLabels )
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` ,"labels": ` )
2022-10-07 21:39:28 +02:00
writeLabelsJSON ( w , ts . sw . Config . Labels )
2024-07-17 13:52:10 +02:00
fmt . Fprintf ( w , ` ,"scrapePool":%s ` , stringsutil . JSONString ( ts . sw . Config . Job ( ) ) )
fmt . Fprintf ( w , ` ,"scrapeUrl":%s ` , stringsutil . JSONString ( ts . sw . Config . ScrapeURL ) )
2020-10-20 20:44:59 +02:00
errMsg := ""
2022-06-06 23:57:05 +02:00
if ts . err != nil {
errMsg = ts . err . Error ( )
2020-10-20 20:44:59 +02:00
}
2024-07-17 13:52:10 +02:00
fmt . Fprintf ( w , ` ,"lastError":%s ` , stringsutil . JSONString ( errMsg ) )
fmt . Fprintf ( w , ` ,"lastScrape":"%s" ` , time . Unix ( ts . scrapeTime / 1000 , ( ts . scrapeTime % 1000 ) * 1e6 ) . Format ( time . RFC3339Nano ) )
2022-06-06 23:57:05 +02:00
fmt . Fprintf ( w , ` ,"lastScrapeDuration":%g ` , ( time . Millisecond * time . Duration ( ts . scrapeDuration ) ) . Seconds ( ) )
fmt . Fprintf ( w , ` ,"lastSamplesScraped":%d ` , ts . samplesScraped )
2020-10-20 20:44:59 +02:00
state := "up"
2022-06-06 23:57:05 +02:00
if ! ts . up {
2020-10-20 20:44:59 +02:00
state = "down"
}
2024-07-17 13:52:10 +02:00
fmt . Fprintf ( w , ` ,"health":%s} ` , stringsutil . JSONString ( state ) )
2022-06-06 23:57:05 +02:00
if i + 1 < len ( tss ) {
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` , ` )
}
}
fmt . Fprintf ( w , ` ] ` )
}
2022-11-30 06:22:12 +01:00
func writeLabelsJSON ( w io . Writer , labels * promutils . Labels ) {
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` { ` )
2022-11-30 06:22:12 +01:00
labelsList := labels . GetLabels ( )
for i , label := range labelsList {
2024-07-17 13:52:10 +02:00
fmt . Fprintf ( w , "%s:%s" , stringsutil . JSONString ( label . Name ) , stringsutil . JSONString ( label . Value ) )
2022-11-30 06:22:12 +01:00
if i + 1 < len ( labelsList ) {
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` , ` )
}
}
fmt . Fprintf ( w , ` } ` )
}
2020-02-23 12:35:47 +01:00
type targetStatus struct {
2024-06-20 13:58:42 +02:00
sw * scrapeWork
up bool
scrapeTime int64
scrapeDuration int64
2024-07-16 12:24:14 +02:00
scrapeResponseSize int
2024-06-20 13:58:42 +02:00
samplesScraped int
scrapesTotal int
scrapesFailed int
err error
2020-02-23 12:35:47 +01:00
}
2023-12-06 21:24:24 +01:00
func ( ts * targetStatus ) getDurationFromLastScrape ( ) string {
if ts . scrapeTime <= 0 {
return "never scraped"
}
d := time . Since ( time . Unix ( ts . scrapeTime / 1000 , ( ts . scrapeTime % 1000 ) * 1e6 ) )
return fmt . Sprintf ( "%.3fs ago" , d . Seconds ( ) )
2020-02-23 12:35:47 +01:00
}
2020-10-20 20:44:59 +02:00
2024-06-20 13:58:42 +02:00
func ( ts * targetStatus ) getSizeFromLastScrape ( ) string {
if ts . scrapeResponseSize <= 0 {
return "never scraped"
}
2024-07-16 12:24:14 +02:00
return fmt . Sprintf ( "%.3fKiB" , float64 ( ts . scrapeResponseSize ) / 1024 )
2024-06-20 13:58:42 +02:00
}
2022-06-06 23:57:05 +02:00
type droppedTargets struct {
2023-12-01 15:37:57 +01:00
mu sync . Mutex
m map [ uint64 ] droppedTarget
2023-12-08 18:03:49 +01:00
// totalTargets contains the total number of dropped targets registered via Register() call.
totalTargets int
2022-06-06 23:57:05 +02:00
}
2020-10-20 20:44:59 +02:00
2022-06-06 23:57:05 +02:00
type droppedTarget struct {
2023-12-06 23:05:29 +01:00
originalLabels * promutils . Labels
relabelConfigs * promrelabel . ParsedConfigs
dropReason targetDropReason
clusterMemberNums [ ] int
2022-06-06 23:57:05 +02:00
}
2023-12-01 15:37:57 +01:00
type targetDropReason string
const (
targetDropReasonRelabeling = targetDropReason ( "relabeling" ) // target dropped because of relabeling
targetDropReasonMissingScrapeURL = targetDropReason ( "missing scrape URL" ) // target dropped because of missing scrape URL
targetDropReasonDuplicate = targetDropReason ( "duplicate" ) // target with the given set of labels already exists
2024-04-18 01:31:37 +02:00
targetDropReasonSharding = targetDropReason ( "sharding" ) // target is dropped becase of sharding https://docs.victoriametrics.com/vmagent/#scraping-big-number-of-targets
2023-12-01 15:37:57 +01:00
)
2022-12-10 11:09:21 +01:00
func ( dt * droppedTargets ) getTargetsList ( ) [ ] droppedTarget {
2022-06-03 15:38:45 +02:00
dt . mu . Lock ( )
2022-12-10 11:09:21 +01:00
dts := make ( [ ] droppedTarget , 0 , len ( dt . m ) )
2022-06-03 15:38:45 +02:00
for _ , v := range dt . m {
2022-12-10 11:09:21 +01:00
dts = append ( dts , v )
2022-06-03 15:38:45 +02:00
}
dt . mu . Unlock ( )
2022-06-06 23:57:05 +02:00
// Sort discovered targets by __address__ label, so they stay in consistent order across calls
2022-12-10 11:09:21 +01:00
sort . Slice ( dts , func ( i , j int ) bool {
addr1 := dts [ i ] . originalLabels . Get ( "__address__" )
addr2 := dts [ j ] . originalLabels . Get ( "__address__" )
2022-06-06 23:57:05 +02:00
return addr1 < addr2
2022-06-03 15:38:45 +02:00
} )
2022-12-10 11:09:21 +01:00
return dts
2020-10-20 20:44:59 +02:00
}
2023-12-01 15:37:57 +01:00
// Register registers dropped target with the given originalLabels.
//
// The relabelConfigs must contain relabel configs, which were applied to originalLabels.
// The reason must contain the reason why the target has been dropped.
2023-12-06 23:05:29 +01:00
func ( dt * droppedTargets ) Register ( originalLabels * promutils . Labels , relabelConfigs * promrelabel . ParsedConfigs , reason targetDropReason , clusterMemberNums [ ] int ) {
2023-12-01 15:37:57 +01:00
if originalLabels == nil {
// Do not register target without originalLabels. This is the case when *dropOriginalLabels is set to true.
2023-01-05 11:49:26 +01:00
return
}
2022-12-10 11:09:21 +01:00
// It is better to have hash collisions instead of spending additional CPU on originalLabels.String() call.
2022-04-20 14:21:12 +02:00
key := labelsHash ( originalLabels )
2020-10-20 20:44:59 +02:00
dt . mu . Lock ( )
2023-12-08 18:03:49 +01:00
if _ , ok := dt . m [ key ] ; ! ok {
dt . totalTargets ++
}
2023-12-01 15:37:57 +01:00
dt . m [ key ] = droppedTarget {
2023-12-06 23:05:29 +01:00
originalLabels : originalLabels ,
relabelConfigs : relabelConfigs ,
dropReason : reason ,
clusterMemberNums : clusterMemberNums ,
2023-12-01 15:37:57 +01:00
}
2023-12-08 18:03:49 +01:00
if len ( dt . m ) > * maxDroppedTargets {
2023-12-01 15:37:57 +01:00
for k := range dt . m {
delete ( dt . m , k )
2023-12-08 18:03:49 +01:00
if len ( dt . m ) <= * maxDroppedTargets {
2023-12-01 15:37:57 +01:00
break
2020-10-20 20:44:59 +02:00
}
}
}
dt . mu . Unlock ( )
}
2023-12-08 18:03:49 +01:00
func ( dt * droppedTargets ) getTotalTargets ( ) int {
dt . mu . Lock ( )
n := dt . totalTargets
dt . mu . Unlock ( )
return n
}
2022-11-30 06:22:12 +01:00
func labelsHash ( labels * promutils . Labels ) uint64 {
2022-04-20 14:21:12 +02:00
d := xxhashPool . Get ( ) . ( * xxhash . Digest )
2023-01-05 11:49:26 +01:00
for _ , label := range labels . GetLabels ( ) {
2022-04-20 14:21:12 +02:00
_ , _ = d . WriteString ( label . Name )
_ , _ = d . WriteString ( label . Value )
}
h := d . Sum64 ( )
d . Reset ( )
xxhashPool . Put ( d )
return h
}
var xxhashPool = & sync . Pool {
2024-07-10 00:14:15 +02:00
New : func ( ) any {
2022-04-20 14:21:12 +02:00
return xxhash . New ( )
} ,
}
2020-10-20 20:44:59 +02:00
// WriteDroppedTargetsJSON writes `droppedTargets` contents to w according to https://prometheus.io/docs/prometheus/latest/querying/api/#targets
func ( dt * droppedTargets ) WriteDroppedTargetsJSON ( w io . Writer ) {
2022-12-10 11:09:21 +01:00
dts := dt . getTargetsList ( )
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` [ ` )
2022-12-10 11:09:21 +01:00
for i , dt := range dts {
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` { "discoveredLabels": ` )
2022-12-10 11:09:21 +01:00
writeLabelsJSON ( w , dt . originalLabels )
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` } ` )
2022-12-10 11:09:21 +01:00
if i + 1 < len ( dts ) {
2020-10-20 20:44:59 +02:00
fmt . Fprintf ( w , ` , ` )
}
}
fmt . Fprintf ( w , ` ] ` )
}
var droppedTargetsMap = & droppedTargets {
2022-04-20 14:21:12 +02:00
m : make ( map [ uint64 ] droppedTarget ) ,
2020-10-20 20:44:59 +02:00
}
2020-12-14 12:36:48 +01:00
type jobTargetsStatuses struct {
2022-06-06 23:57:05 +02:00
jobName string
2020-12-14 12:36:48 +01:00
upCount int
targetsTotal int
2022-02-03 19:22:35 +01:00
targetsStatus [ ] targetStatus
2020-12-14 12:36:48 +01:00
}
2022-06-06 23:57:05 +02:00
func ( tsm * targetStatusMap ) getTargetsStatusByJob ( filter * requestFilter ) * targetsStatusResult {
2020-12-14 12:36:48 +01:00
byJob := make ( map [ string ] [ ] targetStatus )
tsm . mu . Lock ( )
2022-06-06 23:57:05 +02:00
for _ , ts := range tsm . m {
jobName := ts . sw . Config . jobNameOriginal
2023-12-06 08:44:39 +01:00
if filter . originalJobName != "" && jobName != filter . originalJobName {
continue
}
2022-06-06 23:57:05 +02:00
byJob [ jobName ] = append ( byJob [ jobName ] , * ts )
2020-12-14 12:36:48 +01:00
}
2021-06-18 09:53:10 +02:00
jobNames := append ( [ ] string { } , tsm . jobNames ... )
2020-12-14 12:36:48 +01:00
tsm . mu . Unlock ( )
2022-06-06 23:57:05 +02:00
var jts [ ] * jobTargetsStatuses
for jobName , statuses := range byJob {
2020-12-14 12:36:48 +01:00
sort . Slice ( statuses , func ( i , j int ) bool {
2022-02-03 17:57:36 +01:00
return statuses [ i ] . sw . Config . ScrapeURL < statuses [ j ] . sw . Config . ScrapeURL
2020-12-14 12:36:48 +01:00
} )
ups := 0
2022-02-03 19:22:35 +01:00
var targetsStatuses [ ] targetStatus
2020-12-14 12:36:48 +01:00
for _ , ts := range statuses {
if ts . up {
ups ++
}
2022-06-06 23:57:05 +02:00
if filter . showOnlyUnhealthy && ts . up {
continue
}
2022-02-03 19:22:35 +01:00
targetsStatuses = append ( targetsStatuses , ts )
2020-12-14 12:36:48 +01:00
}
2022-06-06 23:57:05 +02:00
jts = append ( jts , & jobTargetsStatuses {
jobName : jobName ,
2020-12-14 12:36:48 +01:00
upCount : ups ,
targetsTotal : len ( statuses ) ,
targetsStatus : targetsStatuses ,
} )
}
sort . Slice ( jts , func ( i , j int ) bool {
2022-06-06 23:57:05 +02:00
return jts [ i ] . jobName < jts [ j ] . jobName
2020-12-14 12:36:48 +01:00
} )
2021-06-18 09:53:10 +02:00
emptyJobs := getEmptyJobs ( jts , jobNames )
2022-04-19 17:26:21 +02:00
var err error
2022-06-06 23:57:05 +02:00
jts , err = filterTargets ( jts , filter . endpointSearch , filter . labelSearch )
if len ( filter . endpointSearch ) > 0 || len ( filter . labelSearch ) > 0 {
2022-04-19 17:26:21 +02:00
// Do not show empty jobs if target filters are set.
emptyJobs = nil
}
2022-12-10 11:09:21 +01:00
dts := droppedTargetsMap . getTargetsList ( )
2022-06-06 23:57:05 +02:00
return & targetsStatusResult {
2023-07-21 04:14:30 +02:00
hasOriginalLabels : ! * dropOriginalLabels ,
2022-12-10 11:09:21 +01:00
jobTargetsStatuses : jts ,
droppedTargets : dts ,
emptyJobs : emptyJobs ,
err : err ,
2022-06-06 23:57:05 +02:00
}
2022-04-19 17:26:21 +02:00
}
2022-06-06 23:57:05 +02:00
func filterTargetsByEndpoint ( jts [ ] * jobTargetsStatuses , searchQuery string ) ( [ ] * jobTargetsStatuses , error ) {
2022-04-19 17:26:21 +02:00
if searchQuery == "" {
return jts , nil
}
finder , err := regexp . Compile ( searchQuery )
if err != nil {
return nil , fmt . Errorf ( "cannot parse %s: %w" , searchQuery , err )
}
2022-06-06 23:57:05 +02:00
var jtsFiltered [ ] * jobTargetsStatuses
2022-04-19 17:26:21 +02:00
for _ , job := range jts {
var tss [ ] targetStatus
for _ , ts := range job . targetsStatus {
if finder . MatchString ( ts . sw . Config . ScrapeURL ) {
tss = append ( tss , ts )
}
}
if len ( tss ) == 0 {
// Skip jobs with zero targets after filtering, so users could see only the requested targets
continue
}
job . targetsStatus = tss
jtsFiltered = append ( jtsFiltered , job )
}
return jtsFiltered , nil
}
2022-06-06 23:57:05 +02:00
func filterTargetsByLabels ( jts [ ] * jobTargetsStatuses , searchQuery string ) ( [ ] * jobTargetsStatuses , error ) {
2022-04-19 17:26:21 +02:00
if searchQuery == "" {
return jts , nil
}
var ie promrelabel . IfExpression
if err := ie . Parse ( searchQuery ) ; err != nil {
return nil , fmt . Errorf ( "cannot parse %s: %w" , searchQuery , err )
}
2022-06-06 23:57:05 +02:00
var jtsFiltered [ ] * jobTargetsStatuses
2022-04-19 17:26:21 +02:00
for _ , job := range jts {
var tss [ ] targetStatus
for _ , ts := range job . targetsStatus {
2022-11-30 06:22:12 +01:00
labels := ts . sw . Config . Labels . GetLabels ( )
if ie . Match ( labels ) {
2022-04-19 17:26:21 +02:00
tss = append ( tss , ts )
}
}
if len ( tss ) == 0 {
// Skip jobs with zero targets after filtering, so users could see only the requested targets
continue
}
job . targetsStatus = tss
jtsFiltered = append ( jtsFiltered , job )
}
return jtsFiltered , nil
}
2022-06-06 23:57:05 +02:00
func filterTargets ( jts [ ] * jobTargetsStatuses , endpointQuery , labelQuery string ) ( [ ] * jobTargetsStatuses , error ) {
2022-04-19 17:26:21 +02:00
var err error
jts , err = filterTargetsByEndpoint ( jts , endpointQuery )
if err != nil {
return nil , err
}
jts , err = filterTargetsByLabels ( jts , labelQuery )
if err != nil {
return nil , err
}
return jts , nil
2021-06-18 09:53:10 +02:00
}
2022-06-06 23:57:05 +02:00
func getEmptyJobs ( jts [ ] * jobTargetsStatuses , jobNames [ ] string ) [ ] string {
2021-06-18 09:53:10 +02:00
jobNamesMap := make ( map [ string ] struct { } , len ( jobNames ) )
for _ , jobName := range jobNames {
jobNamesMap [ jobName ] = struct { } { }
}
for i := range jts {
2022-06-06 23:57:05 +02:00
delete ( jobNamesMap , jts [ i ] . jobName )
2021-06-18 09:53:10 +02:00
}
emptyJobs := make ( [ ] string , 0 , len ( jobNamesMap ) )
for k := range jobNamesMap {
emptyJobs = append ( emptyJobs , k )
}
sort . Strings ( emptyJobs )
return emptyJobs
2020-12-14 12:36:48 +01:00
}
2022-06-06 23:57:05 +02:00
type requestFilter struct {
showOriginalLabels bool
showOnlyUnhealthy bool
endpointSearch string
labelSearch string
2023-12-06 08:44:39 +01:00
originalJobName string
2022-06-06 23:57:05 +02:00
}
func getRequestFilter ( r * http . Request ) * requestFilter {
showOriginalLabels , _ := strconv . ParseBool ( r . FormValue ( "show_original_labels" ) )
showOnlyUnhealthy , _ := strconv . ParseBool ( r . FormValue ( "show_only_unhealthy" ) )
endpointSearch := strings . TrimSpace ( r . FormValue ( "endpoint_search" ) )
labelSearch := strings . TrimSpace ( r . FormValue ( "label_search" ) )
return & requestFilter {
showOriginalLabels : showOriginalLabels ,
showOnlyUnhealthy : showOnlyUnhealthy ,
endpointSearch : endpointSearch ,
labelSearch : labelSearch ,
}
}
type targetsStatusResult struct {
2023-07-21 04:14:30 +02:00
hasOriginalLabels bool
2022-12-10 11:09:21 +01:00
jobTargetsStatuses [ ] * jobTargetsStatuses
droppedTargets [ ] droppedTarget
emptyJobs [ ] string
err error
2022-06-06 23:57:05 +02:00
}
type targetLabels struct {
2023-12-06 23:05:29 +01:00
up bool
originalLabels * promutils . Labels
labels * promutils . Labels
dropReason targetDropReason
clusterMemberNums [ ] int
2022-06-06 23:57:05 +02:00
}
type targetLabelsByJob struct {
jobName string
targets [ ] targetLabels
activeTargets int
droppedTargets int
}
2023-01-05 12:26:47 +01:00
func getMetricRelabelContextByTargetID ( targetID string ) ( * promrelabel . ParsedConfigs , * promutils . Labels , bool ) {
tsmGlobal . mu . Lock ( )
defer tsmGlobal . mu . Unlock ( )
for sw := range tsmGlobal . m {
// The target is uniquely identified by a pointer to its original labels.
if getLabelsID ( sw . Config . OriginalLabels ) == targetID {
return sw . Config . MetricRelabelConfigs , sw . Config . Labels , true
}
}
return nil , nil , false
}
func getTargetRelabelContextByTargetID ( targetID string ) ( * promrelabel . ParsedConfigs , * promutils . Labels , bool ) {
2022-12-10 11:09:21 +01:00
var relabelConfigs * promrelabel . ParsedConfigs
var labels * promutils . Labels
found := false
// Search for relabel context in tsmGlobal (aka active targets)
tsmGlobal . mu . Lock ( )
for sw := range tsmGlobal . m {
// The target is uniquely identified by a pointer to its original labels.
if getLabelsID ( sw . Config . OriginalLabels ) == targetID {
relabelConfigs = sw . Config . RelabelConfigs
labels = sw . Config . OriginalLabels
found = true
break
}
}
tsmGlobal . mu . Unlock ( )
if found {
return relabelConfigs , labels , true
}
// Search for relabel context in droppedTargetsMap (aka deleted targets)
droppedTargetsMap . mu . Lock ( )
for _ , dt := range droppedTargetsMap . m {
if getLabelsID ( dt . originalLabels ) == targetID {
relabelConfigs = dt . relabelConfigs
labels = dt . originalLabels
found = true
break
}
}
droppedTargetsMap . mu . Unlock ( )
return relabelConfigs , labels , found
}
2022-06-06 23:57:05 +02:00
func ( tsr * targetsStatusResult ) getTargetLabelsByJob ( ) [ ] * targetLabelsByJob {
byJob := make ( map [ string ] * targetLabelsByJob )
for _ , jts := range tsr . jobTargetsStatuses {
jobName := jts . jobName
for _ , ts := range jts . targetsStatus {
m := byJob [ jobName ]
if m == nil {
m = & targetLabelsByJob {
jobName : jobName ,
}
byJob [ jobName ] = m
}
m . activeTargets ++
m . targets = append ( m . targets , targetLabels {
2022-12-10 11:09:21 +01:00
up : ts . up ,
originalLabels : ts . sw . Config . OriginalLabels ,
labels : ts . sw . Config . Labels ,
2022-06-06 23:57:05 +02:00
} )
}
2022-06-03 15:38:45 +02:00
}
2022-12-10 11:09:21 +01:00
for _ , dt := range tsr . droppedTargets {
jobName := dt . originalLabels . Get ( "job" )
2022-06-06 23:57:05 +02:00
m := byJob [ jobName ]
if m == nil {
m = & targetLabelsByJob {
jobName : jobName ,
}
byJob [ jobName ] = m
}
m . droppedTargets ++
m . targets = append ( m . targets , targetLabels {
2023-12-06 23:05:29 +01:00
originalLabels : dt . originalLabels ,
dropReason : dt . dropReason ,
clusterMemberNums : dt . clusterMemberNums ,
2022-06-06 23:57:05 +02:00
} )
2022-06-03 15:38:45 +02:00
}
2022-06-06 23:57:05 +02:00
a := make ( [ ] * targetLabelsByJob , 0 , len ( byJob ) )
for _ , tls := range byJob {
a = append ( a , tls )
2022-06-03 15:38:45 +02:00
}
2022-06-06 23:57:05 +02:00
sort . Slice ( a , func ( i , j int ) bool {
return a [ i ] . jobName < a [ j ] . jobName
} )
return a
2020-12-14 12:36:48 +01:00
}