github-mirrors/VictoriaMetrics
1
0
Fork 0
mirror of https://github.com/VictoriaMetrics/VictoriaMetrics.git synced 2024-11-21 14:44:00 +00:00
Code Issues Projects Releases Packages Wiki Activity

lib/promscrape: follow-up after 7e79adfb55

Browse source
This commit is contained in:
Aliaksandr Valialkin 2022-04-12 12:36:17 +03:00
parent 7e79adfb55
commit f3d4671bb6
No known key found for this signature in database
GPG key ID: A72BEC6CD3D0DED1
6 changed files with 16 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
app/vmagent/README.md
View file

@ -358,6 +358,8 @@ spread scrape targets among a cluster of two `vmagent` instances:
/path/to/vmagent -promscrape.cluster.membersCount=2 -promscrape.cluster.memberNum=1 -promscrape.config=/path/to/config.yml ...
```
The `-promscrape.cluster.memberNum` can be set to a StatefulSet pod name when `vmagent` runs in Kubernetes. The pod name must end with a number in the range `0 ... promscrape.cluster.memberNum-1`. For example, `-promscrape.cluster.memberNum=vmagent-0`.
By default each scrape target is scraped only by a single `vmagent` instance in the cluster. If there is a need for replicating scrape targets among multiple `vmagent` instances,
then `-promscrape.cluster.replicationFactor` command-line flag must be set to the desired number of replicas. For example, the following commands
start a cluster of three `vmagent` instances, where each target is scraped by two `vmagent` instances:

13
docs/CHANGELOG.md
View file

@ -15,12 +15,15 @@ The following tip changes can be tested by building VictoriaMetrics components f
## tip
FEATURE: [vmalert](https://docs.victoriametrics.com/vmalert.html): add support for `alert_relabel_configs` option at `-notifier.config`. This option allows configuring relabeling rules for alerts before sending them to configured notifiers. See [these docs](https://docs.victoriametrics.com/vmalert.html#notifier-configuration-file) for details.
**Update notes:** this release introduces backwards-incompatible changes to communication protocol between `vmselect` and `vmstorage` nodes in cluster version of VictoriaMetrics, so `vmselect` and `vmstorage` nodes may log communication errors during the upgrade. These errors should stop after all the `vmselect` and `vmstorage` nodes are updated to new release.
BUGFIX: properly propagate limits at `-search.max*` command-line flags from `vminsert` to `vmstorage`. The limits are `-search.maxUniqueTimeseries`, `-search.maxSeries`, `-search.maxFederateSeries`, `-search.maxExportSeries`, `-search.maxGraphiteSeries` and `-search.maxTSDBStatusSeries`. They weren't propagated to `vmstorage` because of the bug. These limits were introduced in [v1.76.0](https://docs.victoriametrics.com/CHANGELOG.html#v1760).
BUGFIX: fix goroutine leak and possible deadlock when importing invalid data via [native binary format](https://docs.victoriametrics.com/#how-to-import-data-in-native-format). See [this pull request](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/2423).
BUGFIX: [Graphite Render API](https://docs.victoriametrics.com/#graphite-render-api-usage): properly calculate [hitCount](https://graphite.readthedocs.io/en/latest/functions.html#graphite.render.functions.hitcount) function. Previously it could return empty results if there were no original samples in some parts of the selected time range.
BUGFIX: [MetricsQL](https://docs.victoriametrics.com/MetricsQL.html): allow overriding built-in function names inside [WITH templates](https://play.victoriametrics.com/promql/expand-with-exprs). For example, `WITH (sum(a,b) = a + b + 1) sum(x,y)` now expands into `x + y + 1`. Previously such a query would fail with `cannot use reserved name` error. See [this bugreport](https://github.com/VictoriaMetrics/metricsql/issues/5).
* FEATURE: [vmalert](https://docs.victoriametrics.com/vmalert.html): add support for `alert_relabel_configs` option at `-notifier.config`. This option allows configuring relabeling rules for alerts before sending them to configured notifiers. See [these docs](https://docs.victoriametrics.com/vmalert.html#notifier-configuration-file) for details.
* FEATURE: [vmagent](https://docs.victoriametrics.com/vmalert.html): allow passing StatefulSet pod names to `-promscrape.cluster.memberNum` command-line flag. In this case the member number is automatically extracted from the pod name, which must end with the number in the range `0 ... promscrape.cluster.membersCount-1`. For example, `vmagent-0`, `vmagent-1`, etc. See [this feature request](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2359) and [these docs](https://docs.victoriametrics.com/vmagent.html#scraping-big-number-of-targets).
* BUGFIX: [VictoriaMetrics cluster](https://docs.victoriametrics.com/Cluster-VictoriaMetrics.html): properly propagate limits at `-search.max*` command-line flags from `vminsert` to `vmstorage`. The limits are `-search.maxUniqueTimeseries`, `-search.maxSeries`, `-search.maxFederateSeries`, `-search.maxExportSeries`, `-search.maxGraphiteSeries` and `-search.maxTSDBStatusSeries`. They weren't propagated to `vmstorage` because of the bug. These limits were introduced in [v1.76.0](https://docs.victoriametrics.com/CHANGELOG.html#v1760).
* BUGFIX: fix goroutine leak and possible deadlock when importing invalid data via [native binary format](https://docs.victoriametrics.com/#how-to-import-data-in-native-format). See [this pull request](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/2423).
* BUGFIX: [Graphite Render API](https://docs.victoriametrics.com/#graphite-render-api-usage): properly calculate [hitCount](https://graphite.readthedocs.io/en/latest/functions.html#graphite.render.functions.hitcount) function. Previously it could return empty results if there were no original samples in some parts of the selected time range.
* BUGFIX: [MetricsQL](https://docs.victoriametrics.com/MetricsQL.html): allow overriding built-in function names inside [WITH templates](https://play.victoriametrics.com/promql/expand-with-exprs). For example, `WITH (sum(a,b) = a + b + 1) sum(x,y)` now expands into `x + y + 1`. Previously such a query would fail with `cannot use reserved name` error. See [this bugreport](https://github.com/VictoriaMetrics/metricsql/issues/5).
## [v1.76.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.76.0)

2
docs/vmagent.md
View file

@ -362,6 +362,8 @@ spread scrape targets among a cluster of two `vmagent` instances:
/path/to/vmagent -promscrape.cluster.membersCount=2 -promscrape.cluster.memberNum=1 -promscrape.config=/path/to/config.yml ...
```
The `-promscrape.cluster.memberNum` can be set to a StatefulSet pod name when `vmagent` runs in Kubernetes. The pod name must end with a number in the range `0 ... promscrape.cluster.memberNum-1`. For example, `-promscrape.cluster.memberNum=vmagent-0`.
By default each scrape target is scraped only by a single `vmagent` instance in the cluster. If there is a need for replicating scrape targets among multiple `vmagent` instances,
then `-promscrape.cluster.replicationFactor` command-line flag must be set to the desired number of replicas. For example, the following commands
start a cluster of three `vmagent` instances, where each target is scraped by two `vmagent` instances:

6
lib/promscrape/config.go
View file

@ -58,8 +58,7 @@ var (
var clusterMemberID int
// must be called before any scraper
func initClusterMemberID() error {
func mustInitClusterMemberID() {
s := *clusterMemberNum
// special case for kubernetes deployment, where pod-name formatted at some-pod-name-1
// obtain memberNum from last segment
@ -69,10 +68,9 @@ func initClusterMemberID() error {
}
n, err := strconv.ParseInt(s, 10, 64)
if err != nil {
return fmt.Errorf("cannot parse -promscrape.cluster.memberNum=%q: %w", *clusterMemberNum, err)
logger.Fatalf("cannot parse -promscrape.cluster.memberNum=%q: %s", *clusterMemberNum, err)
}
clusterMemberID = int(n)
return nil
}
// Config represents essential parts from Prometheus config defined at https://prometheus.io/docs/prometheus/latest/configuration/configuration/

4
lib/promscrape/scraper.go
View file

@ -52,6 +52,7 @@ func CheckConfig() error {
//
// Scraped data is passed to pushData.
func Init(pushData func(wr *prompbmarshal.WriteRequest)) {
mustInitClusterMemberID()
globalStopChan = make(chan struct{})
scraperWG.Add(1)
go func() {
@ -99,9 +100,6 @@ func runScraper(configFile string, pushData func(wr *prompbmarshal.WriteRequest)
// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1240
sighupCh := procutil.NewSighupChan()
if err := initClusterMemberID(); err != nil {
logger.Fatalf("cannot init clusterMembership: %s", err)
}
logger.Infof("reading Prometheus configs from %q", configFile)
cfg, data, err := loadConfig(configFile)
if err != nil {

2
lib/promscrape/scrapework.go
View file

@ -276,7 +276,7 @@ func (sw *scrapeWork) run(stopCh <-chan struct{}, globalStopCh <-chan struct{})
// scrapes replicated targets at different time offsets. This guarantees that the deduplication consistently leaves samples
// received from the same vmagent replica.
// See https://docs.victoriametrics.com/vmagent.html#scraping-big-number-of-targets
key := fmt.Sprintf("ClusterMemberNum=%d, ScrapeURL=%s, Labels=%s", *clusterMemberNum, sw.Config.ScrapeURL, sw.Config.LabelsString())
key := fmt.Sprintf("ClusterMemberNum=%d, ScrapeURL=%s, Labels=%s", clusterMemberID, sw.Config.ScrapeURL, sw.Config.LabelsString())
h := xxhash.Sum64(bytesutil.ToUnsafeBytes(key))
randSleep = uint64(float64(scrapeInterval) * (float64(h) / (1 << 64)))
sleepOffset := uint64(time.Now().UnixNano()) % uint64(scrapeInterval)