MyFavMirrors/minio
1
0
Fork 0
mirror of https://github.com/minio/minio.git synced 2025-05-22 18:11:50 -04:00
Code Issues Packages Projects Releases Wiki Activity

Add cluster and bucket replication metrics in metrics-v3 (#19546)

Browse Source 
endpoint: /minio/metrics/v3/cluster/replication
metrics:
- average_active_workers
- average_queued_bytes
- average_queued_count
- average_transfer_rate
- current_active_workers
- current_transfer_rate
- last_minute_queued_bytes
- last_minute_queued_count
- max_active_workers
- max_queued_bytes
- max_queued_count
- max_transfer_rate
- recent_backlog_count

endpoint: /minio/metrics/v3/api/bucket/replication
metrics:
- last_hour_failed_bytes
- last_hour_failed_count
- last_minute_failed_bytes
- last_minute_failed_count
- latency_ms
- proxied_delete_tagging_requests_total
- proxied_get_requests_failures
- proxied_get_requests_total
- proxied_get_tagging_requests_failures
- proxied_get_tagging_requests_total
- proxied_head_requests_failures
- proxied_head_requests_total
- proxied_put_tagging_requests_failures
- proxied_put_tagging_requests_total
- sent_bytes
- sent_count
- total_failed_bytes
- total_failed_count
- proxied_delete_tagging_requests_failures
This commit is contained in:
Shireesh Anjal 2024-05-23 13:11:18 +05:30 committed by GitHub
parent 6d5bc045bc
commit 7981509cc8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
6 changed files with 395 additions and 45 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
cmd/metrics-v3-api.go
View File

@ -144,33 +144,33 @@ func loadAPIRequestsNetworkMetrics(ctx context.Context, m MetricValues, _ *metri
// Metric Descriptions for bucket level S3 metrics. // Metric Descriptions for bucket level S3 metrics.
var ( var (
apiBucketTrafficSentBytesMD = NewCounterMD(apiTrafficSentBytes, bucketAPITrafficSentBytesMD = NewCounterMD(apiTrafficSentBytes,
"Total number of bytes received for a bucket", "bucket", "type") "Total number of bytes received for a bucket", "bucket", "type")
apiBucketTrafficRecvBytesMD = NewCounterMD(apiTrafficRecvBytes, bucketAPITrafficRecvBytesMD = NewCounterMD(apiTrafficRecvBytes,
"Total number of bytes sent for a bucket", "bucket", "type") "Total number of bytes sent for a bucket", "bucket", "type")
apiBucketRequestsInFlightMD = NewGaugeMD(apiRequestsInFlightTotal, bucketAPIRequestsInFlightMD = NewGaugeMD(apiRequestsInFlightTotal,
"Total number of requests currently in flight for a bucket", "bucket", "name", "type") "Total number of requests currently in flight for a bucket", "bucket", "name", "type")
apiBucketRequestsTotalMD = NewCounterMD(apiRequestsTotal, bucketAPIRequestsTotalMD = NewCounterMD(apiRequestsTotal,
"Total number of requests for a bucket", "bucket", "name", "type") "Total number of requests for a bucket", "bucket", "name", "type")
apiBucketRequestsCanceledMD = NewCounterMD(apiRequestsCanceledTotal, bucketAPIRequestsCanceledMD = NewCounterMD(apiRequestsCanceledTotal,
"Total number of requests canceled by the client for a bucket", "bucket", "name", "type") "Total number of requests canceled by the client for a bucket", "bucket", "name", "type")
apiBucketRequests4xxErrorsMD = NewCounterMD(apiRequests4xxErrorsTotal, bucketAPIRequests4xxErrorsMD = NewCounterMD(apiRequests4xxErrorsTotal,
"Total number of requests with 4xx errors for a bucket", "bucket", "name", "type") "Total number of requests with 4xx errors for a bucket", "bucket", "name", "type")
apiBucketRequests5xxErrorsMD = NewCounterMD(apiRequests5xxErrorsTotal, bucketAPIRequests5xxErrorsMD = NewCounterMD(apiRequests5xxErrorsTotal,
"Total number of requests with 5xx errors for a bucket", "bucket", "name", "type") "Total number of requests with 5xx errors for a bucket", "bucket", "name", "type")
apiBucketRequestsTTFBSecondsDistributionMD = NewCounterMD(apiRequestsTTFBSecondsDistribution, bucketAPIRequestsTTFBSecondsDistributionMD = NewCounterMD(apiRequestsTTFBSecondsDistribution,
"Distribution of time to first byte across API calls for a bucket", "Distribution of time to first byte across API calls for a bucket",
"bucket", "name", "le", "type") "bucket", "name", "le", "type")
) )
// loadAPIBucketHTTPMetrics - loads bucket level S3 HTTP metrics. // loadBucketAPIHTTPMetrics - loads bucket level S3 HTTP metrics.
// //
// This is a `MetricsLoaderFn`. // This is a `MetricsLoaderFn`.
// //
// This includes bucket level S3 HTTP metrics and S3 network in/out metrics. // This includes bucket level S3 HTTP metrics and S3 network in/out metrics.
func loadAPIBucketHTTPMetrics(ctx context.Context, m MetricValues, _ *metricsCache, buckets []string) error { func loadBucketAPIHTTPMetrics(ctx context.Context, m MetricValues, _ *metricsCache, buckets []string) error {
if len(buckets) == 0 { if len(buckets) == 0 {
return nil return nil
} }
@ -209,10 +209,10 @@ func loadAPIBucketHTTPMetrics(ctx context.Context, m MetricValues, _ *metricsCac
return nil return nil
} }
// loadAPIBucketTTFBMetrics - loads bucket S3 TTFB metrics. // loadBucketAPITTFBMetrics - loads bucket S3 TTFB metrics.
// //
// This is a `MetricsLoaderFn`. // This is a `MetricsLoaderFn`.
func loadAPIBucketTTFBMetrics(ctx context.Context, m MetricValues, _ *metricsCache, buckets []string) error { func loadBucketAPITTFBMetrics(ctx context.Context, m MetricValues, _ *metricsCache, buckets []string) error {
renameLabels := map[string]string{"api": "name"} renameLabels := map[string]string{"api": "name"}
m.SetHistogram(apiRequestsTTFBSecondsDistribution, bucketHTTPRequestsDuration, renameLabels, m.SetHistogram(apiRequestsTTFBSecondsDistribution, bucketHTTPRequestsDuration, renameLabels,
buckets, "type", "s3") buckets, "type", "s3")

155
cmd/metrics-v3-bucket-replication.go Normal file
View File

@ -0,0 +1,155 @@
// Copyright (c) 2015-2024 MinIO, Inc.
//
// This file is part of MinIO Object Storage stack
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package cmd
import (
"context"
)
const (
bucketReplLastHrFailedBytes = "last_hour_failed_bytes"
bucketReplLastHrFailedCount = "last_hour_failed_count"
bucketReplLastMinFailedBytes = "last_minute_failed_bytes"
bucketReplLastMinFailedCount = "last_minute_failed_count"
bucketReplLatencyMs = "latency_ms"
bucketReplProxiedDeleteTaggingRequestsTotal = "proxied_delete_tagging_requests_total"
bucketReplProxiedGetRequestsFailures = "proxied_get_requests_failures"
bucketReplProxiedGetRequestsTotal = "proxied_get_requests_total"
bucketReplProxiedGetTaggingRequestsFailures = "proxied_get_tagging_requests_failures"
bucketReplProxiedGetTaggingRequestsTotal = "proxied_get_tagging_requests_total"
bucketReplProxiedHeadRequestsFailures = "proxied_head_requests_failures"
bucketReplProxiedHeadRequestsTotal = "proxied_head_requests_total"
bucketReplProxiedPutTaggingRequestsFailures = "proxied_put_tagging_requests_failures"
bucketReplProxiedPutTaggingRequestsTotal = "proxied_put_tagging_requests_total"
bucketReplSentBytes = "sent_bytes"
bucketReplSentCount = "sent_count"
bucketReplTotalFailedBytes = "total_failed_bytes"
bucketReplTotalFailedCount = "total_failed_count"
bucketReplProxiedDeleteTaggingRequestsFailures = "proxied_delete_tagging_requests_failures"
bucketL = "bucket"
operationL = "operation"
targetArnL = "targetArn"
)
var (
bucketReplLastHrFailedBytesMD = NewGaugeMD(bucketReplLastHrFailedBytes,
"Total number of bytes failed at least once to replicate in the last hour on a bucket",
bucketL)
bucketReplLastHrFailedCountMD = NewGaugeMD(bucketReplLastHrFailedCount,
"Total number of objects which failed replication in the last hour on a bucket",
bucketL)
bucketReplLastMinFailedBytesMD = NewGaugeMD(bucketReplLastMinFailedBytes,
"Total number of bytes failed at least once to replicate in the last full minute on a bucket",
bucketL)
bucketReplLastMinFailedCountMD = NewGaugeMD(bucketReplLastMinFailedCount,
"Total number of objects which failed replication in the last full minute on a bucket",
bucketL)
bucketReplLatencyMsMD = NewGaugeMD(bucketReplLatencyMs,
"Replication latency on a bucket in milliseconds",
bucketL, operationL, rangeL, targetArnL)
bucketReplProxiedDeleteTaggingRequestsTotalMD = NewCounterMD(bucketReplProxiedDeleteTaggingRequestsTotal,
"Number of DELETE tagging requests proxied to replication target",
bucketL)
bucketReplProxiedGetRequestsFailuresMD = NewCounterMD(bucketReplProxiedGetRequestsFailures,
"Number of failures in GET requests proxied to replication target",
bucketL)
bucketReplProxiedGetRequestsTotalMD = NewCounterMD(bucketReplProxiedGetRequestsTotal,
"Number of GET requests proxied to replication target",
bucketL)
bucketReplProxiedGetTaggingRequestsFailuresMD = NewCounterMD(bucketReplProxiedGetTaggingRequestsFailures,
"Number of failures in GET tagging requests proxied to replication target",
bucketL)
bucketReplProxiedGetTaggingRequestsTotalMD = NewCounterMD(bucketReplProxiedGetTaggingRequestsTotal,
"Number of GET tagging requests proxied to replication target",
bucketL)
bucketReplProxiedHeadRequestsFailuresMD = NewCounterMD(bucketReplProxiedHeadRequestsFailures,
"Number of failures in HEAD requests proxied to replication target",
bucketL)
bucketReplProxiedHeadRequestsTotalMD = NewCounterMD(bucketReplProxiedHeadRequestsTotal,
"Number of HEAD requests proxied to replication target",
bucketL)
bucketReplProxiedPutTaggingRequestsFailuresMD = NewCounterMD(bucketReplProxiedPutTaggingRequestsFailures,
"Number of failures in PUT tagging requests proxied to replication target",
bucketL)
bucketReplProxiedPutTaggingRequestsTotalMD = NewCounterMD(bucketReplProxiedPutTaggingRequestsTotal,
"Number of PUT tagging requests proxied to replication target",
bucketL)
bucketReplSentBytesMD = NewCounterMD(bucketReplSentBytes,
"Total number of bytes replicated to the target",
bucketL)
bucketReplSentCountMD = NewCounterMD(bucketReplSentCount,
"Total number of objects replicated to the target",
bucketL)
bucketReplTotalFailedBytesMD = NewCounterMD(bucketReplTotalFailedBytes,
"Total number of bytes failed at least once to replicate since server start",
bucketL)
bucketReplTotalFailedCountMD = NewCounterMD(bucketReplTotalFailedCount,
"Total number of objects which failed replication since server start",
bucketL)
bucketReplProxiedDeleteTaggingRequestsFailuresMD = NewCounterMD(bucketReplProxiedDeleteTaggingRequestsFailures,
"Number of failures in DELETE tagging requests proxied to replication target",
bucketL)
)
// loadBucketReplicationMetrics - `BucketMetricsLoaderFn` for bucket replication metrics
// such as latency and sent bytes.
func loadBucketReplicationMetrics(ctx context.Context, m MetricValues, c *metricsCache, buckets []string) error {
if globalSiteReplicationSys.isEnabled() {
return nil
}
dataUsageInfo, err := c.dataUsageInfo.Get()
if err != nil {
metricsLogIf(ctx, err)
return nil
}
bucketReplStats := globalReplicationStats.getAllLatest(dataUsageInfo.BucketsUsage)
for _, bucket := range buckets {
labels := []string{bucketL, bucket}
if s, ok := bucketReplStats[bucket]; ok {
stats := s.ReplicationStats
if stats.hasReplicationUsage() {
for arn, stat := range stats.Stats {
m.Set(bucketReplLastHrFailedBytes, float64(stat.Failed.LastHour.Bytes), labels...)
m.Set(bucketReplLastHrFailedCount, float64(stat.Failed.LastHour.Count), labels...)
m.Set(bucketReplLastMinFailedBytes, float64(stat.Failed.LastMinute.Bytes), labels...)
m.Set(bucketReplLastMinFailedCount, float64(stat.Failed.LastMinute.Count), labels...)
m.Set(bucketReplProxiedDeleteTaggingRequestsTotal, float64(s.ProxyStats.RmvTagTotal), labels...)
m.Set(bucketReplProxiedGetRequestsFailures, float64(s.ProxyStats.GetFailedTotal), labels...)
m.Set(bucketReplProxiedGetRequestsTotal, float64(s.ProxyStats.GetTotal), labels...)
m.Set(bucketReplProxiedGetTaggingRequestsFailures, float64(s.ProxyStats.GetTagFailedTotal), labels...)
m.Set(bucketReplProxiedGetTaggingRequestsTotal, float64(s.ProxyStats.GetTagTotal), labels...)
m.Set(bucketReplProxiedHeadRequestsFailures, float64(s.ProxyStats.HeadFailedTotal), labels...)
m.Set(bucketReplProxiedHeadRequestsTotal, float64(s.ProxyStats.HeadTotal), labels...)
m.Set(bucketReplProxiedPutTaggingRequestsFailures, float64(s.ProxyStats.PutTagFailedTotal), labels...)
m.Set(bucketReplProxiedPutTaggingRequestsTotal, float64(s.ProxyStats.PutTagTotal), labels...)
m.Set(bucketReplSentCount, float64(stat.ReplicatedCount), labels...)
m.Set(bucketReplTotalFailedBytes, float64(stat.Failed.Totals.Bytes), labels...)
m.Set(bucketReplTotalFailedCount, float64(stat.Failed.Totals.Count), labels...)
m.Set(bucketReplProxiedDeleteTaggingRequestsFailures, float64(s.ProxyStats.RmvTagFailedTotal), labels...)
m.Set(bucketReplSentBytes, float64(stat.ReplicatedSize), labels...)
SetHistogramValues(m, bucketReplLatencyMs, stat.Latency.getUploadLatency(), bucketL, bucket, operationL, "upload", targetArnL, arn)
}
}
}
}
return nil
}

96
cmd/metrics-v3-replication.go Normal file
View File

@ -0,0 +1,96 @@
// Copyright (c) 2015-2024 MinIO, Inc.
//
// This file is part of MinIO Object Storage stack
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package cmd
import (
"context"
)
const (
replicationAverageActiveWorkers = "average_active_workers"
replicationAverageQueuedBytes = "average_queued_bytes"
replicationAverageQueuedCount = "average_queued_count"
replicationAverageDataTransferRate = "average_data_transfer_rate"
replicationCurrentActiveWorkers = "current_active_workers"
replicationCurrentDataTransferRate = "current_data_transfer_rate"
replicationLastMinuteQueuedBytes = "last_minute_queued_bytes"
replicationLastMinuteQueuedCount = "last_minute_queued_count"
replicationMaxActiveWorkers = "max_active_workers"
replicationMaxQueuedBytes = "max_queued_bytes"
replicationMaxQueuedCount = "max_queued_count"
replicationMaxDataTransferRate = "max_data_transfer_rate"
)
var (
replicationAverageActiveWorkersMD = NewGaugeMD(replicationAverageActiveWorkers,
"Average number of active replication workers")
replicationAverageQueuedBytesMD = NewGaugeMD(replicationAverageQueuedBytes,
"Average number of bytes queued for replication since server start")
replicationAverageQueuedCountMD = NewGaugeMD(replicationAverageQueuedCount,
"Average number of objects queued for replication since server start")
replicationAverageDataTransferRateMD = NewGaugeMD(replicationAverageDataTransferRate,
"Average replication data transfer rate in bytes/sec")
replicationCurrentActiveWorkersMD = NewGaugeMD(replicationCurrentActiveWorkers,
"Total number of active replication workers")
replicationCurrentDataTransferRateMD = NewGaugeMD(replicationCurrentDataTransferRate,
"Current replication data transfer rate in bytes/sec")
replicationLastMinuteQueuedBytesMD = NewGaugeMD(replicationLastMinuteQueuedBytes,
"Number of bytes queued for replication in the last full minute")
replicationLastMinuteQueuedCountMD = NewGaugeMD(replicationLastMinuteQueuedCount,
"Number of objects queued for replication in the last full minute")
replicationMaxActiveWorkersMD = NewGaugeMD(replicationMaxActiveWorkers,
"Maximum number of active replication workers seen since server start")
replicationMaxQueuedBytesMD = NewGaugeMD(replicationMaxQueuedBytes,
"Maximum number of bytes queued for replication since server start")
replicationMaxQueuedCountMD = NewGaugeMD(replicationMaxQueuedCount,
"Maximum number of objects queued for replication since server start")
replicationMaxDataTransferRateMD = NewGaugeMD(replicationMaxDataTransferRate,
"Maximum replication data transfer rate in bytes/sec seen since server start")
)
// loadClusterReplicationMetrics - `MetricsLoaderFn` for cluster replication metrics
// such as transfer rate and objects queued.
func loadClusterReplicationMetrics(ctx context.Context, m MetricValues, c *metricsCache) error {
if globalReplicationStats == nil {
return nil
}
qs := globalReplicationStats.getNodeQueueStatsSummary()
qt := qs.QStats
m.Set(replicationAverageQueuedBytes, float64(qt.Avg.Bytes))
m.Set(replicationAverageQueuedCount, float64(qt.Avg.Count))
m.Set(replicationMaxQueuedBytes, float64(qt.Max.Bytes))
m.Set(replicationMaxQueuedCount, float64(qt.Max.Count))
m.Set(replicationLastMinuteQueuedBytes, float64(qt.Curr.Bytes))
m.Set(replicationLastMinuteQueuedCount, float64(qt.Curr.Count))
qa := qs.ActiveWorkers
m.Set(replicationAverageActiveWorkers, float64(qa.Avg))
m.Set(replicationCurrentActiveWorkers, float64(qa.Curr))
m.Set(replicationMaxActiveWorkers, float64(qa.Max))
if len(qs.XferStats) > 0 {
tots := qs.XferStats[Total]
m.Set(replicationAverageDataTransferRate, tots.Avg)
m.Set(replicationCurrentDataTransferRate, tots.Curr)
m.Set(replicationMaxDataTransferRate, tots.Peak)
}
return nil
}

12
cmd/metrics-v3-types.go
View File

@ -72,6 +72,8 @@ const (
GaugeMT GaugeMT
// HistogramMT - represents a histogram metric. // HistogramMT - represents a histogram metric.
HistogramMT HistogramMT
// rangeL - represents a range label.
rangeL = "range"
) )
func (mt MetricType) String() string { func (mt MetricType) String() string {
@ -225,7 +227,7 @@ func (m *MetricValues) Set(name MetricName, value float64, labels ...string) {
} }
if len(labels)/2 != len(validLabels) { if len(labels)/2 != len(validLabels) {
panic(fmt.Sprintf("not all labels were given values")) panic("not all labels were given values")
} }
v, ok := m.values[name] v, ok := m.values[name]
@ -284,6 +286,14 @@ func (m *MetricValues) SetHistogram(name MetricName, hist *prometheus.HistogramV
} }
} }
// SetHistogramValues - sets values for the given MetricName using the provided map of
// range to value.
func SetHistogramValues[V uint64 | int64 | float64](m MetricValues, name MetricName, values map[string]V, labels ...string) {
for rng, val := range values {
m.Set(name, float64(val), append(labels, rangeL, rng)...)
}
}
// MetricsLoaderFn - represents a function to load metrics from the // MetricsLoaderFn - represents a function to load metrics from the
// metricsCache. // metricsCache.
// //

72
cmd/metrics-v3.go
View File

@ -35,7 +35,9 @@ import (
// for the bucket "mybucket" would be /minio/metrics/v3/bucket/api/mybucket // for the bucket "mybucket" would be /minio/metrics/v3/bucket/api/mybucket
const ( const (
apiRequestsCollectorPath collectorPath = "/api/requests" apiRequestsCollectorPath collectorPath = "/api/requests"
apiBucketCollectorPath collectorPath = "/bucket/api"
bucketAPICollectorPath collectorPath = "/bucket/api"
bucketReplicationCollectorPath collectorPath = "/bucket/replication"
systemNetworkInternodeCollectorPath collectorPath = "/system/network/internode" systemNetworkInternodeCollectorPath collectorPath = "/system/network/internode"
systemDriveCollectorPath collectorPath = "/system/drive" systemDriveCollectorPath collectorPath = "/system/drive"
@ -54,6 +56,7 @@ const (
auditCollectorPath collectorPath = "/audit" auditCollectorPath collectorPath = "/audit"
loggerWebhookCollectorPath collectorPath = "/logger/webhook" loggerWebhookCollectorPath collectorPath = "/logger/webhook"
replicationCollectorPath collectorPath = "/replication"
) )
const ( const (
@ -97,20 +100,45 @@ func newMetricGroups(r *prometheus.Registry) *metricsV3Collection {
loadAPIRequestsNetworkMetrics), loadAPIRequestsNetworkMetrics),
) )
apiBucketMG := NewBucketMetricsGroup(apiBucketCollectorPath, bucketAPIMG := NewBucketMetricsGroup(bucketAPICollectorPath,
[]MetricDescriptor{ []MetricDescriptor{
apiBucketTrafficRecvBytesMD, bucketAPITrafficRecvBytesMD,
apiBucketTrafficSentBytesMD, bucketAPITrafficSentBytesMD,
apiBucketRequestsInFlightMD, bucketAPIRequestsInFlightMD,
apiBucketRequestsTotalMD, bucketAPIRequestsTotalMD,
apiBucketRequestsCanceledMD, bucketAPIRequestsCanceledMD,
apiBucketRequests4xxErrorsMD, bucketAPIRequests4xxErrorsMD,
apiBucketRequests5xxErrorsMD, bucketAPIRequests5xxErrorsMD,
apiBucketRequestsTTFBSecondsDistributionMD, bucketAPIRequestsTTFBSecondsDistributionMD,
}, },
JoinBucketLoaders(loadAPIBucketHTTPMetrics, loadAPIBucketTTFBMetrics), JoinBucketLoaders(loadBucketAPIHTTPMetrics, loadBucketAPITTFBMetrics),
)
bucketReplicationMG := NewBucketMetricsGroup(bucketReplicationCollectorPath,
[]MetricDescriptor{
bucketReplLastHrFailedBytesMD,
bucketReplLastHrFailedCountMD,
bucketReplLastMinFailedBytesMD,
bucketReplLastMinFailedCountMD,
bucketReplLatencyMsMD,
bucketReplProxiedDeleteTaggingRequestsTotalMD,
bucketReplProxiedGetRequestsFailuresMD,
bucketReplProxiedGetRequestsTotalMD,
bucketReplProxiedGetTaggingRequestsFailuresMD,
bucketReplProxiedGetTaggingRequestsTotalMD,
bucketReplProxiedHeadRequestsFailuresMD,
bucketReplProxiedHeadRequestsTotalMD,
bucketReplProxiedPutTaggingRequestsFailuresMD,
bucketReplProxiedPutTaggingRequestsTotalMD,
bucketReplSentBytesMD,
bucketReplSentCountMD,
bucketReplTotalFailedBytesMD,
bucketReplTotalFailedCountMD,
bucketReplProxiedDeleteTaggingRequestsFailuresMD,
},
loadBucketReplicationMetrics,
) )
systemNetworkInternodeMG := NewMetricsGroup(systemNetworkInternodeCollectorPath, systemNetworkInternodeMG := NewMetricsGroup(systemNetworkInternodeCollectorPath,
@ -296,6 +324,24 @@ func newMetricGroups(r *prometheus.Registry) *metricsV3Collection {
loadClusterIAMMetrics, loadClusterIAMMetrics,
) )
clusterReplicationMG := NewMetricsGroup(replicationCollectorPath,
[]MetricDescriptor{
replicationAverageActiveWorkersMD,
replicationAverageQueuedBytesMD,
replicationAverageQueuedCountMD,
replicationAverageDataTransferRateMD,
replicationCurrentActiveWorkersMD,
replicationCurrentDataTransferRateMD,
replicationLastMinuteQueuedBytesMD,
replicationLastMinuteQueuedCountMD,
replicationMaxActiveWorkersMD,
replicationMaxQueuedBytesMD,
replicationMaxQueuedCountMD,
replicationMaxDataTransferRateMD,
},
loadClusterReplicationMetrics,
)
loggerWebhookMG := NewMetricsGroup(loggerWebhookCollectorPath, loggerWebhookMG := NewMetricsGroup(loggerWebhookCollectorPath,
[]MetricDescriptor{ []MetricDescriptor{
webhookFailedMessagesMD, webhookFailedMessagesMD,
@ -316,7 +362,8 @@ func newMetricGroups(r *prometheus.Registry) *metricsV3Collection {
allMetricGroups := []*MetricsGroup{ allMetricGroups := []*MetricsGroup{
apiRequestsMG, apiRequestsMG,
apiBucketMG, bucketAPIMG,
bucketReplicationMG,
systemNetworkInternodeMG, systemNetworkInternodeMG,
systemDriveMG, systemDriveMG,
@ -330,6 +377,7 @@ func newMetricGroups(r *prometheus.Registry) *metricsV3Collection {
clusterErasureSetMG, clusterErasureSetMG,
clusterNotificationMG, clusterNotificationMG,
clusterIAMMG, clusterIAMMG,
clusterReplicationMG,
auditMG, auditMG,
loggerWebhookMG, loggerWebhookMG,

81
docs/metrics/v3.md
View File

@ -31,7 +31,7 @@ These are metrics about requests served by the (current) node.
| Path | Description | | Path | Description |
|-----------------|--------------------------------------------------| |-----------------|--------------------------------------------------|
| `/api/requests` | Metrics over all requests | | `/api/requests` | Metrics over all requests |
| `/api/bucket` | Metrics over all requests split by bucket labels | | `/bucket/api` | Metrics over all requests for a given bucket |
| | | | | |
### Audit metrics ### Audit metrics
@ -122,6 +122,30 @@ The standard metrics group for GoCollector is not shown below.
| `minio_bucket_api_5xx_errors_total` | `counter` | Total number of requests with 5xx errors for a bucket | `bucket,name,type,server,pool_index` | | `minio_bucket_api_5xx_errors_total` | `counter` | Total number of requests with 5xx errors for a bucket | `bucket,name,type,server,pool_index` |
| `minio_bucket_api_ttfb_seconds_distribution` | `counter` | Distribution of time to first byte across API calls for a bucket | `bucket,name,le,type,server,pool_index` | | `minio_bucket_api_ttfb_seconds_distribution` | `counter` | Distribution of time to first byte across API calls for a bucket | `bucket,name,le,type,server,pool_index` |
### `/bucket/replication`
| Name | Type | Help | Labels |
|---------------------------------------------------------------------|-----------|---------------------------------------------------------------------------------------------|-------------------------------------------|
| `minio_bucket_replication_last_hour_failed_bytes` | `gauge` | Total number of bytes failed at least once to replicate in the last hour on a bucket | `bucket,server` |
| `minio_bucket_replication_last_hour_failed_count` | `gauge` | Total number of objects which failed replication in the last hour on a bucket | `bucket,server` |
| `minio_bucket_replication_last_minute_failed_bytes` | `gauge` | Total number of bytes failed at least once to replicate in the last full minute on a bucket | `bucket,server` |
| `minio_bucket_replication_last_minute_failed_count` | `gauge` | Total number of objects which failed replication in the last full minute on a bucket | `bucket,server` |
| `minio_bucket_replication_latency_ms` | `gauge` | Replication latency on a bucket in milliseconds | `bucket,operation,range,targetArn,server` |
| `minio_bucket_replication_proxied_delete_tagging_requests_total` | `counter` | Number of DELETE tagging requests proxied to replication target | `bucket,server` |
| `minio_bucket_replication_proxied_get_requests_failures` | `counter` | Number of failures in GET requests proxied to replication target | `bucket,server` |
| `minio_bucket_replication_proxied_get_requests_total` | `counter` | Number of GET requests proxied to replication target | `bucket,server` |
| `minio_bucket_replication_proxied_get_tagging_requests_failures` | `counter` | Number of failures in GET tagging requests proxied to replication target | `bucket,server` |
| `minio_bucket_replication_proxied_get_tagging_requests_total` | `counter` | Number of GET tagging requests proxied to replication target | `bucket,server` |
| `minio_bucket_replication_proxied_head_requests_failures` | `counter` | Number of failures in HEAD requests proxied to replication target | `bucket,server` |
| `minio_bucket_replication_proxied_head_requests_total` | `counter` | Number of HEAD requests proxied to replication target | `bucket,server` |
| `minio_bucket_replication_proxied_put_tagging_requests_failures` | `counter` | Number of failures in PUT tagging requests proxied to replication target | `bucket,server` |
| `minio_bucket_replication_proxied_put_tagging_requests_total` | `counter` | Number of PUT tagging requests proxied to replication target | `bucket,server` |
| `minio_bucket_replication_sent_bytes` | `counter` | Total number of bytes replicated to the target | `bucket,server` |
| `minio_bucket_replication_sent_count` | `counter` | Total number of objects replicated to the target | `bucket,server` |
| `minio_bucket_replication_total_failed_bytes` | `counter` | Total number of bytes failed at least once to replicate since server start | `bucket,server` |
| `minio_bucket_replication_total_failed_count` | `counter` | Total number of objects which failed replication since server start | `bucket,server` |
| `minio_bucket_replication_proxied_delete_tagging_requests_failures` | `counter` | Number of failures in DELETE tagging requests proxied to replication target | `bucket,server` |
### `/audit` ### `/audit`
| Name | Type | Help | Labels | | Name | Type | Help | Labels |
@ -195,25 +219,25 @@ The standard metrics group for GoCollector is not shown below.
### `/system/process` ### `/system/process`
| Name | Type | Help | Labels | | Name | Type | Help | Labels |
|-------------------------------|-----------|----------------------------------------------------------------------------------------------------------------|----------| |----------------------------------------------------|-----------|----------------------------------------------------------------------------------------------------------------|----------|
| `locks_read_total` | `gauge` | Number of current READ locks on this peer | `server` | | `minio_system_process_locks_read_total` | `gauge` | Number of current READ locks on this peer | `server` |
| `locks_write_total` | `gauge` | Number of current WRITE locks on this peer | `server` | | `minio_system_process_locks_write_total` | `gauge` | Number of current WRITE locks on this peer | `server` |
| `cpu_total_seconds` | `counter` | Total user and system CPU time spent in seconds | `server` | | `minio_system_process_cpu_total_seconds` | `counter` | Total user and system CPU time spent in seconds | `server` |
| `go_routine_total` | `gauge` | Total number of go routines running | `server` | | `minio_system_process_go_routine_total` | `gauge` | Total number of go routines running | `server` |
| `io_rchar_bytes` | `counter` | Total bytes read by the process from the underlying storage system including cache, /proc/[pid]/io rchar | `server` | | `minio_system_process_io_rchar_bytes` | `counter` | Total bytes read by the process from the underlying storage system including cache, /proc/[pid]/io rchar | `server` |
| `io_read_bytes` | `counter` | Total bytes read by the process from the underlying storage system, /proc/[pid]/io read_bytes | `server` | | `minio_system_process_io_read_bytes` | `counter` | Total bytes read by the process from the underlying storage system, /proc/[pid]/io read_bytes | `server` |
| `io_wchar_bytes` | `counter` | Total bytes written by the process to the underlying storage system including page cache, /proc/[pid]/io wchar | `server` | | `minio_system_process_io_wchar_bytes` | `counter` | Total bytes written by the process to the underlying storage system including page cache, /proc/[pid]/io wchar | `server` |
| `io_write_bytes` | `counter` | Total bytes written by the process to the underlying storage system, /proc/[pid]/io write_bytes | `server` | | `minio_system_process_io_write_bytes` | `counter` | Total bytes written by the process to the underlying storage system, /proc/[pid]/io write_bytes | `server` |
| `start_time_seconds` | `gauge` | Start time for MinIO process in seconds since Unix epoc | `server` | | `minio_system_process_start_time_seconds` | `gauge` | Start time for MinIO process in seconds since Unix epoc | `server` |
| `uptime_seconds` | `gauge` | Uptime for MinIO process in seconds | `server` | | `minio_system_process_uptime_seconds` | `gauge` | Uptime for MinIO process in seconds | `server` |
| `file_descriptor_limit_total` | `gauge` | Limit on total number of open file descriptors for the MinIO Server process | `server` | | `minio_system_process_file_descriptor_limit_total` | `gauge` | Limit on total number of open file descriptors for the MinIO Server process | `server` |
| `file_descriptor_open_total` | `gauge` | Total number of open file descriptors by the MinIO Server process | `server` | | `minio_system_process_file_descriptor_open_total` | `gauge` | Total number of open file descriptors by the MinIO Server process | `server` |
| `syscall_read_total` | `counter` | Total read SysCalls to the kernel. /proc/[pid]/io syscr | `server` | | `minio_system_process_syscall_read_total` | `counter` | Total read SysCalls to the kernel. /proc/[pid]/io syscr | `server` |
| `syscall_write_total` | `counter` | Total write SysCalls to the kernel. /proc/[pid]/io syscw | `server` | | `minio_system_process_syscall_write_total` | `counter` | Total write SysCalls to the kernel. /proc/[pid]/io syscw | `server` |
| `resident_memory_bytes` | `gauge` | Resident memory size in bytes | `server` | | `minio_system_process_resident_memory_bytes` | `gauge` | Resident memory size in bytes | `server` |
| `virtual_memory_bytes` | `gauge` | Virtual memory size in bytes | `server` | | `minio_system_process_virtual_memory_bytes` | `gauge` | Virtual memory size in bytes | `server` |
| `virtual_memory_max_bytes` | `gauge` | Maximum virtual memory size in bytes | `server` | | `minio_system_process_virtual_memory_max_bytes` | `gauge` | Maximum virtual memory size in bytes | `server` |
### `/cluster/health` ### `/cluster/health`
@ -302,3 +326,20 @@ The standard metrics group for GoCollector is not shown below.
| `minio_logger_webhook_failed_messages` | `counter` | Number of messages that failed to send | `server,name,endpoint` | | `minio_logger_webhook_failed_messages` | `counter` | Number of messages that failed to send | `server,name,endpoint` |
| `minio_logger_webhook_queue_length` | `gauge` | Webhook queue length | `server,name,endpoint` | | `minio_logger_webhook_queue_length` | `gauge` | Webhook queue length | `server,name,endpoint` |
| `minio_logger_webhook_total_message` | `counter` | Total number of messages sent to this target | `server,name,endpoint` | | `minio_logger_webhook_total_message` | `counter` | Total number of messages sent to this target | `server,name,endpoint` |
### `/replication`
| Name | Type | Help | Labels |
|---------------------------------------------------|---------|-----------------------------------------------------------------------------|----------|
| `minio_replication_average_active_workers` | `gauge` | Average number of active replication workers | `server` |
| `minio_replication_average_queued_bytes` | `gauge` | Average number of bytes queued for replication since server start | `server` |
| `minio_replication_average_queued_count` | `gauge` | Average number of objects queued for replication since server start | `server` |
| `minio_replication_average_data_transfer_rate` | `gauge` | Average replication data transfer rate in bytes/sec | `server` |
| `minio_replication_current_active_workers` | `gauge` | Total number of active replication workers | `server` |
| `minio_replication_current_data_transfer_rate` | `gauge` | Current replication data transfer rate in bytes/sec | `server` |
| `minio_replication_last_minute_queued_bytes` | `gauge` | Number of bytes queued for replication in the last full minute | `server` |
| `minio_replication_last_minute_queued_count` | `gauge` | Number of objects queued for replication in the last full minute | `server` |
| `minio_replication_max_active_workers` | `gauge` | Maximum number of active replication workers seen since server start | `server` |
| `minio_replication_max_queued_bytes` | `gauge` | Maximum number of bytes queued for replication since server start | `server` |
| `minio_replication_max_queued_count` | `gauge` | Maximum number of objects queued for replication since server start | `server` |
| `minio_replication_max_data_transfer_rate` | `gauge` | Maximum replication data transfer rate in bytes/sec seen since server start | `server` |