Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add new agent status metrics; rename opni_monitoring_cluster_info to opni_cluster_info #786

Merged
merged 1 commit into from
Nov 9, 2022
Merged

Add new agent status metrics; rename opni_monitoring_cluster_info to opni_cluster_info #786

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 12 additions & 0 deletions pkg/gateway/sync.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,10 +2,12 @@ package gateway

import (
"context"
"fmt"
"math/rand"
"sync"
"time"

"github.com/prometheus/client_golang/prometheus"
agentv1 "github.com/rancher/opni/pkg/agent"
capabilityv1 "github.com/rancher/opni/pkg/apis/capability/v1"
corev1 "github.com/rancher/opni/pkg/apis/core/v1"
Expand All @@ -17,6 +19,14 @@ import (
"google.golang.org/protobuf/types/known/emptypb"
)

var (
mSyncRequests = prometheus.NewCounterVec(prometheus.CounterOpts{
Namespace: "opni",
Name: "server_sync_requests_total",
Help: "Total number of sync requests sent to agents",
}, []string{"cluster_id", "code", "code_text"})
)

type SyncRequester struct {
capabilityv1.UnsafeNodeManagerServer
mu sync.RWMutex
Expand Down Expand Up @@ -75,6 +85,8 @@ func (f *SyncRequester) RequestSync(ctx context.Context, req *capabilityv1.SyncR
"capabilities", req.GetFilter().GetCapabilityNames(),
).Debug("sending sync request to agent")
_, err := clientSet.SyncNow(ctx, req.GetFilter())
code := status.Code(err)
mSyncRequests.WithLabelValues(req.GetCluster().GetId(), fmt.Sprint(code), code.String()).Inc()
if err != nil {
f.logger.With(
zap.Error(err),
Expand Down
54 changes: 53 additions & 1 deletion pkg/management/collector.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,24 +2,47 @@ package management

import (
"context"
"strings"
"time"

"github.com/prometheus/client_golang/prometheus"
corev1 "github.com/rancher/opni/pkg/apis/core/v1"
managementv1 "github.com/rancher/opni/pkg/apis/management/v1"
"github.com/samber/lo"
)

var (
clusterInfo = prometheus.NewDesc(
"opni_monitoring_cluster_info",
"opni_cluster_info",
"Cluster information",
[]string{"cluster_id", "friendly_name"},
prometheus.Labels{},
)
agentUp = prometheus.NewDesc(
"opni_agent_up",
"Agent connection status",
[]string{"cluster_id"},
prometheus.Labels{},
)
agentReady = prometheus.NewDesc(
"opni_agent_ready",
"Agent readiness status",
[]string{"cluster_id", "conditions"},
prometheus.Labels{},
)
agentSummary = prometheus.NewDesc(
"opni_agent_status_summary",
"Agent status summary",
[]string{"cluster_id", "summary"},
prometheus.Labels{},
)
)

func (s *Server) Describe(c chan<- *prometheus.Desc) {
c <- clusterInfo
c <- agentUp
c <- agentReady
c <- agentSummary
}

func (s *Server) Collect(c chan<- prometheus.Metric) {
Expand All @@ -44,5 +67,34 @@ func (s *Server) Collect(c chan<- prometheus.Metric) {
cluster.Id,
friendlyName,
)

var connected, ready float64
var conditions, summary string
if hs, err := s.GetClusterHealthStatus(ctx, &corev1.Reference{Id: cluster.Id}); err == nil {
connected = lo.Ternary[float64](hs.Status.Connected, 1, 0)
ready = lo.Ternary[float64](hs.Health.Ready, 1, 0)
conditions = strings.Join(hs.Health.Conditions, ",")
summary = hs.Summary()
}
c <- prometheus.MustNewConstMetric(
agentUp,
prometheus.GaugeValue,
connected,
cluster.Id,
)
c <- prometheus.MustNewConstMetric(
agentReady,
prometheus.GaugeValue,
ready,
cluster.Id,
conditions,
)
c <- prometheus.MustNewConstMetric(
agentSummary,
prometheus.GaugeValue,
1,
cluster.Id,
summary,
)
}
}
46 changes: 31 additions & 15 deletions pkg/management/collector_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,18 +30,11 @@ var _ = Describe("Collector", Ordered, Label("slow"), func() {
It("should collect descriptors but no metrics", func() {
descs := make(chan *prometheus.Desc, 100)
tv.ifaces.collector.Describe(descs)
Eventually(descs).Should(Receive(WithTransform(fmt.Stringer.String, Equal(
descriptorString(
"opni_monitoring_cluster_info",
"Cluster information",
[]string{},
[]string{"cluster_id", "friendly_name"},
),
))))
Consistently(descs).ShouldNot(Receive())
Eventually(descs).Should(HaveLen(4))
Consistently(descs).Should(HaveLen(4))
metrics := make(chan prometheus.Metric, 100)
tv.ifaces.collector.Collect(metrics)
Consistently(metrics).ShouldNot(Receive())
Consistently(metrics).Should(BeEmpty())
})
})
When("clusters are present", func() {
Expand All @@ -61,16 +54,39 @@ var _ = Describe("Collector", Ordered, Label("slow"), func() {
},
})

descs := make(chan *prometheus.Desc, 100)
tv.ifaces.collector.Describe(descs)
Expect(descs).To(Receive(WithTransform(fmt.Stringer.String, Equal(
c := make(chan *prometheus.Desc, 100)
tv.ifaces.collector.Describe(c)
Expect(c).To(HaveLen(4))
descs := make([]string, 0, 4)
for i := 0; i < 4; i++ {
descs = append(descs, (<-c).String())
}
Expect(descs).To(ConsistOf(
descriptorString(
"opni_monitoring_cluster_info",
"opni_cluster_info",
"Cluster information",
[]string{},
[]string{"cluster_id", "friendly_name"},
),
))))
descriptorString(
"opni_agent_up",
"Agent connection status",
[]string{},
[]string{"cluster_id"},
),
descriptorString(
"opni_agent_ready",
"Agent readiness status",
[]string{},
[]string{"cluster_id", "conditions"},
),
descriptorString(
"opni_agent_status_summary",
"Agent status summary",
[]string{},
[]string{"cluster_id", "summary"},
),
))

metrics := make(chan prometheus.Metric, 100)
tv.ifaces.collector.Collect(metrics)
Expand Down
34 changes: 17 additions & 17 deletions pkg/resources/monitoring/dashboards/dashboards.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1496,7 +1496,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -3201,7 +3201,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 0,
Expand Down Expand Up @@ -4214,7 +4214,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -6945,7 +6945,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -10595,7 +10595,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -11480,7 +11480,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -13631,7 +13631,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -15369,7 +15369,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -17301,7 +17301,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -19311,7 +19311,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -20534,7 +20534,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 0,
Expand Down Expand Up @@ -22053,7 +22053,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 0,
Expand Down Expand Up @@ -22649,7 +22649,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -23658,7 +23658,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 0,
Expand Down Expand Up @@ -24872,7 +24872,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -25839,7 +25839,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 1,
Expand Down Expand Up @@ -26986,7 +26986,7 @@
"multi": false,
"name": "cluster",
"options": [ ],
"query": "query_result(opni_monitoring_cluster_info)",
"query": "query_result(opni_cluster_info)",
"refresh": 2,
"regex": "/cluster_id=\"(?<value>[^\"]+)|friendly_name=\"(?<text>[^\"]+)/g",
"sort": 0,
Expand Down
2 changes: 1 addition & 1 deletion pkg/resources/monitoring/dashboards/dashboards.jsonnet
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ local patchDashboard(d) = {
list: [
if x.name == 'cluster'
then x {
query: 'query_result(opni_monitoring_cluster_info)',
query: 'query_result(opni_cluster_info)',
regex: '/cluster_id="(?<value>[^"]+)|friendly_name="(?<text>[^"]+)/g',
}
else x
Expand Down
2 changes: 1 addition & 1 deletion pkg/resources/monitoring/dashboards/opni-gateway.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,7 +106,7 @@
"uid": "P0BEE70D62F42845E"
},
"exemplar": false,
"expr": "sum by(cluster_id) (rate(opni_gateway_remote_write_cluster_ingest_bytes[5m])) * on(cluster_id) group_left(friendly_name) opni_monitoring_cluster_info",
"expr": "sum by(cluster_id) (rate(opni_gateway_remote_write_cluster_ingest_bytes[5m])) * on(cluster_id) group_left(friendly_name) opni_cluster_info",
"instant": false,
"interval": "",
"legendFormat": "{{friendly_name}}",
Expand Down
6 changes: 6 additions & 0 deletions pkg/resources/monitoring/grafana.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -178,6 +178,12 @@ func (r *Reconciler) grafana() ([]resources.Resource, error) {
SecurityContext: &corev1.PodSecurityContext{
FSGroup: lo.ToPtr(int64(472)),
},
Env: []corev1.EnvVar{
{
Name: "GF_INSTALL_PLUGINS",
Value: "grafana-polystat-panel,marcusolsson-treemap-panel",
},
},
},
Secrets: []string{"grafana-datasource-cert"},
DataStorage: &grafanav1alpha1.GrafanaDataStorage{
Expand Down
7 changes: 7 additions & 0 deletions plugins/metrics/pkg/cortex/metrics.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,11 +17,18 @@ var (
Name: "remote_write_cluster_ingest_bytes",
Help: "Total number of (compressed) bytes received from remote write requests by cluster ID",
}, []string{"cluster_id"})
mRemoteWriteRequests = prometheus.NewCounterVec(prometheus.CounterOpts{
Namespace: "opni",
Subsystem: "gateway",
Name: "remote_write_requests_total",
Help: "Total number of remote write requests forwarded to Cortex",
}, []string{"cluster_id", "code", "code_text"})
)

func Collectors() []prometheus.Collector {
return []prometheus.Collector{
mIngestBytesTotal,
mIngestBytesByID,
mRemoteWriteRequests,
}
}
Loading