Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commitc8580a4

Browse files
feat: expose current agent connections by type via prometheus (#14612)
1 parent40688e4 commitc8580a4

File tree

5 files changed

+82
-43
lines changed

5 files changed

+82
-43
lines changed

‎agent/agent.go

Lines changed: 11 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1510,6 +1510,8 @@ func (a *agent) Collect(ctx context.Context, networkStats map[netlogtype.Connect
15101510
varmu sync.Mutex
15111511
status:=a.network.Status()
15121512
durations:= []float64{}
1513+
p2pConns:=0
1514+
derpConns:=0
15131515
pingCtx,cancelFunc:=context.WithTimeout(ctx,5*time.Second)
15141516
defercancelFunc()
15151517
fornodeID,peer:=rangestatus.Peer {
@@ -1526,13 +1528,18 @@ func (a *agent) Collect(ctx context.Context, networkStats map[netlogtype.Connect
15261528
wg.Add(1)
15271529
gofunc() {
15281530
deferwg.Done()
1529-
duration,_,_,err:=a.network.Ping(pingCtx,addresses[0].Addr())
1531+
duration,p2p,_,err:=a.network.Ping(pingCtx,addresses[0].Addr())
15301532
iferr!=nil {
15311533
return
15321534
}
15331535
mu.Lock()
15341536
defermu.Unlock()
15351537
durations=append(durations,float64(duration.Microseconds()))
1538+
ifp2p {
1539+
p2pConns++
1540+
}else {
1541+
derpConns++
1542+
}
15361543
}()
15371544
}
15381545
wg.Wait()
@@ -1552,6 +1559,9 @@ func (a *agent) Collect(ctx context.Context, networkStats map[netlogtype.Connect
15521559
// Agent metrics are changing all the time, so there is no need to perform
15531560
// reflect.DeepEqual to see if stats should be transferred.
15541561

1562+
// currentConnections behaves like a hypothetical `GaugeFuncVec` and is only set at collection time.
1563+
a.metrics.currentConnections.WithLabelValues("p2p").Set(float64(p2pConns))
1564+
a.metrics.currentConnections.WithLabelValues("derp").Set(float64(derpConns))
15551565
metricsCtx,cancelFunc:=context.WithTimeout(ctx,5*time.Second)
15561566
defercancelFunc()
15571567
a.logger.Debug(ctx,"collecting agent metrics for stats")

‎agent/agent_test.go

Lines changed: 52 additions & 42 deletions
Original file line numberDiff line numberDiff line change
@@ -2531,17 +2531,17 @@ func TestAgent_Metrics_SSH(t *testing.T) {
25312531
err=session.Shell()
25322532
require.NoError(t,err)
25332533

2534-
expected:= []agentsdk.AgentMetric{
2534+
expected:= []*proto.Stats_Metric{
25352535
{
25362536
Name:"agent_reconnecting_pty_connections_total",
2537-
Type:agentsdk.AgentMetricTypeCounter,
2537+
Type:proto.Stats_Metric_COUNTER,
25382538
Value:0,
25392539
},
25402540
{
25412541
Name:"agent_sessions_total",
2542-
Type:agentsdk.AgentMetricTypeCounter,
2542+
Type:proto.Stats_Metric_COUNTER,
25432543
Value:1,
2544-
Labels: []agentsdk.AgentMetricLabel{
2544+
Labels: []*proto.Stats_Metric_Label{
25452545
{
25462546
Name:"magic_type",
25472547
Value:"ssh",
@@ -2554,30 +2554,46 @@ func TestAgent_Metrics_SSH(t *testing.T) {
25542554
},
25552555
{
25562556
Name:"agent_ssh_server_failed_connections_total",
2557-
Type:agentsdk.AgentMetricTypeCounter,
2557+
Type:proto.Stats_Metric_COUNTER,
25582558
Value:0,
25592559
},
25602560
{
25612561
Name:"agent_ssh_server_sftp_connections_total",
2562-
Type:agentsdk.AgentMetricTypeCounter,
2562+
Type:proto.Stats_Metric_COUNTER,
25632563
Value:0,
25642564
},
25652565
{
25662566
Name:"agent_ssh_server_sftp_server_errors_total",
2567-
Type:agentsdk.AgentMetricTypeCounter,
2567+
Type:proto.Stats_Metric_COUNTER,
25682568
Value:0,
25692569
},
25702570
{
2571-
Name:"coderd_agentstats_startup_script_seconds",
2572-
Type:agentsdk.AgentMetricTypeGauge,
2571+
Name:"coderd_agentstats_currently_reachable_peers",
2572+
Type:proto.Stats_Metric_GAUGE,
25732573
Value:0,
2574-
Labels: []agentsdk.AgentMetricLabel{
2574+
Labels: []*proto.Stats_Metric_Label{
2575+
{
2576+
Name:"connection_type",
2577+
Value:"derp",
2578+
},
2579+
},
2580+
},
2581+
{
2582+
Name:"coderd_agentstats_currently_reachable_peers",
2583+
Type:proto.Stats_Metric_GAUGE,
2584+
Value:1,
2585+
Labels: []*proto.Stats_Metric_Label{
25752586
{
2576-
Name:"success",
2577-
Value:"true",
2587+
Name:"connection_type",
2588+
Value:"p2p",
25782589
},
25792590
},
25802591
},
2592+
{
2593+
Name:"coderd_agentstats_startup_script_seconds",
2594+
Type:proto.Stats_Metric_GAUGE,
2595+
Value:1,
2596+
},
25812597
}
25822598

25832599
varactual []*promgo.MetricFamily
@@ -2586,17 +2602,33 @@ func TestAgent_Metrics_SSH(t *testing.T) {
25862602
iferr!=nil {
25872603
returnfalse
25882604
}
2589-
2590-
iflen(expected)!=len(actual) {
2591-
returnfalse
2605+
count:=0
2606+
for_,m:=rangeactual {
2607+
count+=len(m.GetMetric())
25922608
}
2593-
2594-
returnverifyCollectedMetrics(t,expected,actual)
2609+
returncount==len(expected)
25952610
},testutil.WaitLong,testutil.IntervalFast)
25962611

2597-
require.Len(t,actual,len(expected))
2598-
collected:=verifyCollectedMetrics(t,expected,actual)
2599-
require.True(t,collected,"expected metrics were not collected")
2612+
i:=0
2613+
for_,mf:=rangeactual {
2614+
for_,m:=rangemf.GetMetric() {
2615+
assert.Equal(t,expected[i].Name,mf.GetName())
2616+
assert.Equal(t,expected[i].Type.String(),mf.GetType().String())
2617+
// Value is max expected
2618+
ifexpected[i].Type==proto.Stats_Metric_GAUGE {
2619+
assert.GreaterOrEqualf(t,expected[i].Value,m.GetGauge().GetValue(),"expected %s to be greater than or equal to %f, got %f",expected[i].Name,expected[i].Value,m.GetGauge().GetValue())
2620+
}elseifexpected[i].Type==proto.Stats_Metric_COUNTER {
2621+
assert.GreaterOrEqualf(t,expected[i].Value,m.GetCounter().GetValue(),"expected %s to be greater than or equal to %f, got %f",expected[i].Name,expected[i].Value,m.GetCounter().GetValue())
2622+
}
2623+
forj,lbl:=rangeexpected[i].Labels {
2624+
assert.Equal(t,m.GetLabel()[j],&promgo.LabelPair{
2625+
Name:&lbl.Name,
2626+
Value:&lbl.Value,
2627+
})
2628+
}
2629+
i++
2630+
}
2631+
}
26002632

26012633
_=stdin.Close()
26022634
err=session.Wait()
@@ -2828,28 +2860,6 @@ func TestAgent_ManageProcessPriority(t *testing.T) {
28282860
})
28292861
}
28302862

2831-
funcverifyCollectedMetrics(t*testing.T,expected []agentsdk.AgentMetric,actual []*promgo.MetricFamily)bool {
2832-
t.Helper()
2833-
2834-
fori,e:=rangeexpected {
2835-
assert.Equal(t,e.Name,actual[i].GetName())
2836-
assert.Equal(t,string(e.Type),strings.ToLower(actual[i].GetType().String()))
2837-
2838-
for_,m:=rangeactual[i].GetMetric() {
2839-
assert.Equal(t,e.Value,m.Counter.GetValue())
2840-
2841-
iflen(m.GetLabel())>0 {
2842-
forj,lbl:=rangem.GetLabel() {
2843-
assert.Equal(t,e.Labels[j].Name,lbl.GetName())
2844-
assert.Equal(t,e.Labels[j].Value,lbl.GetValue())
2845-
}
2846-
}
2847-
m.GetLabel()
2848-
}
2849-
}
2850-
returntrue
2851-
}
2852-
28532863
typesyncWriterstruct {
28542864
mu sync.Mutex
28552865
w io.Writer

‎agent/metrics.go

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -19,6 +19,7 @@ type agentMetrics struct {
1919
// startupScriptSeconds is the time in seconds that the start script(s)
2020
// took to run. This is reported once per agent.
2121
startupScriptSeconds*prometheus.GaugeVec
22+
currentConnections*prometheus.GaugeVec
2223
}
2324

2425
funcnewAgentMetrics(registerer prometheus.Registerer)*agentMetrics {
@@ -45,10 +46,19 @@ func newAgentMetrics(registerer prometheus.Registerer) *agentMetrics {
4546
}, []string{"success"})
4647
registerer.MustRegister(startupScriptSeconds)
4748

49+
currentConnections:=prometheus.NewGaugeVec(prometheus.GaugeOpts{
50+
Namespace:"coderd",
51+
Subsystem:"agentstats",
52+
Name:"currently_reachable_peers",
53+
Help:"The number of peers (e.g. clients) that are currently reachable over the encrypted network.",
54+
}, []string{"connection_type"})
55+
registerer.MustRegister(currentConnections)
56+
4857
return&agentMetrics{
4958
connectionsTotal:connectionsTotal,
5059
reconnectingPTYErrors:reconnectingPTYErrors,
5160
startupScriptSeconds:startupScriptSeconds,
61+
currentConnections:currentConnections,
5262
}
5363
}
5464

‎docs/admin/prometheus.md

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -96,6 +96,11 @@ spec:
9696

9797
## Available metrics
9898

99+
`coderd_agentstats_*`metrics must first be enabled with the flag
100+
`--prometheus-collect-agent-stats`, or the environment variable
101+
`CODER_PROMETHEUS_COLLECT_AGENT_STATS`before they can be retrieved from the
102+
deployment. They will always be available from the agent.
103+
99104
<!-- Code generated by 'make docs/admin/prometheus.md'. DO NOT EDIT -->
100105

101106
| Name | Type | Description | Labels |
@@ -107,6 +112,7 @@ spec:
107112
| `coderd_agents_up` | gauge | The number of active agents per workspace. | `template_name` `username` `workspace_name` |
108113
| `coderd_agentstats_connection_count` | gauge | The number of established connections by agent | `agent_name` `username` `workspace_name` |
109114
| `coderd_agentstats_connection_median_latency_seconds` | gauge | The median agent connection latency | `agent_name` `username` `workspace_name` |
115+
| `coderd_agentstats_currently_reachable_peers` | gauge | The number of peers (e.g. clients) that are currently reachable over the encrypted network. | `agent_name` `connection_type` `template_name` `username` `workspace_name` |
110116
| `coderd_agentstats_rx_bytes` | gauge | Agent Rx bytes | `agent_name` `username` `workspace_name` |
111117
| `coderd_agentstats_session_count_jetbrains` | gauge | The number of session established by JetBrains | `agent_name` `username` `workspace_name` |
112118
| `coderd_agentstats_session_count_reconnecting_pty` | gauge | The number of session established by reconnecting PTY | `agent_name` `username` `workspace_name` |

‎scripts/metricsdocgen/metrics

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -63,6 +63,9 @@ coderd_agentstats_connection_count{agent_name="main",username="admin",workspace_
6363
# HELP coderd_agentstats_connection_median_latency_seconds The median agent connection latency
6464
# TYPE coderd_agentstats_connection_median_latency_seconds gauge
6565
coderd_agentstats_connection_median_latency_seconds{agent_name="main",username="admin",workspace_name="workspace1"} 0.001784
66+
# HELP coderd_agentstats_currently_reachable_peers The number of peers (e.g. clients) that are currently reachable over the encrypted network.
67+
# TYPE coderd_agentstats_currently_reachable_peers gauge
68+
coderd_agentstats_currently_reachable_peers{agent_name="main",connection_type="derp",template_name="docker",username="admin",workspace_name="workspace1"} 0
6669
# HELP coderd_agentstats_rx_bytes Agent Rx bytes
6770
# TYPE coderd_agentstats_rx_bytes gauge
6871
coderd_agentstats_rx_bytes{agent_name="main",username="admin",workspace_name="workspace1"} 7731

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp