opensource/coder
3
0
Fork 0
mirror of https://github.com/coder/coder.git synced 2026-09-24 15:04:27 +08:00
Code Issues Packages Projects Releases Wiki Activity

feat: add metrics to workspace agent scripts (#11132)

Browse Source 
* push startup script metrics to agent
This commit is contained in:
Steven Masley
2023-12-13 11:45:43 -06:00
committed by GitHub
parent 41ed581460
commit b7bdb17460
20 changed files with 306 additions and 127 deletions
Download Patch File Download Diff File Expand all files Collapse all files
coderd/coderd.go
+3 -1
View File
@@ -38,8 +38,10 @@ import (
_ "github.com/coder/coder/v2/coderd/apidoc"
"github.com/coder/coder/v2/coderd/externalauth"
"github.com/coder/coder/v2/coderd/healthcheck/derphealth"
"github.com/coder/coder/v2/coderd/prometheusmetrics"
"cdr.dev/slog"
"github.com/coder/coder/v2/buildinfo"
"github.com/coder/coder/v2/cli/clibase"
"github.com/coder/coder/v2/coderd/audit"
@@ -168,7 +170,7 @@ type Options struct {
HTTPClient *http.Client
UpdateAgentMetrics func(ctx context.Context, username, workspaceName, agentName string, metrics []agentsdk.AgentMetric)
UpdateAgentMetrics func(ctx context.Context, labels prometheusmetrics.AgentMetricLabels, metrics []agentsdk.AgentMetric)
StatsBatcher *batchstats.Batcher
WorkspaceAppsStatsCollectorOptions workspaceapps.StatsCollectorOptions
coderd/database/dbauthz/dbauthz.go
+1 -1
View File
@@ -1918,7 +1918,7 @@ func (q *querier) GetWorkspaceBuildsCreatedAfter(ctx context.Context, createdAt
return q.db.GetWorkspaceBuildsCreatedAfter(ctx, createdAt)
}
func (q *querier) GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (database.Workspace, error) {
func (q *querier) GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (database.GetWorkspaceByAgentIDRow, error) {
return fetch(q.log, q.auth, q.db.GetWorkspaceByAgentID)(ctx, agentID)
}
coderd/database/dbauthz/dbauthz_test.go
+40 -10
View File
@@ -1065,21 +1065,30 @@ func (s *MethodTestSuite) TestWorkspace() {
check.Args(ws.ID).Asserts(ws, rbac.ActionRead).Returns(b)
}))
s.Run("GetWorkspaceAgentByID", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
tpl := dbgen.Template(s.T(), db, database.Template{})
ws := dbgen.Workspace(s.T(), db, database.Workspace{
TemplateID: tpl.ID,
})
build := dbgen.WorkspaceBuild(s.T(), db, database.WorkspaceBuild{WorkspaceID: ws.ID, JobID: uuid.New()})
res := dbgen.WorkspaceResource(s.T(), db, database.WorkspaceResource{JobID: build.JobID})
agt := dbgen.WorkspaceAgent(s.T(), db, database.WorkspaceAgent{ResourceID: res.ID})
check.Args(agt.ID).Asserts(ws, rbac.ActionRead).Returns(agt)
}))
s.Run("GetWorkspaceAgentByInstanceID", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
tpl := dbgen.Template(s.T(), db, database.Template{})
ws := dbgen.Workspace(s.T(), db, database.Workspace{
TemplateID: tpl.ID,
})
build := dbgen.WorkspaceBuild(s.T(), db, database.WorkspaceBuild{WorkspaceID: ws.ID, JobID: uuid.New()})
res := dbgen.WorkspaceResource(s.T(), db, database.WorkspaceResource{JobID: build.JobID})
agt := dbgen.WorkspaceAgent(s.T(), db, database.WorkspaceAgent{ResourceID: res.ID})
check.Args(agt.AuthInstanceID.String).Asserts(ws, rbac.ActionRead).Returns(agt)
}))
s.Run("UpdateWorkspaceAgentLifecycleStateByID", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
tpl := dbgen.Template(s.T(), db, database.Template{})
ws := dbgen.Workspace(s.T(), db, database.Workspace{
TemplateID: tpl.ID,
})
build := dbgen.WorkspaceBuild(s.T(), db, database.WorkspaceBuild{WorkspaceID: ws.ID, JobID: uuid.New()})
res := dbgen.WorkspaceResource(s.T(), db, database.WorkspaceResource{JobID: build.JobID})
agt := dbgen.WorkspaceAgent(s.T(), db, database.WorkspaceAgent{ResourceID: res.ID})
@@ -1089,7 +1098,10 @@ func (s *MethodTestSuite) TestWorkspace() {
}).Asserts(ws, rbac.ActionUpdate).Returns()
}))
s.Run("UpdateWorkspaceAgentLogOverflowByID", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
tpl := dbgen.Template(s.T(), db, database.Template{})
ws := dbgen.Workspace(s.T(), db, database.Workspace{
TemplateID: tpl.ID,
})
build := dbgen.WorkspaceBuild(s.T(), db, database.WorkspaceBuild{WorkspaceID: ws.ID, JobID: uuid.New()})
res := dbgen.WorkspaceResource(s.T(), db, database.WorkspaceResource{JobID: build.JobID})
agt := dbgen.WorkspaceAgent(s.T(), db, database.WorkspaceAgent{ResourceID: res.ID})
@@ -1099,7 +1111,10 @@ func (s *MethodTestSuite) TestWorkspace() {
}).Asserts(ws, rbac.ActionUpdate).Returns()
}))
s.Run("UpdateWorkspaceAgentStartupByID", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
tpl := dbgen.Template(s.T(), db, database.Template{})
ws := dbgen.Workspace(s.T(), db, database.Workspace{
TemplateID: tpl.ID,
})
build := dbgen.WorkspaceBuild(s.T(), db, database.WorkspaceBuild{WorkspaceID: ws.ID, JobID: uuid.New()})
res := dbgen.WorkspaceResource(s.T(), db, database.WorkspaceResource{JobID: build.JobID})
agt := dbgen.WorkspaceAgent(s.T(), db, database.WorkspaceAgent{ResourceID: res.ID})
@@ -1111,7 +1126,10 @@ func (s *MethodTestSuite) TestWorkspace() {
}).Asserts(ws, rbac.ActionUpdate).Returns()
}))
s.Run("GetWorkspaceAgentLogsAfter", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
tpl := dbgen.Template(s.T(), db, database.Template{})
ws := dbgen.Workspace(s.T(), db, database.Workspace{
TemplateID: tpl.ID,
})
build := dbgen.WorkspaceBuild(s.T(), db, database.WorkspaceBuild{WorkspaceID: ws.ID, JobID: uuid.New()})
res := dbgen.WorkspaceResource(s.T(), db, database.WorkspaceResource{JobID: build.JobID})
agt := dbgen.WorkspaceAgent(s.T(), db, database.WorkspaceAgent{ResourceID: res.ID})
@@ -1120,7 +1138,10 @@ func (s *MethodTestSuite) TestWorkspace() {
}).Asserts(ws, rbac.ActionRead).Returns([]database.WorkspaceAgentLog{})
}))
s.Run("GetWorkspaceAppByAgentIDAndSlug", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
tpl := dbgen.Template(s.T(), db, database.Template{})
ws := dbgen.Workspace(s.T(), db, database.Workspace{
TemplateID: tpl.ID,
})
build := dbgen.WorkspaceBuild(s.T(), db, database.WorkspaceBuild{WorkspaceID: ws.ID, JobID: uuid.New()})
res := dbgen.WorkspaceResource(s.T(), db, database.WorkspaceResource{JobID: build.JobID})
agt := dbgen.WorkspaceAgent(s.T(), db, database.WorkspaceAgent{ResourceID: res.ID})
@@ -1132,7 +1153,10 @@ func (s *MethodTestSuite) TestWorkspace() {
}).Asserts(ws, rbac.ActionRead).Returns(app)
}))
s.Run("GetWorkspaceAppsByAgentID", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
tpl := dbgen.Template(s.T(), db, database.Template{})
ws := dbgen.Workspace(s.T(), db, database.Workspace{
TemplateID: tpl.ID,
})
build := dbgen.WorkspaceBuild(s.T(), db, database.WorkspaceBuild{WorkspaceID: ws.ID, JobID: uuid.New()})
res := dbgen.WorkspaceResource(s.T(), db, database.WorkspaceResource{JobID: build.JobID})
agt := dbgen.WorkspaceAgent(s.T(), db, database.WorkspaceAgent{ResourceID: res.ID})
@@ -1173,11 +1197,17 @@ func (s *MethodTestSuite) TestWorkspace() {
check.Args(database.GetWorkspaceBuildsByWorkspaceIDParams{WorkspaceID: ws.ID}).Asserts(ws, rbac.ActionRead) // ordering
}))
s.Run("GetWorkspaceByAgentID", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
tpl := dbgen.Template(s.T(), db, database.Template{})
ws := dbgen.Workspace(s.T(), db, database.Workspace{
TemplateID: tpl.ID,
})
build := dbgen.WorkspaceBuild(s.T(), db, database.WorkspaceBuild{WorkspaceID: ws.ID, JobID: uuid.New()})
res := dbgen.WorkspaceResource(s.T(), db, database.WorkspaceResource{JobID: build.JobID})
agt := dbgen.WorkspaceAgent(s.T(), db, database.WorkspaceAgent{ResourceID: res.ID})
check.Args(agt.ID).Asserts(ws, rbac.ActionRead).Returns(ws)
check.Args(agt.ID).Asserts(ws, rbac.ActionRead).Returns(database.GetWorkspaceByAgentIDRow{
Workspace: ws,
TemplateName: tpl.Name,
})
}))
s.Run("GetWorkspaceByOwnerIDAndName", s.Subtest(func(db database.Store, check *expects) {
ws := dbgen.Workspace(s.T(), db, database.Workspace{})
coderd/database/dbmem/dbmem.go
+15 -2
View File
@@ -4293,11 +4293,24 @@ func (q *FakeQuerier) GetWorkspaceBuildsCreatedAfter(_ context.Context, after ti
return workspaceBuilds, nil
}
func (q *FakeQuerier) GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (database.Workspace, error) {
func (q *FakeQuerier) GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (database.GetWorkspaceByAgentIDRow, error) {
q.mutex.RLock()
defer q.mutex.RUnlock()
return q.getWorkspaceByAgentIDNoLock(ctx, agentID)
w, err := q.getWorkspaceByAgentIDNoLock(ctx, agentID)
if err != nil {
return database.GetWorkspaceByAgentIDRow{}, err
}
tpl, err := q.getTemplateByIDNoLock(ctx, w.TemplateID)
if err != nil {
return database.GetWorkspaceByAgentIDRow{}, err
}
return database.GetWorkspaceByAgentIDRow{
Workspace: w,
TemplateName: tpl.Name,
}, nil
}
func (q *FakeQuerier) GetWorkspaceByID(ctx context.Context, id uuid.UUID) (database.Workspace, error) {
coderd/database/dbmetrics/dbmetrics.go
+1 -1
View File
@@ -1124,7 +1124,7 @@ func (m metricsStore) GetWorkspaceBuildsCreatedAfter(ctx context.Context, create
return builds, err
}
func (m metricsStore) GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (database.Workspace, error) {
func (m metricsStore) GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (database.GetWorkspaceByAgentIDRow, error) {
start := time.Now()
workspace, err := m.s.GetWorkspaceByAgentID(ctx, agentID)
m.queryLatencies.WithLabelValues("GetWorkspaceByAgentID").Observe(time.Since(start).Seconds())
coderd/database/dbmock/dbmock.go
+2 -2
View File
@@ -2344,10 +2344,10 @@ func (mr *MockStoreMockRecorder) GetWorkspaceBuildsCreatedAfter(arg0, arg1 inter
}
// GetWorkspaceByAgentID mocks base method.
func (m *MockStore) GetWorkspaceByAgentID(arg0 context.Context, arg1 uuid.UUID) (database.Workspace, error) {
func (m *MockStore) GetWorkspaceByAgentID(arg0 context.Context, arg1 uuid.UUID) (database.GetWorkspaceByAgentIDRow, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "GetWorkspaceByAgentID", arg0, arg1)
ret0, _ := ret[0].(database.Workspace)
ret0, _ := ret[0].(database.GetWorkspaceByAgentIDRow)
ret1, _ := ret[1].(error)
return ret0, ret1
}
coderd/database/modelmethods.go
+4
View File
@@ -148,6 +148,10 @@ func (g Group) RBACObject() rbac.Object {
InOrg(g.OrganizationID)
}
func (w GetWorkspaceByAgentIDRow) RBACObject() rbac.Object {
return w.Workspace.RBACObject()
}
func (w Workspace) RBACObject() rbac.Object {
return rbac.ResourceWorkspace.WithID(w.ID).
InOrg(w.OrganizationID).
coderd/database/querier.go
+1 -1
View File
@@ -231,7 +231,7 @@ type sqlcQuerier interface {
GetWorkspaceBuildParameters(ctx context.Context, workspaceBuildID uuid.UUID) ([]WorkspaceBuildParameter, error)
GetWorkspaceBuildsByWorkspaceID(ctx context.Context, arg GetWorkspaceBuildsByWorkspaceIDParams) ([]WorkspaceBuild, error)
GetWorkspaceBuildsCreatedAfter(ctx context.Context, createdAt time.Time) ([]WorkspaceBuild, error)
GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (Workspace, error)
GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (GetWorkspaceByAgentIDRow, error)
GetWorkspaceByID(ctx context.Context, id uuid.UUID) (Workspace, error)
GetWorkspaceByOwnerIDAndName(ctx context.Context, arg GetWorkspaceByOwnerIDAndNameParams) (Workspace, error)
GetWorkspaceByWorkspaceAppID(ctx context.Context, workspaceAppID uuid.UUID) (Workspace, error)
coderd/database/queries.sql.go
+26 -17
View File
@@ -10539,9 +10539,12 @@ func (q *sqlQuerier) GetDeploymentWorkspaceStats(ctx context.Context) (GetDeploy
const getWorkspaceByAgentID = `-- name: GetWorkspaceByAgentID :one
SELECT
id, created_at, updated_at, owner_id, organization_id, template_id, deleted, name, autostart_schedule, ttl, last_used_at, dormant_at, deleting_at, automatic_updates
workspaces.id, workspaces.created_at, workspaces.updated_at, workspaces.owner_id, workspaces.organization_id, workspaces.template_id, workspaces.deleted, workspaces.name, workspaces.autostart_schedule, workspaces.ttl, workspaces.last_used_at, workspaces.dormant_at, workspaces.deleting_at, workspaces.automatic_updates,
templates.name as template_name
FROM
workspaces
INNER JOIN
templates ON workspaces.template_id = templates.id
WHERE
workspaces.id = (
SELECT
@@ -10567,24 +10570,30 @@ WHERE
)
`
func (q *sqlQuerier) GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (Workspace, error) {
type GetWorkspaceByAgentIDRow struct {
Workspace Workspace `db:"workspace" json:"workspace"`
TemplateName string `db:"template_name" json:"template_name"`
}
func (q *sqlQuerier) GetWorkspaceByAgentID(ctx context.Context, agentID uuid.UUID) (GetWorkspaceByAgentIDRow, error) {
row := q.db.QueryRowContext(ctx, getWorkspaceByAgentID, agentID)
var i Workspace
var i GetWorkspaceByAgentIDRow
err := row.Scan(
&i.ID,
&i.CreatedAt,
&i.UpdatedAt,
&i.OwnerID,
&i.OrganizationID,
&i.TemplateID,
&i.Deleted,
&i.Name,
&i.AutostartSchedule,
&i.Ttl,
&i.LastUsedAt,
&i.DormantAt,
&i.DeletingAt,
&i.AutomaticUpdates,
&i.Workspace.ID,
&i.Workspace.CreatedAt,
&i.Workspace.UpdatedAt,
&i.Workspace.OwnerID,
&i.Workspace.OrganizationID,
&i.Workspace.TemplateID,
&i.Workspace.Deleted,
&i.Workspace.Name,
&i.Workspace.AutostartSchedule,
&i.Workspace.Ttl,
&i.Workspace.LastUsedAt,
&i.Workspace.DormantAt,
&i.Workspace.DeletingAt,
&i.Workspace.AutomaticUpdates,
&i.TemplateName,
)
return i, err
}
coderd/database/queries/workspaces.sql
+4 -1
View File
@@ -46,9 +46,12 @@ WHERE
-- name: GetWorkspaceByAgentID :one
SELECT
*
sqlc.embed(workspaces),
templates.name as template_name
FROM
workspaces
INNER JOIN
templates ON workspaces.template_id = templates.id
WHERE
workspaces.id = (
SELECT
coderd/prometheusmetrics/aggregator.go
+19 -6
View File
@@ -47,6 +47,7 @@ type updateRequest struct {
username string
workspaceName string
agentName string
templateName string
metrics []agentsdk.AgentMetric
@@ -59,6 +60,7 @@ type annotatedMetric struct {
username string
workspaceName string
agentName string
templateName string
expiryDate time.Time
}
@@ -74,7 +76,7 @@ func (am *annotatedMetric) asPrometheus() (prometheus.Metric, error) {
labelValues := make([]string, 0, len(agentMetricsLabels)+len(am.Labels))
labels = append(labels, agentMetricsLabels...)
labelValues = append(labelValues, am.username, am.workspaceName, am.agentName)
labelValues = append(labelValues, am.username, am.workspaceName, am.agentName, am.templateName)
for _, l := range am.Labels {
labels = append(labels, l.Name)
@@ -160,6 +162,7 @@ func (ma *MetricsAggregator) Run(ctx context.Context) func() {
username: req.username,
workspaceName: req.workspaceName,
agentName: req.agentName,
templateName: req.templateName,
AgentMetric: m,
@@ -227,7 +230,16 @@ func (ma *MetricsAggregator) Run(ctx context.Context) func() {
func (*MetricsAggregator) Describe(_ chan<- *prometheus.Desc) {
}
var agentMetricsLabels = []string{usernameLabel, workspaceNameLabel, agentNameLabel}
var agentMetricsLabels = []string{usernameLabel, workspaceNameLabel, agentNameLabel, templateNameLabel}
// AgentMetricLabels are the labels used to decorate an agent's metrics.
// This list should match the list of labels in agentMetricsLabels.
type AgentMetricLabels struct {
Username string
WorkspaceName string
AgentName string
TemplateName string
}
func (ma *MetricsAggregator) Collect(ch chan<- prometheus.Metric) {
output := make(chan []prometheus.Metric, 1)
@@ -246,12 +258,13 @@ func (ma *MetricsAggregator) Collect(ch chan<- prometheus.Metric) {
}
}
func (ma *MetricsAggregator) Update(ctx context.Context, username, workspaceName, agentName string, metrics []agentsdk.AgentMetric) {
func (ma *MetricsAggregator) Update(ctx context.Context, labels AgentMetricLabels, metrics []agentsdk.AgentMetric) {
select {
case ma.updateCh <- updateRequest{
username: username,
workspaceName: workspaceName,
agentName: agentName,
username: labels.Username,
workspaceName: labels.WorkspaceName,
agentName: labels.AgentName,
templateName: labels.TemplateName,
metrics: metrics,
timestamp: time.Now(),
coderd/prometheusmetrics/aggregator_test.go
+20 -4
View File
@@ -2,6 +2,7 @@ package prometheusmetrics_test
import (
"context"
"sort"
"sync/atomic"
"testing"
"time"
@@ -12,6 +13,7 @@ import (
"github.com/stretchr/testify/require"
"cdr.dev/slog/sloggers/slogtest"
"github.com/coder/coder/v2/coderd/prometheusmetrics"
"github.com/coder/coder/v2/codersdk/agentsdk"
"github.com/coder/coder/v2/cryptorand"
@@ -22,8 +24,16 @@ const (
testWorkspaceName = "yogi-workspace"
testUsername = "yogi-bear"
testAgentName = "main-agent"
testTemplateName = "main-template"
)
var testLabels = prometheusmetrics.AgentMetricLabels{
Username: testUsername,
WorkspaceName: testWorkspaceName,
AgentName: testAgentName,
TemplateName: testTemplateName,
}
func TestUpdateMetrics_MetricsDoNotExpire(t *testing.T) {
t.Parallel()
@@ -58,6 +68,7 @@ func TestUpdateMetrics_MetricsDoNotExpire(t *testing.T) {
{Name: "agent_name", Value: testAgentName},
{Name: "username", Value: testUsername},
{Name: "workspace_name", Value: testWorkspaceName},
{Name: "template_name", Value: testTemplateName},
}
expected := []agentsdk.AgentMetric{
{Name: "a_counter_one", Type: agentsdk.AgentMetricTypeCounter, Value: 1, Labels: commonLabels},
@@ -69,13 +80,14 @@ func TestUpdateMetrics_MetricsDoNotExpire(t *testing.T) {
{Name: "hello", Value: "world"},
{Name: "username", Value: testUsername},
{Name: "workspace_name", Value: testWorkspaceName},
{Name: "template_name", Value: testTemplateName},
}},
{Name: "d_gauge_four", Type: agentsdk.AgentMetricTypeGauge, Value: 6, Labels: commonLabels},
}
// when
metricsAggregator.Update(ctx, testUsername, testWorkspaceName, testAgentName, given1)
metricsAggregator.Update(ctx, testUsername, testWorkspaceName, testAgentName, given2)
metricsAggregator.Update(ctx, testLabels, given1)
metricsAggregator.Update(ctx, testLabels, given2)
// then
require.Eventually(t, func() bool {
@@ -119,6 +131,10 @@ func verifyCollectedMetrics(t *testing.T, expected []agentsdk.AgentMetric, actua
}
dtoLabels := asMetricAgentLabels(d.GetLabel())
// dto labels are sorted in alphabetical order.
sort.Slice(e.Labels, func(i, j int) bool {
return e.Labels[i].Name < e.Labels[j].Name
})
require.Equal(t, e.Labels, dtoLabels, d.String())
}
return true
@@ -154,7 +170,7 @@ func TestUpdateMetrics_MetricsExpire(t *testing.T) {
}
// when
metricsAggregator.Update(ctx, testUsername, testWorkspaceName, testAgentName, given)
metricsAggregator.Update(ctx, testLabels, given)
time.Sleep(time.Millisecond * 10) // Ensure that metric is expired
@@ -220,7 +236,7 @@ func Benchmark_MetricsAggregator_Run(b *testing.B) {
b.Logf("N=%d sending %d metrics", b.N, numMetrics)
var nGot atomic.Int64
b.StartTimer()
metricsAggregator.Update(ctx, testUsername, testWorkspaceName, testAgentName, metrics)
metricsAggregator.Update(ctx, testLabels, metrics)
for i := 0; i < numMetrics; i++ {
select {
case <-ctx.Done():
coderd/prometheusmetrics/prometheusmetrics.go
+3 -1
View File
@@ -17,6 +17,7 @@ import (
"tailscale.com/tailcfg"
"cdr.dev/slog"
"github.com/coder/coder/v2/coderd/database"
"github.com/coder/coder/v2/coderd/database/dbauthz"
"github.com/coder/coder/v2/coderd/database/dbtime"
@@ -24,6 +25,7 @@ import (
)
const (
templateNameLabel = "template_name"
agentNameLabel = "agent_name"
usernameLabel = "username"
workspaceNameLabel = "workspace_name"
@@ -154,7 +156,7 @@ func Agents(ctx context.Context, logger slog.Logger, registerer prometheus.Regis
Subsystem: "agents",
Name: "up",
Help: "The number of active agents per workspace.",
}, []string{usernameLabel, workspaceNameLabel, "template_name", "template_version"}))
}, []string{usernameLabel, workspaceNameLabel, templateNameLabel, "template_version"}))
err := registerer.Register(agentsGauge)
if err != nil {
return nil, err
coderd/workspaceagents.go
+14 -5
View File
@@ -37,6 +37,7 @@ import (
"github.com/coder/coder/v2/coderd/externalauth"
"github.com/coder/coder/v2/coderd/httpapi"
"github.com/coder/coder/v2/coderd/httpmw"
"github.com/coder/coder/v2/coderd/prometheusmetrics"
"github.com/coder/coder/v2/coderd/rbac"
"github.com/coder/coder/v2/coderd/util/ptr"
"github.com/coder/coder/v2/codersdk"
@@ -572,7 +573,7 @@ func (api *API) workspaceAgentLogs(rw http.ResponseWriter, r *http.Request) {
return
}
workspace, err := api.Database.GetWorkspaceByAgentID(ctx, workspaceAgent.ID)
row, err := api.Database.GetWorkspaceByAgentID(ctx, workspaceAgent.ID)
if err != nil {
httpapi.Write(ctx, rw, http.StatusInternalServerError, codersdk.Response{
Message: "Internal error fetching workspace by agent id.",
@@ -580,6 +581,7 @@ func (api *API) workspaceAgentLogs(rw http.ResponseWriter, r *http.Request) {
})
return
}
workspace := row.Workspace
api.WebsocketWaitMutex.Lock()
api.WebsocketWaitGroup.Add(1)
@@ -1648,7 +1650,7 @@ func (api *API) workspaceAgentReportStats(rw http.ResponseWriter, r *http.Reques
ctx := r.Context()
workspaceAgent := httpmw.WorkspaceAgent(r)
workspace, err := api.Database.GetWorkspaceByAgentID(ctx, workspaceAgent.ID)
row, err := api.Database.GetWorkspaceByAgentID(ctx, workspaceAgent.ID)
if err != nil {
httpapi.Write(ctx, rw, http.StatusBadRequest, codersdk.Response{
Message: "Failed to get workspace.",
@@ -1656,6 +1658,7 @@ func (api *API) workspaceAgentReportStats(rw http.ResponseWriter, r *http.Reques
})
return
}
workspace := row.Workspace
var req agentsdk.Stats
if !httpapi.Read(ctx, rw, r, &req) {
@@ -1681,7 +1684,7 @@ func (api *API) workspaceAgentReportStats(rw http.ResponseWriter, r *http.Reques
var nextAutostart time.Time
if workspace.AutostartSchedule.String != "" {
templateSchedule, err := (*(api.TemplateScheduleStore.Load())).Get(ctx, api.Database, workspace.TemplateID)
// If the template schedule fails to load, just default to bumping without the next trasition and log it.
// If the template schedule fails to load, just default to bumping without the next transition and log it.
if err != nil {
api.Logger.Warn(ctx, "failed to load template schedule bumping activity, defaulting to bumping by 60min",
slog.F("workspace_id", workspace.ID),
@@ -1727,7 +1730,12 @@ func (api *API) workspaceAgentReportStats(rw http.ResponseWriter, r *http.Reques
return xerrors.Errorf("can't get user: %w", err)
}
api.Options.UpdateAgentMetrics(ctx, user.Username, workspace.Name, workspaceAgent.Name, req.Metrics)
api.Options.UpdateAgentMetrics(ctx, prometheusmetrics.AgentMetricLabels{
Username: user.Username,
WorkspaceName: workspace.Name,
AgentName: workspaceAgent.Name,
TemplateName: row.TemplateName,
}, req.Metrics)
return nil
})
}
@@ -2103,7 +2111,7 @@ func (api *API) workspaceAgentReportLifecycle(rw http.ResponseWriter, r *http.Re
ctx := r.Context()
workspaceAgent := httpmw.WorkspaceAgent(r)
workspace, err := api.Database.GetWorkspaceByAgentID(ctx, workspaceAgent.ID)
row, err := api.Database.GetWorkspaceByAgentID(ctx, workspaceAgent.ID)
if err != nil {
httpapi.Write(ctx, rw, http.StatusBadRequest, codersdk.Response{
Message: "Failed to get workspace.",
@@ -2111,6 +2119,7 @@ func (api *API) workspaceAgentReportLifecycle(rw http.ResponseWriter, r *http.Re
})
return
}
workspace := row.Workspace
var req agentsdk.PostLifecycleRequest
if !httpapi.Read(ctx, rw, r, &req) {