feat: add connection statistics for workspace agents (#6469)

* fix: don't make session counts cumulative

This made for some weird tracking... we want the point-in-time
number of counts!

* Add databasefake query for getting agent stats

* Add deployment stats endpoint

* The query... works?!?

* Fix aggregation query

* Select from multiple tables instead

* Fix continuous stats

* Increase period of stat refreshes

* Add workspace counts to deployment stats

* fmt

* Add a slight bit of responsiveness

* Fix template version editor overflow

* Add refresh button

* Fix font family on button

* Fix latest stat being reported

* Revert agent conn stats

* Fix linting error

* Fix tests

* Fix gen

* Fix migrations

* Block on sending stat updates

* Add test fixtures

* Fix response structure

* make gen
This commit is contained in:
Kyle Carberry
2023-03-08 21:05:45 -06:00
committed by GitHub
parent 9d40d2ffdc
commit 5304b4e483
43 changed files with 1790 additions and 174 deletions
+86 -20
View File
@@ -3,6 +3,7 @@ package metricscache
import (
"context"
"database/sql"
"sync"
"sync/atomic"
"time"
@@ -25,34 +26,56 @@ import (
// take a few hundred milliseconds, which would ruin page load times and
// database performance if in the hot path.
type Cache struct {
database database.Store
log slog.Logger
database database.Store
log slog.Logger
intervals Intervals
deploymentDAUResponses atomic.Pointer[codersdk.DeploymentDAUsResponse]
templateDAUResponses atomic.Pointer[map[uuid.UUID]codersdk.TemplateDAUsResponse]
templateUniqueUsers atomic.Pointer[map[uuid.UUID]int]
templateAverageBuildTime atomic.Pointer[map[uuid.UUID]database.GetTemplateAverageBuildTimeRow]
deploymentStatsResponse atomic.Pointer[codersdk.DeploymentStats]
done chan struct{}
cancel func()
interval time.Duration
}
func New(db database.Store, log slog.Logger, interval time.Duration) *Cache {
if interval <= 0 {
interval = time.Hour
type Intervals struct {
TemplateDAUs time.Duration
DeploymentStats time.Duration
}
func New(db database.Store, log slog.Logger, intervals Intervals) *Cache {
if intervals.TemplateDAUs <= 0 {
intervals.TemplateDAUs = time.Hour
}
if intervals.DeploymentStats <= 0 {
intervals.DeploymentStats = time.Minute
}
ctx, cancel := context.WithCancel(context.Background())
c := &Cache{
database: db,
log: log,
done: make(chan struct{}),
cancel: cancel,
interval: interval,
database: db,
intervals: intervals,
log: log,
done: make(chan struct{}),
cancel: cancel,
}
go c.run(ctx)
go func() {
var wg sync.WaitGroup
defer close(c.done)
wg.Add(1)
go func() {
defer wg.Done()
c.run(ctx, "template daus", intervals.TemplateDAUs, c.refreshTemplateDAUs)
}()
wg.Add(1)
go func() {
defer wg.Done()
c.run(ctx, "deployment stats", intervals.DeploymentStats, c.refreshDeploymentStats)
}()
wg.Wait()
}()
return c
}
@@ -142,7 +165,7 @@ func countUniqueUsers(rows []database.GetTemplateDAUsRow) int {
return len(seen)
}
func (c *Cache) refresh(ctx context.Context) error {
func (c *Cache) refreshTemplateDAUs(ctx context.Context) error {
//nolint:gocritic // This is a system service.
ctx = dbauthz.AsSystemRestricted(ctx)
err := c.database.DeleteOldWorkspaceAgentStats(ctx)
@@ -199,16 +222,51 @@ func (c *Cache) refresh(ctx context.Context) error {
return nil
}
func (c *Cache) run(ctx context.Context) {
defer close(c.done)
func (c *Cache) refreshDeploymentStats(ctx context.Context) error {
from := database.Now().Add(-15 * time.Minute)
agentStats, err := c.database.GetDeploymentWorkspaceAgentStats(ctx, from)
if err != nil {
return err
}
workspaceStats, err := c.database.GetDeploymentWorkspaceStats(ctx)
if err != nil {
return err
}
c.deploymentStatsResponse.Store(&codersdk.DeploymentStats{
AggregatedFrom: from,
CollectedAt: database.Now(),
NextUpdateAt: database.Now().Add(c.intervals.DeploymentStats),
Workspaces: codersdk.WorkspaceDeploymentStats{
Pending: workspaceStats.PendingWorkspaces,
Building: workspaceStats.BuildingWorkspaces,
Running: workspaceStats.RunningWorkspaces,
Failed: workspaceStats.FailedWorkspaces,
Stopped: workspaceStats.StoppedWorkspaces,
ConnectionLatencyMS: codersdk.WorkspaceConnectionLatencyMS{
P50: agentStats.WorkspaceConnectionLatency50,
P95: agentStats.WorkspaceConnectionLatency95,
},
RxBytes: agentStats.WorkspaceRxBytes,
TxBytes: agentStats.WorkspaceTxBytes,
},
SessionCount: codersdk.SessionCountDeploymentStats{
VSCode: agentStats.SessionCountVSCode,
SSH: agentStats.SessionCountSSH,
JetBrains: agentStats.SessionCountJetBrains,
ReconnectingPTY: agentStats.SessionCountReconnectingPTY,
},
})
return nil
}
ticker := time.NewTicker(c.interval)
func (c *Cache) run(ctx context.Context, name string, interval time.Duration, refresh func(context.Context) error) {
ticker := time.NewTicker(interval)
defer ticker.Stop()
for {
for r := retry.New(time.Millisecond*100, time.Minute); r.Wait(ctx); {
start := time.Now()
err := c.refresh(ctx)
err := refresh(ctx)
if err != nil {
if ctx.Err() != nil {
return
@@ -218,9 +276,9 @@ func (c *Cache) run(ctx context.Context) {
}
c.log.Debug(
ctx,
"metrics refreshed",
name+" metrics refreshed",
slog.F("took", time.Since(start)),
slog.F("interval", c.interval),
slog.F("interval", interval),
)
break
}
@@ -322,3 +380,11 @@ func (c *Cache) TemplateBuildTimeStats(id uuid.UUID) codersdk.TemplateBuildTimeS
},
}
}
func (c *Cache) DeploymentStats() (codersdk.DeploymentStats, bool) {
deploymentStats := c.deploymentStatsResponse.Load()
if deploymentStats == nil {
return codersdk.DeploymentStats{}, false
}
return *deploymentStats, true
}
+33 -2
View File
@@ -164,7 +164,9 @@ func TestCache_TemplateUsers(t *testing.T) {
t.Parallel()
var (
db = dbfake.New()
cache = metricscache.New(db, slogtest.Make(t, nil), testutil.IntervalFast)
cache = metricscache.New(db, slogtest.Make(t, nil), metricscache.Intervals{
TemplateDAUs: testutil.IntervalFast,
})
)
defer cache.Close()
@@ -286,7 +288,9 @@ func TestCache_BuildTime(t *testing.T) {
var (
db = dbfake.New()
cache = metricscache.New(db, slogtest.Make(t, nil), testutil.IntervalFast)
cache = metricscache.New(db, slogtest.Make(t, nil), metricscache.Intervals{
TemplateDAUs: testutil.IntervalFast,
})
)
defer cache.Close()
@@ -370,3 +374,30 @@ func TestCache_BuildTime(t *testing.T) {
})
}
}
func TestCache_DeploymentStats(t *testing.T) {
t.Parallel()
db := dbfake.New()
cache := metricscache.New(db, slogtest.Make(t, nil), metricscache.Intervals{
DeploymentStats: testutil.IntervalFast,
})
_, err := db.InsertWorkspaceAgentStat(context.Background(), database.InsertWorkspaceAgentStatParams{
ID: uuid.New(),
AgentID: uuid.New(),
CreatedAt: database.Now(),
ConnectionCount: 1,
RxBytes: 1,
TxBytes: 1,
SessionCountVSCode: 1,
})
require.NoError(t, err)
var stat codersdk.DeploymentStats
require.Eventually(t, func() bool {
var ok bool
stat, ok = cache.DeploymentStats()
return ok
}, testutil.WaitLong, testutil.IntervalMedium)
require.Equal(t, int64(1), stat.SessionCount.VSCode)
}