database_observability: report health of component and collectors

CHANGELOG.md

@@ -21,6 +21,7 @@ Main (unreleased)
 ### Enhancements
 
 - Update `prometheus.write.queue` to support v2 for cpu performance. (@mattdurham)
+- (_Experimental_) Add health reporting to `database_observability.mysql` component (@cristiangreco)
 
 v1.6.0-rc.0
 -----------------
@@ -132,26 +133,26 @@ v1.6.0-rc.0
     https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/36071
   - `otelcol.exporter.datadog`: Stop prefixing `http_server_duration`, `http_server_request_size` and `http_server_response_size` with `otelcol`.
     https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/36265
-    These metrics can be from SDKs rather than collector. Stop prefixing them to be consistent with 
+    These metrics can be from SDKs rather than collector. Stop prefixing them to be consistent with
     https://opentelemetry.io/docs/collector/internal-telemetry/#lists-of-internal-metrics
   - `otelcol.receiver.datadog`: Add json handling for the `api/v2/series` endpoint in the datadogreceiver.
     https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/36218
-  - `otelcol.processor.span`: Add a new `keep_original_name` configuration argument 
+  - `otelcol.processor.span`: Add a new `keep_original_name` configuration argument
     to keep the original span name when extracting attributes from the span name.
     https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/36397
   - `pkg/ottl`: Respect the `depth` option when flattening slices using `flatten`.
     The `depth` option is also now required to be at least `1`.
     https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/36198
   - `otelcol.exporter.loadbalancing`: Shutdown exporters during collector shutdown. This fixes a memory leak.
     https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/36024
-  - `otelcol.processor.k8sattributes`: New `wait_for_metadata` and `wait_for_metadata_timeout` configuration arguments, 
+  - `otelcol.processor.k8sattributes`: New `wait_for_metadata` and `wait_for_metadata_timeout` configuration arguments,
     which block the processor startup until metadata is received from Kubernetes.
     https://github.com/open-telemetry/opentelemetry-collector-contrib/issues/32556
-  - `otelcol.processor.k8sattributes`: Enable the `k8sattr.fieldExtractConfigRegex.disallow` for all Alloy instances, 
+  - `otelcol.processor.k8sattributes`: Enable the `k8sattr.fieldExtractConfigRegex.disallow` for all Alloy instances,
     to retain the behavior of `regex` argument in the `annotation` and `label` blocks.
     When the feature gate is "deprecated" in the upstream Collector, Alloy users will need to use the transform processor instead.
     https://github.com/open-telemetry/opentelemetry-collector-contrib/issues/25128
-  - `otelcol.receiver.vcenter`: The existing code did not honor TLS settings beyond 'insecure'. 
+  - `otelcol.receiver.vcenter`: The existing code did not honor TLS settings beyond 'insecure'.
     All TLS client config should now be honored.
     https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/36482
   - `otelcol.receiver.opencensus`: Do not report error message when OpenCensus receiver is shutdown cleanly.

internal/component/database_observability/mysql/collector/connection_info.go

@@ -8,6 +8,7 @@ import (
 
 	"github.com/go-sql-driver/mysql"
 	"github.com/prometheus/client_golang/prometheus"
+	"go.uber.org/atomic"
 )
 
 var rdsRegex = regexp.MustCompile(`(?P<identifier>[^\.]+)\.([^\.]+)\.(?P<region>[^\.]+)\.rds\.amazonaws\.com`)
@@ -21,6 +22,8 @@ type ConnectionInfo struct {
 	DSN        string
 	Registry   *prometheus.Registry
 	InfoMetric *prometheus.GaugeVec
+
+	running *atomic.Bool
 }
 
 func NewConnectionInfo(args ConnectionInfoArguments) (*ConnectionInfo, error) {
@@ -35,15 +38,22 @@ func NewConnectionInfo(args ConnectionInfoArguments) (*ConnectionInfo, error) {
 		DSN:        args.DSN,
 		Registry:   args.Registry,
 		InfoMetric: infoMetric,
+		running:    &atomic.Bool{},
 	}, nil
 }
 
+func (c *ConnectionInfo) Name() string {
+	return "ConnectionInfo"
+}
+
 func (c *ConnectionInfo) Start(ctx context.Context) error {
 	cfg, err := mysql.ParseDSN(c.DSN)
 	if err != nil {
 		return err
 	}
 
+	c.running.Store(true)
+
 	var (
 		providerName         = "unknown"
 		providerRegion       = "unknown"
@@ -66,6 +76,11 @@ func (c *ConnectionInfo) Start(ctx context.Context) error {
 	return nil
 }
 
+func (c *ConnectionInfo) Stopped() bool {
+	return !c.running.Load()
+}
+
 func (c *ConnectionInfo) Stop() {
 	c.Registry.Unregister(c.InfoMetric)
+	c.running.Store(false)
 }

internal/component/database_observability/mysql/collector/query_sample.go

@@ -5,12 +5,12 @@ import (
 	"database/sql"
 	"fmt"
 	"strings"
-	"sync/atomic"
 	"time"
 
 	"github.com/go-kit/log"
 	"github.com/prometheus/common/model"
 	"github.com/xwb1989/sqlparser"
+	"go.uber.org/atomic"
 
 	"github.com/grafana/alloy/internal/component/common/loki"
 	"github.com/grafana/alloy/internal/component/database_observability"
@@ -57,11 +57,15 @@ func NewQuerySample(args QuerySampleArguments) (*QuerySample, error) {
 		dbConnection:    args.DB,
 		collectInterval: args.CollectInterval,
 		entryHandler:    args.EntryHandler,
-		logger:          args.Logger,
+		logger:          log.With(args.Logger, "collector", "QuerySample"),
 		running:         &atomic.Bool{},
 	}, nil
 }
 
+func (c *QuerySample) Name() string {
+	return "QuerySample"
+}
+
 func (c *QuerySample) Start(ctx context.Context) error {
 	level.Debug(c.logger).Log("msg", "QuerySample collector started")
 
@@ -81,6 +85,7 @@ func (c *QuerySample) Start(ctx context.Context) error {
 		for {
 			if err := c.fetchQuerySamples(c.ctx); err != nil {
 				level.Error(c.logger).Log("msg", "collector stopping due to error", "err", err)
+				c.Stop()
 				break
 			}
 
@@ -127,7 +132,7 @@ func (c *QuerySample) fetchQuerySamples(ctx context.Context) error {
 		}
 
 		if strings.HasSuffix(sampleText, "...") {
-			level.Info(c.logger).Log("msg", "skipping parsing truncated query", "digest", digest)
+			level.Debug(c.logger).Log("msg", "skipping parsing truncated query", "digest", digest)
 			continue
 		}
 

internal/component/database_observability/mysql/collector/schema_table.go

@@ -10,6 +10,7 @@ import (
 	"github.com/grafana/loki/v3/pkg/logproto"
 	"github.com/hashicorp/golang-lru/v2/expirable"
 	"github.com/prometheus/common/model"
+	"go.uber.org/atomic"
 
 	"github.com/grafana/alloy/internal/component/common/loki"
 	"github.com/grafana/alloy/internal/component/database_observability"
@@ -66,10 +67,10 @@ type SchemaTable struct {
 	// TODO(cristian): allow configuring cache size (currently unlimited).
 	cache *expirable.LRU[string, tableInfo]
 
-	logger log.Logger
-
-	ctx    context.Context
-	cancel context.CancelFunc
+	logger  log.Logger
+	running *atomic.Bool
+	ctx     context.Context
+	cancel  context.CancelFunc
 }
 
 type tableInfo struct {
@@ -86,18 +87,29 @@ func NewSchemaTable(args SchemaTableArguments) (*SchemaTable, error) {
 		collectInterval: args.CollectInterval,
 		entryHandler:    args.EntryHandler,
 		cache:           expirable.NewLRU[string, tableInfo](0, nil, args.CacheTTL),
-		logger:          args.Logger,
+		logger:          log.With(args.Logger, "collector", "SchemaTable"),
+		running:         &atomic.Bool{},
 	}, nil
 }
 
+func (c *SchemaTable) Name() string {
+	return "SchemaTable"
+}
+
 func (c *SchemaTable) Start(ctx context.Context) error {
 	level.Debug(c.logger).Log("msg", "SchemaTable collector started")
 
+	c.running.Store(true)
 	ctx, cancel := context.WithCancel(ctx)
 	c.ctx = ctx
 	c.cancel = cancel
 
 	go func() {
+		defer func() {
+			c.Stop()
+			c.running.Store(false)
+		}()
+
 		ticker := time.NewTicker(c.collectInterval)
 
 		for {
@@ -119,6 +131,10 @@ func (c *SchemaTable) Start(ctx context.Context) error {
 	return nil
 }
 
+func (c *SchemaTable) Stopped() bool {
+	return !c.running.Load()
+}
+
 // Stop should be kept idempotent
 func (c *SchemaTable) Stop() {
 	c.cancel()
@@ -155,6 +171,11 @@ func (c *SchemaTable) extractSchema(ctx context.Context) error {
 		}
 	}
 
+	if len(schemas) == 0 {
+		level.Info(c.logger).Log("msg", "no schema detected from information_schema.schemata")
+		return nil
+	}
+
 	tables := []tableInfo{}
 
 	for _, schema := range schemas {

internal/component/database_observability/mysql/component.go

@@ -15,6 +15,7 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/client_golang/prometheus/promhttp"
 	"github.com/prometheus/common/model"
+	"go.uber.org/atomic"
 
 	"github.com/grafana/alloy/internal/component"
 	"github.com/grafana/alloy/internal/component/common/loki"
@@ -79,12 +80,15 @@ type Exports struct {
 }
 
 var (
-	_ component.Component    = (*Component)(nil)
-	_ http_service.Component = (*Component)(nil)
+	_ component.Component       = (*Component)(nil)
+	_ http_service.Component    = (*Component)(nil)
+	_ component.HealthComponent = (*Component)(nil)
 )
 
 type Collector interface {
+	Name() string
 	Start(context.Context) error
+	Stopped() bool
 	Stop()
 }
 
@@ -98,6 +102,7 @@ type Component struct {
 	baseTarget   discovery.Target
 	collectors   []Collector
 	dbConnection *sql.DB
+	healthErr    *atomic.String
 }
 
 func New(opts component.Options, args Arguments) (*Component, error) {
@@ -107,6 +112,7 @@ func New(opts component.Options, args Arguments) (*Component, error) {
 		receivers: args.ForwardTo,
 		handler:   loki.NewLogsReceiver(),
 		registry:  prometheus.NewRegistry(),
+		healthErr: atomic.NewString(""),
 	}
 
 	baseTarget, err := c.getBaseTarget()
@@ -184,6 +190,16 @@ func (c *Component) Update(args component.Arguments) error {
 
 	c.args = args.(Arguments)
 
+	if err := c.startCollectors(); err != nil {
+		c.healthErr.Store(err.Error())
+		return err
+	}
+
+	c.healthErr.Store("")
+	return nil
+}
+
+func (c *Component) startCollectors() error {
 	dbConnection, err := sql.Open("mysql", formatDSN(string(c.args.DataSourceName), "parseTime=true"))
 	if err != nil {
 		return err
@@ -254,6 +270,40 @@ func (c *Component) Handler() http.Handler {
 	return promhttp.HandlerFor(c.registry, promhttp.HandlerOpts{})
 }
 
+func (c *Component) CurrentHealth() component.Health {
+	if err := c.healthErr.Load(); err != "" {
+		return component.Health{
+			Health:     component.HealthTypeUnhealthy,
+			Message:    err,
+			UpdateTime: time.Now(),
+		}
+	}
+
+	var unhealthyCollectors []string
+
+	c.mut.RLock()
+	for _, collector := range c.collectors {
+		if collector.Stopped() {
+			unhealthyCollectors = append(unhealthyCollectors, collector.Name())
+		}
+	}
+	c.mut.RUnlock()
+
+	if len(unhealthyCollectors) > 0 {
+		return component.Health{
+			Health:     component.HealthTypeUnhealthy,
+			Message:    "One or more collectors are unhealthy: [" + strings.Join(unhealthyCollectors, ", ") + "]",
+			UpdateTime: time.Now(),
+		}
+	}
+
+	return component.Health{
+		Health:     component.HealthTypeHealthy,
+		Message:    "All collectors are healthy",
+		UpdateTime: time.Now(),
+	}
+}
+
 // instanceKey returns network(hostname:port)/dbname of the MySQL server.
 // This is the same key as used by the mysqld_exporter integration.
 func (c *Component) instanceKey() string {