Merge pull request #382 from pace/bun-poc

Add Bun backend

Author: monstermunchkin

backend/postgres/README.md

@@ -15,32 +15,12 @@ Configuration for the PostgreSQL connection pool of the microservice.
     * postgres user to access the database
 * `POSTGRES_DB` default: `postgres`
     * database to access
-* `POSTGRES_MAX_RETRIES` default: `5`
-    * Maximum number of retries before giving up
-* `POSTGRES_RETRY_STATEMENT_TIMEOUT` default: `false`
-    * Whether to retry queries cancelled because of statement_timeout
-* `POSTGRES_MIN_RETRY_BACKOFF` default: `250ms`
-    *  Minimum backoff between each retry
-* `POSTGRES_MAX_RETRY_BACKOFF` default: `4s`
-    * Maximum backoff between each retry
 * `POSTGRES_DIAL_TIMEOUT` default: `5s`
     * Dial timeout for establishing new connections
 * `POSTGRES_READ_TIMEOUT` default: `30s`
     *  Timeout for socket reads. If reached, commands will fail with a timeout instead of blocking
 * `POSTGRES_WRITE_TIMEOUT` default: `30s`
     * Timeout for socket writes. If reached, commands will fail with a timeout instead of blocking.
-* `POSTGRES_POOL_SIZE` default: `100`
-    * Maximum number of socket connections
-* `POSTGRES_MIN_IDLE_CONNECTIONS` default: `10`
-    * Minimum number of idle connections which is useful when establishing new connection is slow
-* `POSTGRES_MAX_CONN_AGE` default: `30m`
-    * Connection age at which client retires (closes) the connection
-* `POSTGRES_POOL_TIMEOUT` default: `31s`
-    * Time for which client waits for free connection if all connections are busy before returning an error
-* `POSTGRES_IDLE_TIMEOUT` default: `5m`
-    * Amount of time after which client closes idle connections
-* `POSTGRES_IDLE_CHECK_FREQUENCY` default: `1m`
-    * Frequency of idle checks made by idle connections reaper
 * `POSTGRES_HEALTH_CHECK_TABLE_NAME` default: `healthcheck`
     * Name of the Table that is created to try if database is writeable
 * `POSTGRES_HEALTH_CHECK_RESULT_TTL` default: `10s`
@@ -53,7 +33,6 @@ Prometheus metrics exposed.
 * `pace_postgres_query_total{database}` Collects stats about the number of postgres queries made
 * `pace_postgres_query_failed{database}` Collects stats about the number of postgres queries failed
 * `pace_postgres_query_duration_seconds{database}` Collects performance metrics for each postgres query
-* `pace_postgres_query_rows_total{database}` Collects stats about the number of rows returned by a postgres query
 * `pace_postgres_query_affected_total{database}` Collects stats about the number of rows affected by a postgres query
 * `pace_postgres_connection_pool_hits{database}` Collects number of times free connection was found in the pool
 * `pace_postgres_connection_pool_misses{database}` Collects number of times free connection was NOT found in the pool

backend/postgres/errors.go

@@ -7,32 +7,34 @@ import (
 	"io"
 	"net"
 
-	"github.com/go-pg/pg"
+	"github.com/uptrace/bun/driver/pgdriver"
 )
 
 var ErrNotUnique = errors.New("not unique")
 
 func IsErrConnectionFailed(err error) bool {
-	// go-pg has this check internally for network errors
+	// bun has this check internally for network errors
 	if errors.Is(err, io.EOF) {
 		return true
 	}
 
-	// go-pg has this check internally for network errors
+	// bun has this check internally for network errors
 	_, ok := err.(net.Error)
 	if ok {
 		return true
 	}
 
-	// go-pg has similar check for integrity violation issues, here we check network issues
-	pgErr, ok := err.(pg.Error)
-	if ok {
+	// bun has similar check for integrity violation issues, here we check network issues
+	var pgErr pgdriver.Error
+
+	if errors.As(err, &pgErr) {
 		code := pgErr.Field('C')
 		// We check on error codes of Class 08 — Connection Exception.
 		// https://www.postgresql.org/docs/10/errcodes-appendix.html
-		if code[0:2] == "08" {
+		if len(code) > 2 && code[0:2] == "08" {
 			return true
 		}
 	}
+
 	return false
 }

backend/postgres/errors_test.go

@@ -1,12 +1,12 @@
 package postgres_test
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"io"
 	"testing"
 
-	"github.com/go-pg/pg"
 	"github.com/stretchr/testify/require"
 
 	pbpostgres "github.com/pace/bricks/backend/postgres"
@@ -19,13 +19,10 @@ func TestIsErrConnectionFailed(t *testing.T) {
 	})
 
 	t.Run("connection failed (net.Error)", func(t *testing.T) {
-		db := pbpostgres.CustomConnectionPool(&pg.Options{}) // invalid connection
-		_, err := db.Exec("")
-		require.True(t, pbpostgres.IsErrConnectionFailed(err))
-	})
+		ctx := context.Background()
 
-	t.Run("connection failed (pg.Error)", func(t *testing.T) {
-		err := error(mockPGError{m: map[byte]string{'C': "08000"}})
+		db := pbpostgres.NewDB(ctx, pbpostgres.WithHost("foobar")) // invalid connection
+		_, err := db.Exec("")
 		require.True(t, pbpostgres.IsErrConnectionFailed(err))
 	})
 
@@ -34,11 +31,3 @@ func TestIsErrConnectionFailed(t *testing.T) {
 		require.False(t, pbpostgres.IsErrConnectionFailed(err))
 	})
 }
-
-type mockPGError struct {
-	m map[byte]string
-}
-
-func (err mockPGError) Field(k byte) string      { return err.m[k] }
-func (err mockPGError) IntegrityViolation() bool { return false }
-func (err mockPGError) Error() string            { return fmt.Sprintf("%+v", err.m) }

backend/postgres/health.go

@@ -0,0 +1,91 @@
+// Copyright © 2024 by PACE Telematics GmbH. All rights reserved.
+
+package postgres
+
+import (
+	"context"
+	"database/sql"
+	"time"
+
+	"github.com/pace/bricks/maintenance/health/servicehealthcheck"
+	"github.com/uptrace/bun"
+)
+
+type queryExecutor interface {
+	Exec(ctx context.Context, dest ...interface{}) (sql.Result, error)
+}
+
+// HealthCheck checks the state of a postgres connection. It must not be changed
+// after it was registered as a health check.
+type HealthCheck struct {
+	state servicehealthcheck.ConnectionState
+
+	createTableQueryExecutor queryExecutor
+	deleteQueryExecutor      queryExecutor
+	dropTableQueryExecutor   queryExecutor
+	insertQueryExecutor      queryExecutor
+	selectQueryExecutor      queryExecutor
+}
+
+type healthcheck struct {
+	bun.BaseModel
+
+	OK bool `bun:"column:ok"`
+}
+
+// NewHealthCheck creates a new HealthCheck instance.
+func NewHealthCheck(db *bun.DB) *HealthCheck {
+	return &HealthCheck{
+		createTableQueryExecutor: db.NewCreateTable().Model((*healthcheck)(nil)).ModelTableExpr(cfg.HealthCheckTableName).IfNotExists(),
+		deleteQueryExecutor:      db.NewDelete().ModelTableExpr(cfg.HealthCheckTableName).Where("TRUE"),
+		dropTableQueryExecutor:   db.NewDropTable().ModelTableExpr(cfg.HealthCheckTableName).IfExists(),
+		insertQueryExecutor:      db.NewInsert().ModelTableExpr(cfg.HealthCheckTableName).Model(&healthcheck{OK: true}),
+		selectQueryExecutor:      db.NewRaw("SELECT 1;"),
+	}
+}
+
+// Init initializes the test table
+func (h *HealthCheck) Init(ctx context.Context) error {
+	_, err := h.createTableQueryExecutor.Exec(ctx)
+	return err
+}
+
+// HealthCheck performs the read test on the database. If enabled, it performs a
+// write test as well.
+func (h *HealthCheck) HealthCheck(ctx context.Context) servicehealthcheck.HealthCheckResult {
+	if time.Since(h.state.LastChecked()) <= cfg.HealthCheckResultTTL {
+		// the last result of the Health Check is still not outdated
+		return h.state.GetState()
+	}
+
+	// Readcheck
+	if _, err := h.selectQueryExecutor.Exec(ctx); err != nil {
+		h.state.SetErrorState(err)
+		return h.state.GetState()
+	}
+
+	// writecheck - add Data to configured Table
+	if _, err := h.insertQueryExecutor.Exec(ctx); err != nil {
+		h.state.SetErrorState(err)
+		return h.state.GetState()
+	}
+
+	// and while we're at it, check delete as well (so as not to clutter the database
+	// because UPSERT is impractical here
+	if _, err := h.deleteQueryExecutor.Exec(ctx); err != nil {
+		h.state.SetErrorState(err)
+		return h.state.GetState()
+	}
+
+	// If no error occurred set the State of this Health Check to healthy
+	h.state.SetHealthy()
+
+	return h.state.GetState()
+}
+
+// CleanUp drops the test table.
+func (h *HealthCheck) CleanUp(ctx context.Context) error {
+	_, err := h.dropTableQueryExecutor.Exec(ctx)
+
+	return err
+}

backend/postgres/health_postgres.go

@@ -4,19 +4,20 @@ package postgres
 
 import (
 	"context"
+	"database/sql"
 	"io"
 	"net/http"
 	"net/http/httptest"
 	"strings"
 	"testing"
 	"time"
 
-	"github.com/go-pg/pg/orm"
+	"github.com/stretchr/testify/require"
+
 	http2 "github.com/pace/bricks/http"
 	"github.com/pace/bricks/maintenance/errors"
 	"github.com/pace/bricks/maintenance/health/servicehealthcheck"
 	"github.com/pace/bricks/maintenance/log"
-	"github.com/stretchr/testify/require"
 )
 
 func setup() *http.Response {
@@ -52,7 +53,7 @@ type testPool struct {
 	err error
 }
 
-func (t *testPool) Exec(ctx context.Context, query interface{}, params ...interface{}) (res orm.Result, err error) {
+func (t *testPool) Exec(ctx context.Context, dest ...any) (sql.Result, error) {
 	return nil, t.err
 }
 
@@ -63,16 +64,25 @@ func TestHealthCheckCaching(t *testing.T) {
 	cfg.HealthCheckResultTTL = time.Minute
 	requiredErr := errors.New("TestHealthCheckCaching")
 	pool := &testPool{err: requiredErr}
-	h := &HealthCheck{Pool: pool}
+	h := &HealthCheck{
+		createTableQueryExecutor: pool,
+		deleteQueryExecutor:      pool,
+		dropTableQueryExecutor:   pool,
+		insertQueryExecutor:      pool,
+		selectQueryExecutor:      pool,
+	}
+
 	res := h.HealthCheck(ctx)
 	// get the error for the first time
 	require.Equal(t, servicehealthcheck.Err, res.State)
 	require.Equal(t, "TestHealthCheckCaching", res.Msg)
+
 	res = h.HealthCheck(ctx)
 	pool.err = nil
 	// getting the cached error
 	require.Equal(t, servicehealthcheck.Err, res.State)
 	require.Equal(t, "TestHealthCheckCaching", res.Msg)
+
 	// Resetting the TTL to get a uncached result
 	cfg.HealthCheckResultTTL = 0
 	res = h.HealthCheck(ctx)