feat: Allow for pubsub connections to optionally be taken from pool

options.go

@@ -147,6 +147,11 @@ type Options struct {
 
 	// Add suffix to client name. Default is empty.
 	IdentitySuffix string
+
+	// Use connections from pool instead of creating new ones. Note that after use these connections will not be
+	// returned to the pool. Useful for managing the total Redis connection limit for a mix of Pubsub & other commands.
+	// Applies only to non-cluster client. Default is false.
+	PubsubFromPool bool
 }
 
 func (opt *Options) init() {

pubsub.go

@@ -12,6 +12,9 @@ import (
 	"github.com/redis/go-redis/v9/internal/proto"
 )
 
+type PubsubNewConnFunc func(ctx context.Context, channels []string) (*pool.Conn, error)
+type PubsubCloseConnFunc func(*pool.Conn) error
+
 // PubSub implements Pub/Sub commands as described in
 // http://redis.io/topics/pubsub. Message receiving is NOT safe
 // for concurrent use by multiple goroutines.
@@ -21,8 +24,8 @@ import (
 type PubSub struct {
 	opt *Options
 
-	newConn   func(ctx context.Context, channels []string) (*pool.Conn, error)
-	closeConn func(*pool.Conn) error
+	newConn   PubsubNewConnFunc
+	closeConn PubsubCloseConnFunc
 
 	mu        sync.Mutex
 	cn        *pool.Conn

pubsub_test.go

@@ -567,4 +567,104 @@ var _ = Describe("PubSub", func() {
 		Expect(msg.Channel).To(Equal("mychannel"))
 		Expect(msg.Payload).To(Equal(text))
 	})
+
+	It("should not use connections from pool", func() {
+		statsBefore := client.PoolStats()
+
+		pubsub := client.Subscribe(ctx, "mychannel")
+		defer pubsub.Close()
+
+		stats := client.PoolStats()
+		// A connection has been created
+		Expect(stats.TotalConns - statsBefore.TotalConns).To(Equal(uint32(1)))
+		// But it's not taken from the pool
+		poolFetchesBefore := statsBefore.Hits + statsBefore.Misses
+		poolFetchesAfter := stats.Hits + stats.Misses
+		Expect(poolFetchesAfter - poolFetchesBefore).To(Equal(uint32(0)))
+
+		pubsub.Close()
+
+		stats = client.PoolStats()
+		// The connection no longer exists
+		Expect(stats.TotalConns - statsBefore.TotalConns).To(Equal(uint32(0)))
+		Expect(stats.IdleConns - statsBefore.IdleConns).To(Equal(uint32(0)))
+	})
+})
+
+var _ = Describe("PubSub with PubsubFromPool set", func() {
+	var client *redis.Client
+
+	BeforeEach(func() {
+		opt := redisOptions()
+		opt.MinIdleConns = 0
+		opt.ConnMaxLifetime = 0
+		opt.PubsubFromPool = true
+		// zero value ends up using default so set small instead
+		opt.PoolTimeout = time.Microsecond
+		client = redis.NewClient(opt)
+		Expect(client.FlushDB(ctx).Err()).NotTo(HaveOccurred())
+	})
+
+	AfterEach(func() {
+		Expect(client.Close()).NotTo(HaveOccurred())
+	})
+
+	It("should use connection from pool", func() {
+		statsBefore := client.PoolStats()
+
+		pubsub := client.Subscribe(ctx, "mychannel")
+		defer pubsub.Close()
+
+		stats := client.PoolStats()
+		// A connection has been taken from the pool
+		Expect(stats.Hits - statsBefore.Hits).To(Equal(uint32(1)))
+		statsDuring := client.PoolStats()
+
+		pubsub.Close()
+
+		stats = client.PoolStats()
+		// It's not returned to the idle pool ..
+		Expect(statsDuring.IdleConns - stats.IdleConns).To(Equal(uint32(0)))
+		// .. and has been terminated
+		Expect(statsDuring.TotalConns - stats.TotalConns).To(Equal(uint32(1)))
+	})
+
+	It("should respect pool size limit", func() {
+		poolSize := client.Options().PoolSize
+		statsBefore := client.PoolStats()
+
+		var pubsubs []*redis.PubSub
+		for i := 0; i < poolSize; i++ {
+			pubsub := client.Subscribe(ctx, "mychannel")
+			defer pubsub.Close()
+			pubsubs = append(pubsubs, pubsub)
+		}
+
+		statsDuring := client.PoolStats()
+		poolFetchesBefore := statsBefore.Hits + statsBefore.Misses
+		poolFetchesAfter := statsDuring.Hits + statsDuring.Misses
+
+		// A total of poolSize connections should have been taken from the pool (new or existing)
+		Expect(poolFetchesAfter - poolFetchesBefore).To(Equal(uint32(poolSize)))
+
+		// The next pubsub connection should fail to connect (waiting for pool)
+		extraPubsub := client.Subscribe(ctx, "mychannel")
+		defer extraPubsub.Close()
+		Expect(client.PoolStats().Timeouts - statsDuring.Timeouts).To(Equal(uint32(1)))
+
+		// As should retries
+		err := extraPubsub.Ping(ctx)
+		Expect(err).To(MatchError(ContainSubstring("connection pool timeout")))
+		Expect(client.PoolStats().Timeouts - statsDuring.Timeouts).To(Equal(uint32(2)))
+
+		for _, pubsub := range pubsubs {
+			pubsub.Close()
+		}
+
+		stats := client.PoolStats()
+		// Connections are not returned to the idle pool ..
+		Expect(statsDuring.IdleConns - stats.IdleConns).To(Equal(uint32(0)))
+		// .. and have been terminated
+		Expect(statsDuring.TotalConns - stats.TotalConns).To(Equal(uint32(poolSize)))
+	})
 })

redis.go

@@ -199,6 +199,9 @@ type baseClient struct {
 	opt      *Options
 	connPool pool.Pooler
 
+	pubsubNewConn   PubsubNewConnFunc
+	pubsubCloseConn PubsubCloseConnFunc
+
 	onClose func() error // hook called when client is closed
 }
 
@@ -368,6 +371,13 @@ func (c *baseClient) releaseConn(ctx context.Context, cn *pool.Conn, err error)
 	}
 }
 
+func (c *baseClient) removeConn(ctx context.Context, cn *pool.Conn, err error) {
+	if c.opt.Limiter != nil {
+		c.opt.Limiter.ReportResult(err)
+	}
+	c.connPool.Remove(ctx, cn, err)
+}
+
 func (c *baseClient) withConn(
 	ctx context.Context, fn func(context.Context, *pool.Conn) error,
 ) error {
@@ -649,6 +659,28 @@ func (c *Client) init() {
 		pipeline:   c.baseClient.processPipeline,
 		txPipeline: c.baseClient.processTxPipeline,
 	})
+
+	if c.opt.PubsubFromPool {
+		// Take connections from pool and remove them from pool afterwards. (Pubsub & other connections are managed
+		// together.)
+		c.pubsubNewConn = func(ctx context.Context, channels []string) (*pool.Conn, error) {
+			return c.getConn(ctx)
+		}
+		c.pubsubCloseConn = func(conn *pool.Conn) error {
+			c.removeConn(context.TODO(), conn, nil)
+			return nil
+		}
+	} else {
+		// Make brand new connection from pool and close it afterwards. (Pubsub & other connections are managed
+		// independently other than that pubsub connection can no longer be created once the pool is full.)
+		c.pubsubNewConn = func(ctx context.Context, channels []string) (*pool.Conn, error) {
+			return c.newConn(ctx)
+		}
+		// wrapping in closure since pool has not been initialised yet
+		c.pubsubCloseConn = func(conn *pool.Conn) error {
+			return c.connPool.CloseConn(conn)
+		}
+	}
 }
 
 func (c *Client) WithTimeout(timeout time.Duration) *Client {
@@ -720,10 +752,8 @@ func (c *Client) pubSub() *PubSub {
 	pubsub := &PubSub{
 		opt: c.opt,
 
-		newConn: func(ctx context.Context, channels []string) (*pool.Conn, error) {
-			return c.newConn(ctx)
-		},
-		closeConn: c.connPool.CloseConn,
+		newConn:   c.pubsubNewConn,
+		closeConn: c.pubsubCloseConn,
 	}
 	pubsub.init()
 	return pubsub