Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

fix(eventbus): Refine auto-reconnection and bugfix #761

Merged
merged 5 commits into from
Jul 22, 2020
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 19 additions & 9 deletions eventbus/driver/nats.go
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,9 @@ import (
type natsStreamingConnection struct {
natsConn *nats.Conn
stanConn stan.Conn

natsConnected bool
stanConnected bool
}

func (nsc *natsStreamingConnection) Close() error {
Expand All @@ -37,10 +40,10 @@ func (nsc *natsStreamingConnection) Close() error {
}

func (nsc *natsStreamingConnection) IsClosed() bool {
if nsc.natsConn == nil {
if nsc.natsConn == nil || nsc.stanConn == nil || !nsc.natsConnected || !nsc.stanConnected || nsc.natsConn.IsClosed() {
return true
}
return nsc.natsConn.IsClosed()
return false
}

func (nsc *natsStreamingConnection) Publish(subject string, data []byte) error {
Expand Down Expand Up @@ -71,13 +74,16 @@ func NewNATSStreaming(url, clusterID, subject, clientID string, auth *Auth, logg

func (n *natsStreaming) Connect() (Connection, error) {
log := n.logger.WithField("clientID", n.clientID)
conn := &natsStreamingConnection{}
opts := []nats.Option{
nats.MaxReconnects(-1),
nats.ReconnectWait(3 * time.Second),
// Do not reconnect here but handle reconnction outside
nats.NoReconnect(),
nats.DisconnectErrHandler(func(nc *nats.Conn, err error) {
conn.natsConnected = false
log.Errorf("NATS connection lost, reason: %v", err)
}),
nats.ReconnectHandler(func(nnc *nats.Conn) {
conn.natsConnected = true
log.Info("Reconnected to NATS server")
}),
}
Expand All @@ -96,20 +102,22 @@ func (n *natsStreaming) Connect() (Connection, error) {
return nil, err
}
log.Info("Connected to NATS server.")
conn.natsConn = nc
conn.natsConnected = true

sc, err := stan.Connect(n.clusterID, n.clientID, stan.NatsConn(nc), stan.Pings(5, 60),
stan.SetConnectionLostHandler(func(_ stan.Conn, reason error) {
conn.stanConnected = false
log.Errorf("NATS streaming connection lost, reason: %v", reason)
}))
if err != nil {
log.Errorf("Failed to connect to NATS streaming server, %v", err)
return nil, err
}
log.Info("Connected to NATS streaming server.")
return &natsStreamingConnection{
natsConn: nc,
stanConn: sc,
}, nil
conn.stanConn = sc
conn.stanConnected = true
return conn, nil
}

func (n *natsStreaming) Publish(conn Connection, message []byte) error {
Expand Down Expand Up @@ -138,7 +146,7 @@ func (n *natsStreaming) SubscribeEventSources(ctx context.Context, conn Connecti
n.processEventSourceMsg(m, msgHolder, filter, action, log)
}, stan.DurableName(durableName),
stan.SetManualAckMode(),
stan.StartAt(pb.StartPosition_NewOnly),
stan.StartAt(pb.StartPosition_LastReceived),
stan.AckWait(1*time.Second),
stan.MaxInflight(len(msgHolder.depNames)+2))
if err != nil {
Expand All @@ -152,10 +160,12 @@ func (n *natsStreaming) SubscribeEventSources(ctx context.Context, conn Connecti
log.Info("existing, unsubscribing and closing connection...")
_ = sub.Close()
log.Infof("subscription on subject %s closed", n.subject)
return nil
case <-closeCh:
log.Info("closing subscription...")
_ = sub.Close()
log.Infof("subscription on subject %s closed", n.subject)
return nil
}
}
}
Expand Down
30 changes: 17 additions & 13 deletions eventsources/eventing.go
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,7 @@ import (

cloudevents "github.com/cloudevents/sdk-go/v2"
"github.com/google/uuid"
"github.com/pkg/errors"

"github.com/argoproj/argo-events/common/logging"
"github.com/argoproj/argo-events/eventbus"
Expand Down Expand Up @@ -255,27 +256,27 @@ func (e *EventSourceAdaptor) Start(ctx context.Context, stopCh <-chan struct{})
defer e.eventBusConn.Close()

// Daemon to reconnect
go func(ctx context.Context) {
go func() {
logger.Info("starting eventbus connection daemon...")
ticker := time.NewTicker(5 * time.Second)
for {
select {
case <-ctx.Done():
case <-cctx.Done():
logger.Info("exiting eventbus connection daemon...")
return
default:
time.Sleep(3 * time.Second)
}
if e.eventBusConn == nil || e.eventBusConn.IsClosed() {
logger.Info("NATS connection lost, reconnecting...")
e.eventBusConn, err = driver.Connect()
if err != nil {
logger.WithError(err).Errorln("failed to reconnect to eventbus")
continue
case <-ticker.C:
if e.eventBusConn == nil || e.eventBusConn.IsClosed() {
logger.Info("NATS connection lost, reconnecting...")
e.eventBusConn, err = driver.Connect()
if err != nil {
logger.WithError(err).Errorln("failed to reconnect to eventbus")
continue
}
logger.Info("reconnected the NATS streaming server...")
}
logger.Info("reconnected the NATS streaming server...")
}
}
}(cctx)
}()

for _, ss := range servers {
for _, server := range ss {
Expand All @@ -300,6 +301,9 @@ func (e *EventSourceAdaptor) Start(ctx context.Context, stopCh <-chan struct{})
if err != nil {
return err
}
if e.eventBusConn == nil || e.eventBusConn.IsClosed() {
return errors.New("failed to publish event, eventbus connection closed")
}
return driver.Publish(e.eventBusConn, eventBody)
})
logger.WithField(logging.LabelEventSourceName, s.GetEventSourceName()).
Expand Down
10 changes: 2 additions & 8 deletions go.mod
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,6 @@ require (
github.com/ahmetb/gen-crd-api-reference-docs v0.2.0
github.com/antonmedv/expr v1.8.8
github.com/apache/openwhisk-client-go v0.0.0-20190915054138-716c6f973eb2
github.com/argoproj/pkg v0.0.0-20200319004004-f46beff7cd54 // indirect
github.com/aws/aws-sdk-go v1.30.7
github.com/cloudevents/sdk-go/v2 v2.1.0
github.com/cloudfoundry/jibber_jabber v0.0.0-20151120183258-bcc4c8345a21 // indirect
Expand All @@ -36,7 +35,6 @@ require (
github.com/google/go-cmp v0.4.0
github.com/google/go-github/v31 v31.0.0
github.com/google/gofuzz v1.1.0 // indirect
github.com/google/shlex v0.0.0-20191202100458-e7afc7fbc510 // indirect
github.com/google/uuid v1.1.1
github.com/googleapis/gnostic v0.4.0 // indirect
github.com/gopherjs/gopherjs v0.0.0-20200217142428-fce0ec30dd00 // indirect
Expand All @@ -49,10 +47,9 @@ require (
github.com/imdario/mergo v0.3.9
github.com/joncalhoun/qson v0.0.0-20200422171543-84433dcd3da0
github.com/json-iterator/go v1.1.9 // indirect
github.com/k0kubun/colorstring v0.0.0-20150214042306-9440f1994b88 // indirect
github.com/kballard/go-shellquote v0.0.0-20180428030007-95032a82bc51 // indirect
github.com/klauspost/compress v1.10.4 // indirect
github.com/konsorten/go-windows-terminal-sequences v1.0.2 // indirect
github.com/magiconair/properties v1.8.1 // indirect
github.com/mailru/easyjson v0.7.1 // indirect
github.com/mattn/go-colorable v0.1.6 // indirect
github.com/mattn/go-isatty v0.0.12
Expand Down Expand Up @@ -80,16 +77,14 @@ require (
github.com/sirupsen/logrus v1.5.0
github.com/smartystreets/assertions v0.0.0-20190401211740-f487f9de1cd3 // indirect
github.com/smartystreets/goconvey v1.6.4
github.com/spf13/jwalterweatherman v1.1.0 // indirect
github.com/spf13/viper v1.3.2
github.com/streadway/amqp v1.0.0
github.com/stretchr/testify v1.5.1
github.com/stripe/stripe-go v70.15.0+incompatible
github.com/tidwall/gjson v1.6.0
github.com/tidwall/sjson v1.1.1
github.com/xanzy/go-gitlab v0.33.0
github.com/yudai/gojsondiff v1.0.0 // indirect
github.com/yudai/golcs v0.0.0-20170316035057-ecda9a501e82 // indirect
github.com/yudai/pp v2.0.1+incompatible // indirect
go.opencensus.io v0.22.3 // indirect
go.uber.org/zap v1.14.1
golang.org/x/crypto v0.0.0-20200429183012-4b2356b1ed79
Expand All @@ -114,7 +109,6 @@ require (
k8s.io/code-generator v0.17.5
k8s.io/gengo v0.0.0-20190822140433-26a664648505
k8s.io/kube-openapi v0.0.0-20200316234421-82d701f24f9d
k8s.io/kubernetes v1.17.5 // indirect
k8s.io/utils v0.0.0-20200327001022-6496210b90e8 // indirect
sigs.k8s.io/controller-runtime v0.5.4
sigs.k8s.io/controller-tools v0.2.5
Expand Down
Loading