TrueCloudLab/frostfs-node
20
1
Fork 27
Code Issues 101 Pull requests 5 Projects Releases 1 Activity Actions

[#195] morph: use blocking unlimited pool for notifications

Browse source 
With non-blocking pool restricted by 10 in capacity, the probability of
dropping events is unexpectedly big. Notifications are an essential part of the FrostFS,
we should not drop anything, especially new epochs.
```
Mar 31 07:07:03 vedi neofs-ir[19164]: 2023-03-31T07:07:03.901Z        debug        subscriber/subscriber.go:154        new notification event from sidechain        {"name": "NewEpoch"}
Mar 31 07:07:03 vedi neofs-ir[19164]: 2023-03-31T07:07:03.901Z        warn        event/listener.go:248        listener worker pool drained        {"chain": "morph", "capacity": 10}
```

Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
This commit is contained in:
Evgenii Stratonikov 2023-04-03 10:44:05 +03:00
parent ed28ce24cd
commit 03bef2bc98
3 changed files with 7 additions and 33 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

13
cmd/frostfs-node/morph.go
View file

@ -181,14 +181,6 @@ func waitNotaryDeposit(ctx context.Context, c *cfg, tx util.Uint256) error {
}
func listenMorphNotifications(ctx context.Context, c *cfg) {
// listenerPoolCap is a capacity of a
// worker pool inside the listener. It
// is used to prevent blocking in neo-go:
// the client cannot make RPC requests if
// the notification channel is not being
// read by another goroutine.
const listenerPoolCap = 10
var (
err error
subs subscriber.Subscriber
@ -208,9 +200,8 @@ func listenMorphNotifications(ctx context.Context, c *cfg) {
fatalOnErr(err)
lis, err := event.NewListener(event.ListenerParams{
Logger: c.log,
Subscriber: subs,
WorkerPoolCapacity: listenerPoolCap,
Logger: c.log,
Subscriber: subs,
})
fatalOnErr(err)

13
pkg/innerring/innerring.go
View file

@ -407,14 +407,6 @@ func New(ctx context.Context, log *logger.Logger, cfg *viper.Viper, errChan chan
}
func createListener(ctx context.Context, cli *client.Client, p *chainParams) (event.Listener, error) {
// listenerPoolCap is a capacity of a
// worker pool inside the listener. It
// is used to prevent blocking in neo-go:
// the client cannot make RPC requests if
// the notification channel is not being
// read by another goroutine.
const listenerPoolCap = 10
var (
sub subscriber.Subscriber
err error
@ -430,9 +422,8 @@ func createListener(ctx context.Context, cli *client.Client, p *chainParams) (ev
}
listener, err := event.NewListener(event.ListenerParams{
Logger: &logger.Logger{Logger: p.log.With(zap.String("chain", p.name))},
Subscriber: sub,
WorkerPoolCapacity: listenerPoolCap,
Logger: &logger.Logger{Logger: p.log.With(zap.String("chain", p.name))},
Subscriber: sub,
})
if err != nil {
return nil, err

14
pkg/morph/event/listener.go
View file

@ -609,11 +609,6 @@ func (l *listener) RegisterBlockHandler(handler BlockHandler) {
// NewListener create the notification event listener instance and returns Listener interface.
func NewListener(p ListenerParams) (Listener, error) {
// defaultPoolCap is a default worker
// pool capacity if it was not specified
// via params
const defaultPoolCap = 10
switch {
case p.Logger == nil:
return nil, fmt.Errorf("%s: %w", newListenerFailMsg, errNilLogger)
@ -621,12 +616,9 @@ func NewListener(p ListenerParams) (Listener, error) {
return nil, fmt.Errorf("%s: %w", newListenerFailMsg, errNilSubscriber)
}
poolCap := p.WorkerPoolCapacity
if poolCap == 0 {
poolCap = defaultPoolCap
}
pool, err := ants.NewPool(poolCap, ants.WithNonblocking(true))
// The pool here must be blocking, otherwise notifications could be dropped.
// The default capacity is 0, which means "infinite".
pool, err := ants.NewPool(p.WorkerPoolCapacity)
if err != nil {
return nil, fmt.Errorf("could not init worker pool: %w", err)
}