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.

Sign up for GitHub

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

maint: Add jitter to peer traffic, fix startup #1227

Merged
merged 1 commit into from
Jul 17, 2024
Merged

maint: Add jitter to peer traffic, fix startup #1227

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
10 changes: 7 additions & 3 deletions cmd/refinery/main.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,7 @@ import (
"github.com/facebookgo/startstop"
libhoney "github.com/honeycombio/libhoney-go"
"github.com/honeycombio/libhoney-go/transmission"
"github.com/jonboulle/clockwork"
"github.com/sirupsen/logrus"

"github.com/honeycombio/refinery/app"
Expand Down Expand Up @@ -121,8 +122,9 @@ func main() {
}
switch ptype {
case "file":
peers = &peer.FilePeers{Cfg: c}
// we know FilePeers doesn't need to be Started, so we can ask it how many peers we have.
// we want to use the file peers object to ask if we have only one peer, so we need to instantiate it
// with a dummy metrics object. we'll replace it with the real one later.
peers = &peer.FilePeers{Cfg: c, Metrics: &metrics.NullMetrics{}}
// if we only have one, we can use the local pubsub implementation.
peerList, err := peers.GetPeers()
if err != nil {
Expand All @@ -133,6 +135,8 @@ func main() {
} else {
pubsubber = &pubsub.GoRedisPubSub{}
}
// now erase the peers Metrics object so that it will get injected with the right one later
peers.(*peer.FilePeers).Metrics = nil
case "redis":
pubsubber = &pubsub.GoRedisPubSub{}
peers = &peer.RedisPubsubPeers{}
Expand Down Expand Up @@ -252,7 +256,7 @@ func main() {
{Value: promMetrics, Name: "promMetrics"},
{Value: oTelMetrics, Name: "otelMetrics"},
{Value: tracer, Name: "tracer"},

{Value: clockwork.NewRealClock()},
{Value: metricsSingleton, Name: "metrics"},
{Value: genericMetricsRecorder, Name: "genericMetrics"},
{Value: upstreamMetricsRecorder, Name: "upstreamMetrics"},
Expand Down
23 changes: 12 additions & 11 deletions internal/peer/pubsub_redis.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@ import (
"context"
"errors"
"fmt"
"math/rand"
"net"
"os"
"strings"
Expand All @@ -20,18 +21,14 @@ import (

const (
// refreshCacheInterval is how frequently this host will re-register itself
// with Redis. This should happen about 3x during each timeout phase in order
// to allow multiple timeouts to fail and yet still keep the host in the mix.
// Falling out of Redis will result in re-hashing the host-trace affinity and
// will cause broken traces for those that fall on both sides of the rehashing.
// This is why it's important to ensure hosts stay in the pool.
// by publishing their address. This should happen about 3x during each
// timeout phase in order to allow multiple timeouts to fail and yet still
// keep the host in the mix.
refreshCacheInterval = 3 * time.Second

// peerEntryTimeout is how long redis will wait before expiring a peer that
// doesn't check in. The ratio of refresh to peer timeout should be 1/3. Redis
// timeouts are in seconds and entries can last up to 2 seconds longer than
// their expected timeout (in my load testing), so the lower bound for this
// timer should be ... 5sec?
// peerEntryTimeout is how long we will wait before expiring a peer that
// doesn't check in. The ratio of refresh to peer timeout should be about
// 1/3; we overshoot because we add a jitter to the refresh interval.
peerEntryTimeout = 10 * time.Second
)

Expand Down Expand Up @@ -135,7 +132,11 @@ func (p *RedisPubsubPeers) Start() error {

// periodically refresh our presence in the list of peers, and update peers as they come in
go func() {
ticker := p.Clock.NewTicker(refreshCacheInterval)
// we want our refresh cache interval to vary from peer to peer so they
// don't always hit redis at the same time, so we add a random jitter of up
// to 20% of the interval
interval := refreshCacheInterval + time.Duration(rand.Int63n(int64(refreshCacheInterval/5)))
ticker := p.Clock.NewTicker(interval)
for {
select {
case <-p.done:
Expand Down
Loading