alertmanager/cluster/channel.go
stuart nelson 445fbdf1a8
gossip large messages via SendReliable (#1415)
* Gossip large messages via SendReliable

For messages beyond half of the maximum gossip
packet size, send the message to all peer nodes
via TCP.

The choice of "larger than half the max gossip
size" is relatively arbitrary. From brief testing,
the overhead from memberlist on a packet seemed to
only use ~3 of the available 1400 bytes, and most
gossip messages seem to be <<500 bytes.

* Add tests for oversized/normal message gossiping

* Make oversize metric names consistent

* Remove errant printf in test

* Correctly increment WaitGroup

* Add comment for OversizedMessage func

* Add metric for oversized messages dropped

Code was added to drop oversized messages if the
buffered channel they are sent on is full. This
is a good thing to surface as a metric.

* Add counter for total oversized messages sent

* Change full queue log level to debug

Was previously a warning, which isn't necessary
now that there is a metric tracking it.

Signed-off-by: stuart nelson <stuartnelson3@gmail.com>
2018-06-15 13:40:21 +02:00

150 lines
4.8 KiB
Go

// Copyright 2018 Prometheus Team
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package cluster
import (
"sync"
"time"
"github.com/go-kit/kit/log"
"github.com/go-kit/kit/log/level"
"github.com/gogo/protobuf/proto"
"github.com/hashicorp/memberlist"
"github.com/prometheus/alertmanager/cluster/clusterpb"
"github.com/prometheus/client_golang/prometheus"
)
// Channel allows clients to send messages for a specific state type that will be
// broadcasted in a best-effort manner.
type Channel struct {
key string
send func([]byte)
peers func() []*memberlist.Node
sendOversize func(*memberlist.Node, []byte) error
msgc chan []byte
logger log.Logger
oversizeGossipMessageFailureTotal prometheus.Counter
oversizeGossipMessageDroppedTotal prometheus.Counter
oversizeGossipMessageSentTotal prometheus.Counter
oversizeGossipDuration prometheus.Histogram
}
// NewChannel creates a new Channel struct, which handles sending normal and
// oversize messages to peers.
func NewChannel(
key string,
send func([]byte),
peers func() []*memberlist.Node,
sendOversize func(*memberlist.Node, []byte) error,
logger log.Logger,
stopc chan struct{},
reg prometheus.Registerer,
) *Channel {
oversizeGossipMessageFailureTotal := prometheus.NewCounter(prometheus.CounterOpts{
Name: "alertmanager_oversized_gossip_message_failure_total",
Help: "Number of oversized gossip message sends that failed.",
ConstLabels: prometheus.Labels{"key": key},
})
oversizeGossipMessageSentTotal := prometheus.NewCounter(prometheus.CounterOpts{
Name: "alertmanager_oversized_gossip_message_sent_total",
Help: "Number of oversized gossip message sent.",
ConstLabels: prometheus.Labels{"key": key},
})
oversizeGossipMessageDroppedTotal := prometheus.NewCounter(prometheus.CounterOpts{
Name: "alertmanager_oversized_gossip_message_dropped_total",
Help: "Number of oversized gossip messages that were dropped due to a full message queue.",
ConstLabels: prometheus.Labels{"key": key},
})
oversizeGossipDuration := prometheus.NewHistogram(prometheus.HistogramOpts{
Name: "alertmanager_oversize_gossip_message_duration_seconds",
Help: "Duration of oversized gossip message requests.",
ConstLabels: prometheus.Labels{"key": key},
})
reg.MustRegister(oversizeGossipDuration, oversizeGossipMessageFailureTotal, oversizeGossipMessageDroppedTotal, oversizeGossipMessageSentTotal)
c := &Channel{
key: key,
send: send,
peers: peers,
logger: logger,
msgc: make(chan []byte, 200),
sendOversize: sendOversize,
oversizeGossipMessageFailureTotal: oversizeGossipMessageFailureTotal,
oversizeGossipMessageDroppedTotal: oversizeGossipMessageDroppedTotal,
oversizeGossipMessageSentTotal: oversizeGossipMessageSentTotal,
oversizeGossipDuration: oversizeGossipDuration,
}
go c.handleOverSizedMessages(stopc)
return c
}
// handleOverSizedMessages prevents memberlist from opening too many parallel
// TCP connections to its peers.
func (c *Channel) handleOverSizedMessages(stopc chan struct{}) {
var wg sync.WaitGroup
for {
select {
case b := <-c.msgc:
for _, n := range c.peers() {
wg.Add(1)
go func(n *memberlist.Node) {
defer wg.Done()
c.oversizeGossipMessageSentTotal.Inc()
start := time.Now()
if err := c.sendOversize(n, b); err != nil {
level.Debug(c.logger).Log("msg", "failed to send reliable", "key", c.key, "node", n, "err", err)
c.oversizeGossipMessageFailureTotal.Inc()
return
}
c.oversizeGossipDuration.Observe(time.Since(start).Seconds())
}(n)
}
wg.Wait()
case <-stopc:
return
}
}
}
// Broadcast enqueues a message for broadcasting.
func (c *Channel) Broadcast(b []byte) {
b, err := proto.Marshal(&clusterpb.Part{Key: c.key, Data: b})
if err != nil {
return
}
if OversizedMessage(b) {
select {
case c.msgc <- b:
default:
level.Debug(c.logger).Log("msg", "oversized gossip channel full")
c.oversizeGossipMessageDroppedTotal.Inc()
}
} else {
c.send(b)
}
}
// OversizedMessage indicates whether or not the byte payload should be sent
// via TCP.
func OversizedMessage(b []byte) bool {
return len(b) > maxGossipPacketSize/2
}