plugeth/p2p/tracker/tracker.go
Péter Szilágyi 1fb9a6dd32
eth/protocols, prp/tracker: add support for req/rep rtt tracking (#22608)
* eth/protocols, prp/tracker: add support for req/rep rtt tracking

* p2p/tracker: sanity cap the number of pending requests

* pap/tracker: linter <3

* p2p/tracker: disable entire tracker if no metrics are enabled
2021-04-22 11:42:46 +03:00

204 lines
6.8 KiB
Go

// Copyright 2021 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum library is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// The go-ethereum library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
package tracker
import (
"container/list"
"fmt"
"sync"
"time"
"github.com/ethereum/go-ethereum/log"
"github.com/ethereum/go-ethereum/metrics"
)
const (
// trackedGaugeName is the prefix of the per-packet request tracking.
trackedGaugeName = "p2p/tracked"
// lostMeterName is the prefix of the per-packet request expirations.
lostMeterName = "p2p/lost"
// staleMeterName is the prefix of the per-packet stale responses.
staleMeterName = "p2p/stale"
// waitHistName is the prefix of the per-packet (req only) waiting time histograms.
waitHistName = "p2p/wait"
// maxTrackedPackets is a huge number to act as a failsafe on the number of
// pending requests the node will track. It should never be hit unless an
// attacker figures out a way to spin requests.
maxTrackedPackets = 100000
)
// request tracks sent network requests which have not yet received a response.
type request struct {
peer string
version uint // Protocol version
reqCode uint64 // Protocol message code of the request
resCode uint64 // Protocol message code of the expected response
time time.Time // Timestamp when the request was made
expire *list.Element // Expiration marker to untrack it
}
// Tracker is a pending network request tracker to measure how much time it takes
// a remote peer to respond.
type Tracker struct {
protocol string // Protocol capability identifier for the metrics
timeout time.Duration // Global timeout after which to drop a tracked packet
pending map[uint64]*request // Currently pending requests
expire *list.List // Linked list tracking the expiration order
wake *time.Timer // Timer tracking the expiration of the next item
lock sync.Mutex // Lock protecting from concurrent updates
}
// New creates a new network request tracker to monitor how much time it takes to
// fill certain requests and how individual peers perform.
func New(protocol string, timeout time.Duration) *Tracker {
return &Tracker{
protocol: protocol,
timeout: timeout,
pending: make(map[uint64]*request),
expire: list.New(),
}
}
// Track adds a network request to the tracker to wait for a response to arrive
// or until the request it cancelled or times out.
func (t *Tracker) Track(peer string, version uint, reqCode uint64, resCode uint64, id uint64) {
if !metrics.Enabled {
return
}
t.lock.Lock()
defer t.lock.Unlock()
// If there's a duplicate request, we've just random-collided (or more probably,
// we have a bug), report it. We could also add a metric, but we're not really
// expecting ourselves to be buggy, so a noisy warning should be enough.
if _, ok := t.pending[id]; ok {
log.Error("Network request id collision", "protocol", t.protocol, "version", version, "code", reqCode, "id", id)
return
}
// If we have too many pending requests, bail out instead of leaking memory
if pending := len(t.pending); pending >= maxTrackedPackets {
log.Error("Request tracker exceeded allowance", "pending", pending, "peer", peer, "protocol", t.protocol, "version", version, "code", reqCode)
return
}
// Id doesn't exist yet, start tracking it
t.pending[id] = &request{
peer: peer,
version: version,
reqCode: reqCode,
resCode: resCode,
time: time.Now(),
expire: t.expire.PushBack(id),
}
g := fmt.Sprintf("%s/%s/%d/%#02x", trackedGaugeName, t.protocol, version, reqCode)
metrics.GetOrRegisterGauge(g, nil).Inc(1)
// If we've just inserted the first item, start the expiration timer
if t.wake == nil {
t.wake = time.AfterFunc(t.timeout, t.clean)
}
}
// clean is called automatically when a preset time passes without a response
// being dleivered for the first network request.
func (t *Tracker) clean() {
t.lock.Lock()
defer t.lock.Unlock()
// Expire anything within a certain threshold (might be no items at all if
// we raced with the delivery)
for t.expire.Len() > 0 {
// Stop iterating if the next pending request is still alive
var (
head = t.expire.Front()
id = head.Value.(uint64)
req = t.pending[id]
)
if time.Since(req.time) < t.timeout+5*time.Millisecond {
break
}
// Nope, dead, drop it
t.expire.Remove(head)
delete(t.pending, id)
g := fmt.Sprintf("%s/%s/%d/%#02x", trackedGaugeName, t.protocol, req.version, req.reqCode)
metrics.GetOrRegisterGauge(g, nil).Dec(1)
m := fmt.Sprintf("%s/%s/%d/%#02x", lostMeterName, t.protocol, req.version, req.reqCode)
metrics.GetOrRegisterMeter(m, nil).Mark(1)
}
t.schedule()
}
// schedule starts a timer to trigger on the expiration of the first network
// packet.
func (t *Tracker) schedule() {
if t.expire.Len() == 0 {
t.wake = nil
return
}
t.wake = time.AfterFunc(time.Until(t.pending[t.expire.Front().Value.(uint64)].time.Add(t.timeout)), t.clean)
}
// Fulfil fills a pending request, if any is available, reporting on various metrics.
func (t *Tracker) Fulfil(peer string, version uint, code uint64, id uint64) {
if !metrics.Enabled {
return
}
t.lock.Lock()
defer t.lock.Unlock()
// If it's a non existing request, track as stale response
req, ok := t.pending[id]
if !ok {
m := fmt.Sprintf("%s/%s/%d/%#02x", staleMeterName, t.protocol, version, code)
metrics.GetOrRegisterMeter(m, nil).Mark(1)
return
}
// If the response is funky, it might be some active attack
if req.peer != peer || req.version != version || req.resCode != code {
log.Warn("Network response id collision",
"have", fmt.Sprintf("%s:%s/%d:%d", peer, t.protocol, version, code),
"want", fmt.Sprintf("%s:%s/%d:%d", peer, t.protocol, req.version, req.resCode),
)
return
}
// Everything matches, mark the request serviced and meter it
t.expire.Remove(req.expire)
if req.expire.Prev() == nil {
t.wake.Stop()
t.schedule()
}
g := fmt.Sprintf("%s/%s/%d/%#02x", trackedGaugeName, t.protocol, req.version, req.reqCode)
metrics.GetOrRegisterGauge(g, nil).Dec(1)
h := fmt.Sprintf("%s/%s/%d/%#02x", waitHistName, t.protocol, req.version, req.reqCode)
sampler := func() metrics.Sample {
return metrics.ResettingSample(
metrics.NewExpDecaySample(1028, 0.015),
)
}
metrics.GetOrRegisterHistogramLazy(h, nil, sampler).Update(time.Since(req.time).Microseconds())
}