Files
coredns/plugin/pkg/proxy/persistent.go
Ville Vesilehto f3983c1111 perf(proxy): use mutex-based connection pool (#7790)
* perf(proxy): use mutex-based connection pool

The proxy package (used for example by the forward plugin) utilized
an actor model where a single connManager goroutine managed
connection pooling via unbuffered channels (dial, yield, ret). This
design serialized all connection acquisition and release operations
through a single goroutine, creating a bottleneck under high
concurrency. This was observable as a performance degradation when
using a single upstream backend compared to multiple backends
(which sharded the bottleneck).

Changes:
- Removed dial, yield, and ret channels from the Transport struct.
- Removed the connManager goroutine's request processing loop.
- Implemented Dial() and Yield() using a sync.Mutex to protect the
  connection slice, allowing for fast concurrent access without
  context switching.
- Downgraded connManager to a simple background cleanup loop that
  only handles connection expiration on a ticker.
- Updated plugin/pkg/proxy/connect.go to use direct method calls
  instead of channel sends.
- Updated tests to reflect the removal of internal channels.

Benchmarks show that this change eliminates the single-backend
bottleneck. Now a single upstream backend performs on par with
multiple backends, and overall throughput is improved.

The implementation aligns with standard Go patterns for connection
pooling (e.g., net/http.Transport).

Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>

* fix: address PR review for persistent.go

- Named mutex field instead of embedding, to not expose
  Lock() and Unlock()
- Move stop check outside of lock in Yield()
- Close() without a separate goroutine
- Change stop channel to struct

Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>

* fix: address code review feedback for conn pool

- Switch from LIFO to FIFO connection selection for source port
  diversity, reducing DNS cache poisoning risk (RFC 5452).
- Remove "clear entire cache" optimization as it was LIFO-specific.
  FIFO naturally iterates and skips expired connections.
- Remove all goroutines for closing connections; collect connections
  while holding lock, close synchronously after releasing lock.

Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>

* fix: remove unused error consts

No longer utilised after refactoring the channel based approach.

Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>

* feat(forward): add max_idle_conns option

Add configurable connection pool limit for the forward plugin via
the max_idle_conns Corefile option.

Changes:
- Add SetMaxIdleConns to proxy
- Add maxIdleConns field to Forward struct
- Add max_idle_conns parsing in forward plugin setup
- Apply setting to each proxy during configuration
- Update forward plugin README with new option

By default the value is 0 (unbounded). When set, excess
connections returned to the pool are closed immediately
rather than cached.

Also add a yield related test.

Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>

* chore(proxy): simple Dial by closing conns inline

Remove toClose slice collection to reduce complexity. Instead close
expired connections directly while iterating. Reduces complexity with
negligible lock-time impact.

Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>

* chore: fewer explicit Unlock calls

Cleaner and less chance of forgetting to unlock on new possible
code paths.

Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>

---------

Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
2026-01-13 17:49:46 -08:00

148 lines
3.5 KiB
Go

package proxy
import (
"crypto/tls"
"sort"
"sync"
"time"
"github.com/miekg/dns"
)
// a persistConn hold the dns.Conn and the last used time.
type persistConn struct {
c *dns.Conn
used time.Time
}
// Transport hold the persistent cache.
type Transport struct {
avgDialTime int64 // kind of average time of dial time
conns [typeTotalCount][]*persistConn // Buckets for udp, tcp and tcp-tls.
expire time.Duration // After this duration a connection is expired.
maxIdleConns int // Max idle connections per transport type; 0 means unlimited.
addr string
tlsConfig *tls.Config
proxyName string
mu sync.Mutex
stop chan struct{}
}
func newTransport(proxyName, addr string) *Transport {
t := &Transport{
avgDialTime: int64(maxDialTimeout / 2),
conns: [typeTotalCount][]*persistConn{},
expire: defaultExpire,
addr: addr,
stop: make(chan struct{}),
proxyName: proxyName,
}
return t
}
// connManager manages the persistent connection cache for UDP and TCP.
func (t *Transport) connManager() {
ticker := time.NewTicker(defaultExpire)
defer ticker.Stop()
for {
select {
case <-ticker.C:
t.cleanup(false)
case <-t.stop:
t.cleanup(true)
return
}
}
}
// closeConns closes connections.
func closeConns(conns []*persistConn) {
for _, pc := range conns {
pc.c.Close()
}
}
// cleanup removes connections from cache.
func (t *Transport) cleanup(all bool) {
var toClose []*persistConn
t.mu.Lock()
staleTime := time.Now().Add(-t.expire)
for transtype, stack := range t.conns {
if len(stack) == 0 {
continue
}
if all {
t.conns[transtype] = nil
toClose = append(toClose, stack...)
continue
}
if stack[0].used.After(staleTime) {
continue
}
// connections in stack are sorted by "used"
good := sort.Search(len(stack), func(i int) bool {
return stack[i].used.After(staleTime)
})
t.conns[transtype] = stack[good:]
toClose = append(toClose, stack[:good]...)
}
t.mu.Unlock()
// Close connections after releasing lock
closeConns(toClose)
}
// Yield returns the connection to transport for reuse.
func (t *Transport) Yield(pc *persistConn) {
// Check if transport is stopped before acquiring lock
select {
case <-t.stop:
// If stopped, don't return to pool, just close
pc.c.Close()
return
default:
}
pc.used = time.Now() // update used time
t.mu.Lock()
defer t.mu.Unlock()
transtype := t.transportTypeFromConn(pc)
if t.maxIdleConns > 0 && len(t.conns[transtype]) >= t.maxIdleConns {
pc.c.Close()
return
}
t.conns[transtype] = append(t.conns[transtype], pc)
}
// Start starts the transport's connection manager.
func (t *Transport) Start() { go t.connManager() }
// Stop stops the transport's connection manager.
func (t *Transport) Stop() { close(t.stop) }
// SetExpire sets the connection expire time in transport.
func (t *Transport) SetExpire(expire time.Duration) { t.expire = expire }
// SetMaxIdleConns sets the maximum idle connections per transport type.
// A value of 0 means unlimited (default).
func (t *Transport) SetMaxIdleConns(n int) { t.maxIdleConns = n }
// SetTLSConfig sets the TLS config in transport.
func (t *Transport) SetTLSConfig(cfg *tls.Config) { t.tlsConfig = cfg }
// GetTLSConfig returns the TLS config in transport.
func (t *Transport) GetTLSConfig() *tls.Config { return t.tlsConfig }
const (
defaultExpire = 10 * time.Second
minDialTimeout = 1 * time.Second
maxDialTimeout = 30 * time.Second
)