support gRPC proxy to remote backends

This commit is contained in:
Pavel Sviderski
2024-11-08 15:49:21 +10:00
parent b23699470b
commit 4493bb3c27
3 changed files with 366 additions and 23 deletions
+95 -7
View File
@@ -3,35 +3,123 @@ package proxy
import ( import (
"context" "context"
"github.com/siderolabs/grpc-proxy/proxy" "github.com/siderolabs/grpc-proxy/proxy"
"google.golang.org/grpc/codes"
"google.golang.org/grpc/metadata"
"google.golang.org/grpc/status"
"net"
"strconv"
"sync" "sync"
) )
// Director manages routing of gRPC requests between local and remote backends. // Director manages routing of gRPC requests between local and remote backends.
type Director struct { type Director struct {
localTarget string localBackend *LocalBackend
localBackend proxy.Backend remotePort int
remoteBackends sync.Map remoteBackends sync.Map
// mu synchronizes access to localAddress.
mu sync.RWMutex mu sync.RWMutex
localAddress string
} }
func NewDirector(localSockPath string) *Director { func NewDirector(localSockPath string, remotePort int) *Director {
return &Director{ return &Director{
localBackend: NewLocalBackend(localSockPath), localBackend: NewLocalBackend(localSockPath),
remotePort: remotePort,
} }
} }
// UpdateLocalAddress updates the local machine address used to identify which requests should be proxied // UpdateLocalAddress updates the local machine address used to identify which requests should be proxied
// to the local gRPC server. // to the local gRPC server.
func (d *Director) UpdateLocalAddress(target string) { func (d *Director) UpdateLocalAddress(addr string) {
d.mu.Lock() d.mu.Lock()
defer d.mu.Unlock() defer d.mu.Unlock()
d.localTarget = target d.localAddress = addr
} }
// Director implements proxy.StreamDirector for grpc-proxy, routing requests to local or remote backends based // Director implements proxy.StreamDirector for grpc-proxy, routing requests to local or remote backends based
// on gRPC metadata in the context. // on gRPC metadata in the context. Each machine metadata is injected into the response messages by the proxy
// if the request is proxied to multiple backends.
func (d *Director) Director(ctx context.Context, fullMethodName string) (proxy.Mode, []proxy.Backend, error) { func (d *Director) Director(ctx context.Context, fullMethodName string) (proxy.Mode, []proxy.Backend, error) {
md, ok := metadata.FromIncomingContext(ctx)
if !ok {
return proxy.One2One, []proxy.Backend{d.localBackend}, nil return proxy.One2One, []proxy.Backend{d.localBackend}, nil
}
// If the request is already proxied, send it to the local backend.
if _, ok = md["proxy-authority"]; ok {
return proxy.One2One, []proxy.Backend{d.localBackend}, nil
}
// If the request metadata doesn't contain machines to proxy to, send it to the local backend.
machines, ok := md["machines"]
if !ok {
return proxy.One2One, []proxy.Backend{d.localBackend}, nil
}
if len(machines) == 0 {
return proxy.One2One, nil, status.Error(codes.InvalidArgument, "no machines specified")
}
d.mu.RLock()
localAddress := d.localAddress
d.mu.RUnlock()
backends := make([]proxy.Backend, len(machines))
for i, addr := range machines {
if addr == localAddress {
backends[i] = d.localBackend
continue
}
target := net.JoinHostPort(addr, strconv.Itoa(d.remotePort))
backend, err := d.remoteBackend(target)
if err != nil {
return proxy.One2One, nil, status.Error(codes.Internal, err.Error())
}
backends[i] = backend
}
if len(backends) == 1 {
return proxy.One2One, backends, nil
}
return proxy.One2Many, backends, nil
}
// remoteBackend returns a RemoteBackend for the given target from the cache or creates a new one.
func (d *Director) remoteBackend(target string) (*RemoteBackend, error) {
b, ok := d.remoteBackends.Load(target)
if ok {
return b.(*RemoteBackend), nil
}
backend, err := NewRemoteBackend(target)
if err != nil {
return nil, err
}
existing, loaded := d.remoteBackends.LoadOrStore(target, backend)
if loaded {
// A concurrent remoteBackend call built a different backend.
backend.Close()
return existing.(*RemoteBackend), nil
}
return backend, nil
}
// FlushRemoteBackends closes all remote backend connections and removes them from the cache.
func (d *Director) FlushRemoteBackends() {
d.remoteBackends.Range(func(key, value interface{}) bool {
backend, ok := value.(*RemoteBackend)
if !ok {
return true
}
backend.Close()
d.remoteBackends.Delete(key)
return true
})
}
// Close closes all backend connections.
func (d *Director) Close() {
d.localBackend.Close()
d.FlushRemoteBackends()
} }
+28 -14
View File
@@ -2,6 +2,7 @@ package proxy
import ( import (
"context" "context"
"fmt"
"github.com/siderolabs/grpc-proxy/proxy" "github.com/siderolabs/grpc-proxy/proxy"
"google.golang.org/grpc" "google.golang.org/grpc"
"google.golang.org/grpc/credentials/insecure" "google.golang.org/grpc/credentials/insecure"
@@ -26,43 +27,56 @@ func NewLocalBackend(sockPath string) *LocalBackend {
} }
} }
func (l *LocalBackend) String() string { func (b *LocalBackend) String() string {
return "local" return "local"
} }
// GetConnection returns a gRPC connection to the local server listening on the Unix socket. // GetConnection returns a gRPC connection to the local server listening on the Unix socket.
func (l *LocalBackend) GetConnection(ctx context.Context, _ string) (context.Context, *grpc.ClientConn, error) { func (b *LocalBackend) GetConnection(ctx context.Context, _ string) (context.Context, *grpc.ClientConn, error) {
md, _ := metadata.FromIncomingContext(ctx) md, _ := metadata.FromIncomingContext(ctx)
// TODO: delete
fmt.Printf("### local backend metadata: %+v\n", md)
outCtx := metadata.NewOutgoingContext(ctx, md) outCtx := metadata.NewOutgoingContext(ctx, md)
l.mu.RLock() b.mu.RLock()
if l.conn != nil { if b.conn != nil {
l.mu.RUnlock() defer b.mu.RUnlock()
return outCtx, l.conn, nil return outCtx, b.conn, nil
} }
l.mu.RUnlock() b.mu.RUnlock()
l.mu.Lock() b.mu.Lock()
defer l.mu.Unlock() defer b.mu.Unlock()
var err error var err error
l.conn, err = grpc.NewClient( b.conn, err = grpc.NewClient(
"unix://"+l.sockPath, "unix://"+b.sockPath,
grpc.WithTransportCredentials(insecure.NewCredentials()), grpc.WithTransportCredentials(insecure.NewCredentials()),
grpc.WithDefaultCallOptions( grpc.WithDefaultCallOptions(
grpc.ForceCodecV2(proxy.Codec()), grpc.ForceCodecV2(proxy.Codec()),
), ),
) )
return outCtx, l.conn, err return outCtx, b.conn, err
} }
// AppendInfo is called to enhance response from the backend with additional data. // AppendInfo is called to enhance response from the backend with additional data.
func (l *LocalBackend) AppendInfo(_ bool, resp []byte) ([]byte, error) { func (b *LocalBackend) AppendInfo(_ bool, resp []byte) ([]byte, error) {
return resp, nil return resp, nil
} }
// BuildError is called to convert error from upstream into response field. // BuildError is called to convert error from upstream into response field.
func (l *LocalBackend) BuildError(bool, error) ([]byte, error) { func (b *LocalBackend) BuildError(bool, error) ([]byte, error) {
return nil, nil return nil, nil
} }
// Close closes the upstream gRPC connection.
func (b *LocalBackend) Close() {
b.mu.Lock()
defer b.mu.Unlock()
if b.conn != nil {
b.conn.Close()
b.conn = nil
}
}
+241
View File
@@ -0,0 +1,241 @@
package proxy
import (
"context"
"fmt"
"github.com/siderolabs/grpc-proxy/proxy"
"google.golang.org/grpc"
"google.golang.org/grpc/backoff"
"google.golang.org/grpc/credentials/insecure"
"google.golang.org/grpc/metadata"
"google.golang.org/grpc/status"
"google.golang.org/protobuf/encoding/protowire"
"google.golang.org/protobuf/proto"
"net"
"net/netip"
"sync"
"time"
"uncloud/internal/machine/api/pb"
)
// RemoteBackend is a proxy.Backend implementation that proxies to a remote gRPC server, injecting machine metadata
// into the response.
//
// Based on the Talos apid implementation:
// https://github.com/siderolabs/talos/blob/59a78da42cdea8fbccc35d0851f9b0eef928261b/internal/app/apid/pkg/backend/apid.go
type RemoteBackend struct {
target string
mu sync.RWMutex
conn *grpc.ClientConn
}
var _ proxy.Backend = (*RemoteBackend)(nil)
// NewRemoteBackend creates a new instance of RemoteBackend for the given target which must have the format [IPv6]:port.
func NewRemoteBackend(target string) (*RemoteBackend, error) {
host, _, err := net.SplitHostPort(target)
if err != nil {
return nil, fmt.Errorf("target must have the format [IPv6]:port: %s", target)
}
addr, err := netip.ParseAddr(host)
if err != nil || !addr.Is6() {
return nil, fmt.Errorf("target host must be a valid IPv6 address: %s", host)
}
return &RemoteBackend{target: target}, nil
}
func (b *RemoteBackend) String() string {
return b.target
}
// GetConnection returns a gRPC connection to the remote server.
func (b *RemoteBackend) GetConnection(ctx context.Context, _ string) (context.Context, *grpc.ClientConn, error) {
md, _ := metadata.FromIncomingContext(ctx)
if authority := md[":authority"]; len(authority) > 0 {
md.Set("proxy-authority", authority...)
} else {
md.Set("proxy-authority", "unknown")
}
delete(md, ":authority")
delete(md, "machines")
outCtx := metadata.NewOutgoingContext(ctx, md)
b.mu.RLock()
if b.conn != nil {
defer b.mu.RUnlock()
return outCtx, b.conn, nil
}
b.mu.RUnlock()
b.mu.Lock()
defer b.mu.Unlock()
// Override the max delay to avoid excessive backoff when the another node is unavailable, e.g. rebooted
// or the WireGuard connection is temporary down.
//
// Default max delay is 2 minutes, which is too long for our use case.
backoffConfig := backoff.DefaultConfig
// The maximum wait time between attempts.
backoffConfig.MaxDelay = 15 * time.Second
var err error
b.conn, err = grpc.NewClient(
b.target,
grpc.WithTransportCredentials(insecure.NewCredentials()),
grpc.WithConnectParams(grpc.ConnectParams{
Backoff: backoffConfig,
// Not published as a constant in gRPC library.
// See: https://github.com/grpc/grpc-go/blob/d5dee5fdbdeb52f6ea10b37b2cc7ce37814642d7/clientconn.go#L55-L56
// Each connection attempt can take up to MinConnectTimeout.
MinConnectTimeout: 20 * time.Second,
}),
grpc.WithDefaultCallOptions(
grpc.ForceCodecV2(proxy.Codec()),
),
)
return outCtx, b.conn, err
}
// AppendInfo is called to enhance response from the backend with additional data.
//
// AppendInfo enhances upstream response with machine metadata (target).
//
// This method depends on grpc protobuf response structure, each response should
// look like:
//
// message SomeResponse {
// repeated SomeReply messages = 1; // please note field ID == 1
// }
//
// message SomeReply {
// common.Metadata metadata = 1;
// <other fields go here ...>
// }
//
// As 'SomeReply' is repeated in 'SomeResponse', if we concatenate protobuf representation
// of several 'SomeResponse' messages, we still get valid 'SomeResponse' representation but with more
// entries (feature of protobuf binary representation).
//
// If we look at binary representation of any unary 'SomeResponse' message, it will always contain one
// protobuf field with field ID 1 (see above) and type 2 (embedded message SomeReply is encoded
// as string with length). So if we want to add fields to 'SomeReply', we can simply read field
// header, adjust length for new 'SomeReply' representation, and prepend new field header.
//
// At the same time, we can add 'common.Metadata' structure to 'SomeReply' by simply
// appending or prepending 'common.Metadata' as a single field. This requires 'metadata'
// field to be not defined in original response. (This is due to the fact that protobuf message
// representation is concatenation of each field representation).
//
// To build only single field (Metadata) we use helper message which contains exactly this
// field with same field ID as in every other 'SomeReply':
//
// message Empty {
// common.Metadata metadata = 1;
// }
//
// As streaming replies are not wrapped into 'SomeResponse' with 'repeated', handling is simpler: we just
// need to append Empty with details.
//
// So AppendInfo does the following: validates that response contains field ID 1 encoded as string,
// cuts field header, rest is representation of some reply. Marshal 'Empty' as protobuf,
// which builds 'common.Metadata' field, append it to original response message, build new header
// for new length of some response, and add back new field header.
func (b *RemoteBackend) AppendInfo(streaming bool, resp []byte) ([]byte, error) {
payload, err := proto.Marshal(&pb.Empty{
Metadata: &pb.Metadata{
Machine: b.target,
},
})
if streaming {
return append(resp, payload...), err
}
const (
metadataField = 1 // field number in proto definition for repeated response
metadataType = 2 // "string" for embedded messages
)
// decode protobuf embedded header
typ, n1 := protowire.ConsumeVarint(resp)
if n1 < 0 {
return nil, protowire.ParseError(n1)
}
_, n2 := protowire.ConsumeVarint(resp[n1:]) // length
if n2 < 0 {
return nil, protowire.ParseError(n2)
}
if typ != (metadataField<<3)|metadataType {
return nil, fmt.Errorf("unexpected message format: %d", typ)
}
if n1+n2 > len(resp) {
return nil, fmt.Errorf("unexpected message size: %d", len(resp))
}
// cut off embedded message header
resp = resp[n1+n2:]
// build new embedded message header
prefix := protowire.AppendVarint(
protowire.AppendVarint(nil, (metadataField<<3)|metadataType),
uint64(len(resp)+len(payload)),
)
resp = append(prefix, resp...)
return append(resp, payload...), err
}
// BuildError converts upstream error into message from upstream, so that multiple
// successful and failure responses might be returned.
//
// This simply relies on the fact that any response contains 'Empty' message.
// So if 'Empty' is unmarshalled into any other reply message, all the fields
// are undefined but 'Metadata':
//
// message Empty {
// common.Metadata metadata = 1;
// }
//
// message EmptyResponse {
// repeated Empty messages = 1;
// }
//
// Streaming responses are not wrapped into Empty, so we simply marshall EmptyResponse
// message.
func (b *RemoteBackend) BuildError(streaming bool, err error) ([]byte, error) {
var resp proto.Message = &pb.Empty{
Metadata: &pb.Metadata{
Machine: b.target,
Error: err.Error(),
Status: status.Convert(err).Proto(),
},
}
if !streaming {
resp = &pb.EmptyResponse{
Messages: []*pb.Empty{
resp.(*pb.Empty),
},
}
}
return proto.Marshal(resp)
}
// Close closes the upstream gRPC connection.
func (b *RemoteBackend) Close() {
b.mu.Lock()
defer b.mu.Unlock()
if b.conn != nil {
b.conn.Close()
b.conn = nil
}
}