[#307] add package trace, integrate it with logging, and adopt it throughout zrepl

package trace:

- introduce the concept of tasks and spans, tracked as linked list within ctx
    - see package-level docs for an overview of the concepts
    - **main feature 1**: unique stack of task and span IDs
        - makes it easy to follow a series of log entries in concurrent code
    - **main feature 2**: ability to produce a chrome://tracing-compatible trace file
        - either via an env variable or a `zrepl pprof` subcommand
        - this is not a CPU profile, we already have go pprof for that
        - but it is very useful to visually inspect where the
          replication / snapshotter / pruner spends its time
          ( fixes #307 )

usage in package daemon/logging:

- goal: every log entry should have a trace field with the ID stack from package trace

- make `logging.GetLogger(ctx, Subsys)` the authoritative `logger.Logger` factory function
    - the context carries a linked list of injected fields which
      `logging.GetLogger` adds to the logger it returns
    - `logging.GetLogger` also uses package `trace` to get the
      task-and-span-stack and injects it into the returned logger's fields
This commit is contained in:
Christian Schwarz
2020-04-11 15:49:41 +02:00
parent bcb5965617
commit 10a14a8c50
75 changed files with 1934 additions and 462 deletions
-1
View File
@@ -143,7 +143,6 @@ func (c *Client) ReqSend(ctx context.Context, req *pdu.SendReq) (*pdu.SendRes, i
}
func (c *Client) ReqRecv(ctx context.Context, req *pdu.ReceiveReq, stream io.ReadCloser) (*pdu.ReceiveRes, error) {
defer c.log.Debug("ReqRecv returns")
conn, err := c.getWire(ctx)
if err != nil {
+37 -1
View File
@@ -33,16 +33,33 @@ type Handler interface {
type Logger = logger.Logger
type ContextInterceptorData interface {
FullMethod() string
ClientIdentity() string
}
type ContextInterceptor = func(ctx context.Context, data ContextInterceptorData, handler func(ctx context.Context))
type Server struct {
h Handler
wi WireInterceptor
ci ContextInterceptor
log Logger
}
func NewServer(wi WireInterceptor, logger Logger, handler Handler) *Server {
var noopContextInteceptor = func(ctx context.Context, _ ContextInterceptorData, handler func(context.Context)) {
handler(ctx)
}
// wi and ci may be nil
func NewServer(wi WireInterceptor, ci ContextInterceptor, logger Logger, handler Handler) *Server {
if ci == nil {
ci = noopContextInteceptor
}
return &Server{
h: handler,
wi: wi,
ci: ci,
log: logger,
}
}
@@ -93,6 +110,14 @@ func (s *Server) Serve(ctx context.Context, l transport.AuthenticatedListener) {
}
}
type contextInterceptorData struct {
fullMethod string
clientIdentity string
}
func (d contextInterceptorData) FullMethod() string { return d.fullMethod }
func (d contextInterceptorData) ClientIdentity() string { return d.clientIdentity }
func (s *Server) serveConn(nc *transport.AuthConn) {
s.log.Debug("serveConn begin")
defer s.log.Debug("serveConn done")
@@ -117,6 +142,17 @@ func (s *Server) serveConn(nc *transport.AuthConn) {
}
endpoint := string(header)
data := contextInterceptorData{
fullMethod: endpoint,
clientIdentity: nc.ClientIdentity(),
}
s.ci(ctx, data, func(ctx context.Context) {
s.serveConnRequest(ctx, endpoint, c)
})
}
func (s *Server) serveConnRequest(ctx context.Context, endpoint string, c *stream.Conn) {
reqStructured, err := c.ReadStreamedMessage(ctx, RequestStructuredMaxSize, ReqStructured)
if err != nil {
s.log.WithError(err).Error("error reading structured part")
@@ -112,7 +112,7 @@ func server() {
orDie(err)
l := tcpListener{nl.(*net.TCPListener), "fakeclientidentity"}
srv := dataconn.NewServer(nil, logger.NewStderrDebugLogger(), devNullHandler{})
srv := dataconn.NewServer(nil, nil, logger.NewStderrDebugLogger(), devNullHandler{})
ctx := context.Background()
+5
View File
@@ -7,6 +7,7 @@ import (
"io"
"net"
"strings"
"sync"
"sync/atomic"
"unicode/utf8"
@@ -80,9 +81,13 @@ func doWriteStream(ctx context.Context, c *heartbeatconn.Conn, stream io.Reader,
err error
}
var wg sync.WaitGroup
defer wg.Wait()
reads := make(chan read, 5)
var stopReading uint32
wg.Add(1)
go func() {
defer wg.Done()
defer close(reads)
for atomic.LoadUint32(&stopReading) == 0 {
buffer := bufpool.Get(1 << FramePayloadShift)
@@ -99,10 +99,23 @@ func (*transportCredentials) OverrideServerName(string) error {
panic("not implemented")
}
type ContextInterceptor = func(ctx context.Context) context.Context
type ContextInterceptorData interface {
FullMethod() string
ClientIdentity() string
}
func NewInterceptors(logger Logger, clientIdentityKey interface{}, ctxInterceptor ContextInterceptor) (unary grpc.UnaryServerInterceptor, stream grpc.StreamServerInterceptor) {
unary = func(ctx context.Context, req interface{}, info *grpc.UnaryServerInfo, handler grpc.UnaryHandler) (resp interface{}, err error) {
type contextInterceptorData struct {
fullMethod string
clientIdentity string
}
func (d contextInterceptorData) FullMethod() string { return d.fullMethod }
func (d contextInterceptorData) ClientIdentity() string { return d.clientIdentity }
type Interceptor = func(ctx context.Context, data ContextInterceptorData, handler func(ctx context.Context))
func NewInterceptors(logger Logger, clientIdentityKey interface{}, interceptor Interceptor) (unary grpc.UnaryServerInterceptor, stream grpc.StreamServerInterceptor) {
unary = func(ctx context.Context, req interface{}, info *grpc.UnaryServerInfo, handler grpc.UnaryHandler) (interface{}, error) {
logger.WithField("fullMethod", info.FullMethod).Debug("request")
p, ok := peer.FromContext(ctx)
if !ok {
@@ -115,10 +128,18 @@ func NewInterceptors(logger Logger, clientIdentityKey interface{}, ctxIntercepto
}
logger.WithField("peer_client_identity", a.clientIdentity).Debug("peer client identity")
ctx = context.WithValue(ctx, clientIdentityKey, a.clientIdentity)
if ctxInterceptor != nil {
ctx = ctxInterceptor(ctx)
data := contextInterceptorData{
fullMethod: info.FullMethod,
clientIdentity: a.clientIdentity,
}
return handler(ctx, req)
var (
resp interface{}
err error
)
interceptor(ctx, data, func(ctx context.Context) {
resp, err = handler(ctx, req) // no-shadow
})
return resp, err
}
stream = func(srv interface{}, ss grpc.ServerStream, info *grpc.StreamServerInfo, handler grpc.StreamHandler) error {
panic("unimplemented")
@@ -50,7 +50,7 @@ func ClientConn(cn transport.Connecter, log Logger) *grpc.ClientConn {
}
// NewServer is a convenience interface around the TransportCredentials and Interceptors interface.
func NewServer(authListener transport.AuthenticatedListener, clientIdentityKey interface{}, logger grpcclientidentity.Logger, ctxInterceptor grpcclientidentity.ContextInterceptor) (srv *grpc.Server, serve func() error) {
func NewServer(authListener transport.AuthenticatedListener, clientIdentityKey interface{}, logger grpcclientidentity.Logger, ctxInterceptor grpcclientidentity.Interceptor) (srv *grpc.Server, serve func() error) {
ka := grpc.KeepaliveParams(keepalive.ServerParameters{
Time: StartKeepalivesAfterInactivityDuration,
Timeout: KeepalivePeerTimeout,
+28
View File
@@ -10,6 +10,7 @@ import (
"sync/atomic"
"time"
"github.com/zrepl/zrepl/daemon/logging/trace"
"google.golang.org/grpc"
"github.com/google/uuid"
@@ -83,6 +84,9 @@ func (c *Client) Close() {
// callers must ensure that the returned io.ReadCloser is closed
// TODO expose dataClient interface to the outside world
func (c *Client) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error) {
ctx, endSpan := trace.WithSpan(ctx, "rpc.client.Send")
defer endSpan()
// TODO the returned sendStream may return a read error created by the remote side
res, stream, err := c.dataClient.ReqSend(ctx, r)
if err != nil {
@@ -97,34 +101,58 @@ func (c *Client) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, io.Rea
}
func (c *Client) Receive(ctx context.Context, req *pdu.ReceiveReq, stream io.ReadCloser) (*pdu.ReceiveRes, error) {
ctx, endSpan := trace.WithSpan(ctx, "rpc.client.Receive")
defer endSpan()
return c.dataClient.ReqRecv(ctx, req, stream)
}
func (c *Client) ListFilesystems(ctx context.Context, in *pdu.ListFilesystemReq) (*pdu.ListFilesystemRes, error) {
ctx, endSpan := trace.WithSpan(ctx, "rpc.client.ListFilesystems")
defer endSpan()
return c.controlClient.ListFilesystems(ctx, in)
}
func (c *Client) ListFilesystemVersions(ctx context.Context, in *pdu.ListFilesystemVersionsReq) (*pdu.ListFilesystemVersionsRes, error) {
ctx, endSpan := trace.WithSpan(ctx, "rpc.client.ListFilesystemVersions")
defer endSpan()
return c.controlClient.ListFilesystemVersions(ctx, in)
}
func (c *Client) DestroySnapshots(ctx context.Context, in *pdu.DestroySnapshotsReq) (*pdu.DestroySnapshotsRes, error) {
ctx, endSpan := trace.WithSpan(ctx, "rpc.client.DestroySnapshots")
defer endSpan()
return c.controlClient.DestroySnapshots(ctx, in)
}
func (c *Client) ReplicationCursor(ctx context.Context, in *pdu.ReplicationCursorReq) (*pdu.ReplicationCursorRes, error) {
ctx, endSpan := trace.WithSpan(ctx, "rpc.client.ReplicationCursor")
defer endSpan()
return c.controlClient.ReplicationCursor(ctx, in)
}
func (c *Client) SendCompleted(ctx context.Context, in *pdu.SendCompletedReq) (*pdu.SendCompletedRes, error) {
ctx, endSpan := trace.WithSpan(ctx, "rpc.client.SendCompleted")
defer endSpan()
return c.controlClient.SendCompleted(ctx, in)
}
func (c *Client) HintMostRecentCommonAncestor(ctx context.Context, in *pdu.HintMostRecentCommonAncestorReq) (*pdu.HintMostRecentCommonAncestorRes, error) {
ctx, endSpan := trace.WithSpan(ctx, "rpc.client.HintMostRecentCommonAncestor")
defer endSpan()
return c.controlClient.HintMostRecentCommonAncestor(ctx, in)
}
func (c *Client) WaitForConnectivity(ctx context.Context) error {
ctx, endSpan := trace.WithSpan(ctx, "rpc.client.WaitForConnectivity")
defer endSpan()
ctx, cancel := context.WithCancel(ctx)
defer cancel()
msg := uuid.New().String()
+6 -12
View File
@@ -3,17 +3,12 @@ package rpc
import (
"context"
"github.com/zrepl/zrepl/daemon/logging"
"github.com/zrepl/zrepl/logger"
)
type Logger = logger.Logger
type contextKey int
const (
contextKeyLoggers contextKey = iota
)
/// All fields must be non-nil
type Loggers struct {
General Logger
@@ -21,11 +16,10 @@ type Loggers struct {
Data Logger
}
func WithLoggers(ctx context.Context, loggers Loggers) context.Context {
ctx = context.WithValue(ctx, contextKeyLoggers, loggers)
return ctx
}
func GetLoggersOrPanic(ctx context.Context) Loggers {
return ctx.Value(contextKeyLoggers).(Loggers)
return Loggers{
General: logging.GetLogger(ctx, logging.SubsysRPC),
Control: logging.GetLogger(ctx, logging.SubsysRPCControl),
Data: logging.GetLogger(ctx, logging.SubsysRPCData),
}
}
+23 -6
View File
@@ -7,6 +7,7 @@ import (
"github.com/zrepl/zrepl/endpoint"
"github.com/zrepl/zrepl/replication/logic/pdu"
"github.com/zrepl/zrepl/rpc/dataconn"
"github.com/zrepl/zrepl/rpc/grpcclientidentity"
"github.com/zrepl/zrepl/rpc/grpcclientidentity/grpchelper"
"github.com/zrepl/zrepl/rpc/versionhandshake"
"github.com/zrepl/zrepl/transport"
@@ -30,7 +31,20 @@ type Server struct {
dataServerServe serveFunc
}
type HandlerContextInterceptor func(ctx context.Context) context.Context
type HandlerContextInterceptorData interface {
FullMethod() string
ClientIdentity() string
}
type interceptorData struct {
prefixMethod string
wrapped HandlerContextInterceptorData
}
func (d interceptorData) ClientIdentity() string { return d.wrapped.ClientIdentity() }
func (d interceptorData) FullMethod() string { return d.prefixMethod + d.wrapped.FullMethod() }
type HandlerContextInterceptor func(ctx context.Context, data HandlerContextInterceptorData, handler func(ctx context.Context))
// config must be valid (use its Validate function).
func NewServer(handler Handler, loggers Loggers, ctxInterceptor HandlerContextInterceptor) *Server {
@@ -38,7 +52,10 @@ func NewServer(handler Handler, loggers Loggers, ctxInterceptor HandlerContextIn
// setup control server
controlServerServe := func(ctx context.Context, controlListener transport.AuthenticatedListener, errOut chan<- error) {
controlServer, serve := grpchelper.NewServer(controlListener, endpoint.ClientIdentityKey, loggers.Control, ctxInterceptor)
var controlCtxInterceptor grpcclientidentity.Interceptor = func(ctx context.Context, data grpcclientidentity.ContextInterceptorData, handler func(ctx context.Context)) {
ctxInterceptor(ctx, interceptorData{"control://", data}, handler)
}
controlServer, serve := grpchelper.NewServer(controlListener, endpoint.ClientIdentityKey, loggers.Control, controlCtxInterceptor)
pdu.RegisterReplicationServer(controlServer, handler)
// give time for graceful stop until deadline expires, then hard stop
@@ -59,12 +76,12 @@ func NewServer(handler Handler, loggers Loggers, ctxInterceptor HandlerContextIn
dataServerClientIdentitySetter := func(ctx context.Context, wire *transport.AuthConn) (context.Context, *transport.AuthConn) {
ci := wire.ClientIdentity()
ctx = context.WithValue(ctx, endpoint.ClientIdentityKey, ci)
if ctxInterceptor != nil {
ctx = ctxInterceptor(ctx) // SHADOWING
}
return ctx, wire
}
dataServer := dataconn.NewServer(dataServerClientIdentitySetter, loggers.Data, handler)
var dataCtxInterceptor dataconn.ContextInterceptor = func(ctx context.Context, data dataconn.ContextInterceptorData, handler func(ctx context.Context)) {
ctxInterceptor(ctx, interceptorData{"data://", data}, handler)
}
dataServer := dataconn.NewServer(dataServerClientIdentitySetter, dataCtxInterceptor, loggers.Data, handler)
dataServerServe := func(ctx context.Context, dataListener transport.AuthenticatedListener, errOut chan<- error) {
dataServer.Serve(ctx, dataListener)
errOut <- nil // TODO bad design of dataServer?
+2 -14
View File
@@ -15,27 +15,15 @@ import (
"net"
"time"
"github.com/zrepl/zrepl/daemon/logging"
"github.com/zrepl/zrepl/logger"
"github.com/zrepl/zrepl/transport"
)
type contextKey int
const (
contextKeyLog contextKey = 1 + iota
)
type Logger = logger.Logger
func WithLogger(ctx context.Context, log Logger) context.Context {
return context.WithValue(ctx, contextKeyLog, log)
}
func getLog(ctx context.Context) Logger {
if l, ok := ctx.Value(contextKeyLog).(Logger); ok {
return l
}
return logger.NewNullLogger()
return logging.GetLogger(ctx, logging.SubsysTransportMux)
}
type acceptRes struct {