boulder/grpc/interceptors.go

93 lines
2.9 KiB
Go

package grpc
import (
"errors"
"strings"
"time"
"github.com/grpc-ecosystem/go-grpc-prometheus"
"github.com/letsencrypt/boulder/metrics"
"github.com/jmhodges/clock"
"golang.org/x/net/context"
"google.golang.org/grpc"
)
// serverInterceptor is a gRPC interceptor that adds statsd and Prometheus
// metrics to requests handled by a gRPC server.
type serverInterceptor struct {
stats metrics.Scope
clk clock.Clock
}
func cleanMethod(m string, trimService bool) string {
m = strings.TrimLeft(m, "-")
if trimService {
s := strings.Split(m, "-")
if len(s) == 1 {
return m
}
return s[len(s)-1]
}
return strings.Replace(m, "-", "_", -1)
}
func (si *serverInterceptor) intercept(ctx context.Context, req interface{}, info *grpc.UnaryServerInfo, handler grpc.UnaryHandler) (interface{}, error) {
if info == nil {
si.stats.Inc("NoInfo", 1)
return nil, errors.New("passed nil *grpc.UnaryServerInfo")
}
s := si.clk.Now()
methodScope := si.stats.NewScope(cleanMethod(info.FullMethod, true))
methodScope.Inc("Calls", 1)
methodScope.GaugeDelta("InProgress", 1)
resp, err := grpc_prometheus.UnaryServerInterceptor(ctx, req, info, handler)
methodScope.TimingDuration("Latency", si.clk.Since(s))
methodScope.GaugeDelta("InProgress", -1)
if err != nil {
methodScope.Inc("Failed", 1)
}
return resp, err
}
// clientInterceptor is a gRPC interceptor that adds statsd and Prometheus
// metrics to sent requests, and disables FailFast. We disable FailFast because
// non-FailFast mode is most similar to the old AMQP RPC layer: If a client
// makes a request while all backends are briefly down (e.g. for a restart), the
// request doesn't necessarily fail. A backend can service the request if it
// comes back up within the timeout. Under gRPC the same effect is achieved by
// retries up to the Context deadline.
type clientInterceptor struct {
stats metrics.Scope
clk clock.Clock
timeout time.Duration
}
// intercept fulfils the grpc.UnaryClientInterceptor interface, it should be noted that while this API
// is currently experimental the metrics it reports should be kept as stable as can be, *within reason*.
func (ci *clientInterceptor) intercept(
ctx context.Context,
method string,
req,
reply interface{},
cc *grpc.ClientConn,
invoker grpc.UnaryInvoker,
opts ...grpc.CallOption) error {
localCtx, cancel := context.WithTimeout(ctx, ci.timeout)
defer cancel()
s := ci.clk.Now()
methodScope := ci.stats.NewScope(cleanMethod(method, false))
methodScope.Inc("Calls", 1)
methodScope.GaugeDelta("InProgress", 1)
// Disable fail-fast so RPCs will retry until deadline, even if all backends
// are down.
opts = append(opts, grpc.FailFast(false))
err := grpc_prometheus.UnaryClientInterceptor(localCtx, method, req, reply, cc, invoker, opts...)
methodScope.TimingDuration("Latency", ci.clk.Since(s))
methodScope.GaugeDelta("InProgress", -1)
if err != nil {
methodScope.Inc("Failed", 1)
}
return err
}