diff --git a/CHANGELOG.md b/CHANGELOG.md index f45fb2e44af..655005269f4 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -20,9 +20,10 @@ Accepted into CNCF: ### Added +- [#1478](https://github.com/thanos-io/thanos/pull/1478) Thanos components now exposes gRPC server metrics as soon as server starts, to provide more reliable data for instrumentation. - [#1378](https://github.com/thanos-io/thanos/pull/1378) Thanos Receive now exposes `thanos_receive_config_hash`, `thanos_receive_config_last_reload_successful` and `thanos_receive_config_last_reload_success_timestamp_seconds` metrics to track latest configuration change -- [#1268](https://github.com/thanos-io/thanos/pull/1268) Thanos Sidecar added support for newest Prometheus streaming remote read added [here](https://github.com/prometheus/prometheus/pull/5703). This massively improves memory required by single - request for both Prometheus and sidecar. Single requests now should take constant amount of memory on sidecar, so resource consumption prediction is now straightforward. This will be used if you have Prometheus `2.13` or `2.12-master`. +- [#1268](https://github.com/thanos-io/thanos/pull/1268) Thanos Sidecar added support for newest Prometheus streaming remote read added [here](https://github.com/prometheus/prometheus/pull/5703). This massively improves memory required by single + request for both Prometheus and sidecar. Single requests now should take constant amount of memory on sidecar, so resource consumption prediction is now straightforward. This will be used if you have Prometheus `2.13` or `2.12-master`. - [#1358](https://github.com/thanos-io/thanos/pull/1358) Added `part_size` configuration option for HTTP multipart requests minimum part size for S3 storage type - [#1363](https://github.com/thanos-io/thanos/pull/1363) Thanos Receive now exposes `thanos_receive_hashring_nodes` and `thanos_receive_hashring_tenants` metrics to monitor status of hash-rings - [#1395](https://github.com/thanos-io/thanos/pull/1395) Thanos Sidecar added `/-/ready` and `/-/healthy` endpoints to Thanos sidecar. @@ -39,7 +40,7 @@ Accepted into CNCF: - [BUGFIX] prometheus_tsdb_compactions_failed_total is now incremented on any compaction failure. tsdb#613 - [BUGFIX] PromQL: Correctly display {__name__="a"}. - [#1338](https://github.com/thanos-io/thanos/pull/1338) Thanos Query still warns on store API duplicate, but allows a single one from duplicated set. This is gracefully warn about the problematic logic and not disrupt immediately. -- [#1385](https://github.com/thanos-io/thanos/pull/1385) Thanos Compact exposes flag to disable downsampling `downsampling.disable`. +- [#1385](https://github.com/thanos-io/thanos/pull/1385) Thanos Compact exposes flag to disable downsampling `downsampling.disable`. ### Fixed diff --git a/cmd/thanos/main.go b/cmd/thanos/main.go index 2e6f2bf413f..b44a2c9fc09 100644 --- a/cmd/thanos/main.go +++ b/cmd/thanos/main.go @@ -33,6 +33,7 @@ import ( "github.com/prometheus/common/version" "github.com/thanos-io/thanos/pkg/prober" "github.com/thanos-io/thanos/pkg/runutil" + "github.com/thanos-io/thanos/pkg/store/storepb" "github.com/thanos-io/thanos/pkg/tracing" "github.com/thanos-io/thanos/pkg/tracing/client" "go.uber.org/automaxprocs/maxprocs" @@ -244,41 +245,8 @@ func registerMetrics(mux *http.ServeMux, g prometheus.Gatherer) { mux.Handle("/metrics", promhttp.HandlerFor(g, promhttp.HandlerOpts{})) } -// defaultGRPCServerOpts returns default gRPC server opts that includes: -// - request histogram -// - tracing -// - panic recovery with panic counter -func defaultGRPCServerOpts(logger log.Logger, reg *prometheus.Registry, tracer opentracing.Tracer, cert, key, clientCA string) ([]grpc.ServerOption, error) { - met := grpc_prometheus.NewServerMetrics() - met.EnableHandlingTimeHistogram( - grpc_prometheus.WithHistogramBuckets([]float64{ - 0.001, 0.01, 0.05, 0.1, 0.2, 0.4, 0.8, 1.6, 3.2, 6.4, - }), - ) - - panicsTotal := prometheus.NewCounter(prometheus.CounterOpts{ - Name: "thanos_grpc_req_panics_recovered_total", - Help: "Total number of gRPC requests recovered from internal panic.", - }) - grpcPanicRecoveryHandler := func(p interface{}) (err error) { - panicsTotal.Inc() - level.Error(logger).Log("msg", "recovered from panic", "panic", p, "stack", debug.Stack()) - return status.Errorf(codes.Internal, "%s", p) - } - reg.MustRegister(met, panicsTotal) - opts := []grpc.ServerOption{ - grpc.MaxSendMsgSize(math.MaxInt32), - grpc_middleware.WithUnaryServerChain( - met.UnaryServerInterceptor(), - tracing.UnaryServerInterceptor(tracer), - grpc_recovery.UnaryServerInterceptor(grpc_recovery.WithRecoveryHandler(grpcPanicRecoveryHandler)), - ), - grpc_middleware.WithStreamServerChain( - met.StreamServerInterceptor(), - tracing.StreamServerInterceptor(tracer), - grpc_recovery.StreamServerInterceptor(grpc_recovery.WithRecoveryHandler(grpcPanicRecoveryHandler)), - ), - } +func defaultGRPCServerOpts(logger log.Logger, cert, key, clientCA string) ([]grpc.ServerOption, error) { + opts := []grpc.ServerOption{} if key == "" && cert == "" { if clientCA != "" { @@ -325,6 +293,45 @@ func defaultGRPCServerOpts(logger log.Logger, reg *prometheus.Registry, tracer o return append(opts, grpc.Creds(credentials.NewTLS(tlsCfg))), nil } +func newStoreGRPCServer(logger log.Logger, reg *prometheus.Registry, tracer opentracing.Tracer, srv storepb.StoreServer, opts []grpc.ServerOption) *grpc.Server { + met := grpc_prometheus.NewServerMetrics() + met.EnableHandlingTimeHistogram( + grpc_prometheus.WithHistogramBuckets([]float64{ + 0.001, 0.01, 0.05, 0.1, 0.2, 0.4, 0.8, 1.6, 3.2, 6.4, + }), + ) + panicsTotal := prometheus.NewCounter(prometheus.CounterOpts{ + Name: "thanos_grpc_req_panics_recovered_total", + Help: "Total number of gRPC requests recovered from internal panic.", + }) + reg.MustRegister(met, panicsTotal) + + grpcPanicRecoveryHandler := func(p interface{}) (err error) { + panicsTotal.Inc() + level.Error(logger).Log("msg", "recovered from panic", "panic", p, "stack", debug.Stack()) + return status.Errorf(codes.Internal, "%s", p) + } + opts = append(opts, + grpc.MaxSendMsgSize(math.MaxInt32), + grpc_middleware.WithUnaryServerChain( + met.UnaryServerInterceptor(), + tracing.UnaryServerInterceptor(tracer), + grpc_recovery.UnaryServerInterceptor(grpc_recovery.WithRecoveryHandler(grpcPanicRecoveryHandler)), + ), + grpc_middleware.WithStreamServerChain( + met.StreamServerInterceptor(), + tracing.StreamServerInterceptor(tracer), + grpc_recovery.StreamServerInterceptor(grpc_recovery.WithRecoveryHandler(grpcPanicRecoveryHandler)), + ), + ) + + s := grpc.NewServer(opts...) + storepb.RegisterStoreServer(s, srv) + met.InitializeMetrics(s) + + return s +} + // TODO Remove once all components are migrated to the new defaultHTTPListener. // metricHTTPListenGroup is a run.Group that servers HTTP endpoint with only Prometheus metrics. func metricHTTPListenGroup(g *run.Group, logger log.Logger, reg *prometheus.Registry, httpBindAddr string) error { diff --git a/cmd/thanos/query.go b/cmd/thanos/query.go index 4577ac0b3ab..0eb2ac03995 100644 --- a/cmd/thanos/query.go +++ b/cmd/thanos/query.go @@ -34,7 +34,6 @@ import ( v1 "github.com/thanos-io/thanos/pkg/query/api" "github.com/thanos-io/thanos/pkg/runutil" "github.com/thanos-io/thanos/pkg/store" - "github.com/thanos-io/thanos/pkg/store/storepb" "github.com/thanos-io/thanos/pkg/tracing" "github.com/thanos-io/thanos/pkg/ui" "google.golang.org/grpc" @@ -441,13 +440,11 @@ func runQuery( } logger := log.With(logger, "component", component.Query.String()) - opts, err := defaultGRPCServerOpts(logger, reg, tracer, srvCert, srvKey, srvClientCA) + opts, err := defaultGRPCServerOpts(logger, srvCert, srvKey, srvClientCA) if err != nil { return errors.Wrapf(err, "build gRPC server") } - - s := grpc.NewServer(opts...) - storepb.RegisterStoreServer(s, proxy) + s := newStoreGRPCServer(logger, reg, tracer, proxy, opts) g.Add(func() error { level.Info(logger).Log("msg", "Listening for StoreAPI gRPC", "address", grpcBindAddr) diff --git a/cmd/thanos/receive.go b/cmd/thanos/receive.go index 78c8c1bd1b0..85cd5a05326 100644 --- a/cmd/thanos/receive.go +++ b/cmd/thanos/receive.go @@ -24,7 +24,6 @@ import ( "github.com/thanos-io/thanos/pkg/runutil" "github.com/thanos-io/thanos/pkg/shipper" "github.com/thanos-io/thanos/pkg/store" - "github.com/thanos-io/thanos/pkg/store/storepb" "google.golang.org/grpc" kingpin "gopkg.in/alecthomas/kingpin.v2" ) @@ -267,12 +266,11 @@ func runReceive( db := localStorage.Get() tsdbStore := store.NewTSDBStore(log.With(logger, "component", "thanos-tsdb-store"), reg, db, component.Receive, lset) - opts, err := defaultGRPCServerOpts(logger, reg, tracer, cert, key, clientCA) + opts, err := defaultGRPCServerOpts(logger, cert, key, clientCA) if err != nil { return errors.Wrap(err, "setup gRPC server") } - s = grpc.NewServer(opts...) - storepb.RegisterStoreServer(s, tsdbStore) + s := newStoreGRPCServer(logger, reg, tracer, tsdbStore, opts) level.Info(logger).Log("msg", "listening for StoreAPI gRPC", "address", grpcBindAddr) return errors.Wrap(s.Serve(l), "serve gRPC") diff --git a/cmd/thanos/rule.go b/cmd/thanos/rule.go index 6226ca44795..43ce9a0390a 100644 --- a/cmd/thanos/rule.go +++ b/cmd/thanos/rule.go @@ -50,7 +50,6 @@ import ( "github.com/thanos-io/thanos/pkg/store/storepb" "github.com/thanos-io/thanos/pkg/tracing" "github.com/thanos-io/thanos/pkg/ui" - "google.golang.org/grpc" kingpin "gopkg.in/alecthomas/kingpin.v2" ) @@ -493,12 +492,11 @@ func runRule( store := store.NewTSDBStore(logger, reg, db, component.Rule, lset) - opts, err := defaultGRPCServerOpts(logger, reg, tracer, cert, key, clientCA) + opts, err := defaultGRPCServerOpts(logger, cert, key, clientCA) if err != nil { return errors.Wrap(err, "setup gRPC options") } - s := grpc.NewServer(opts...) - storepb.RegisterStoreServer(s, store) + s := newStoreGRPCServer(logger, reg, tracer, store, opts) g.Add(func() error { return errors.Wrap(s.Serve(l), "serve gRPC") diff --git a/cmd/thanos/sidecar.go b/cmd/thanos/sidecar.go index 3e4655c5d14..2a3c5afe9e3 100644 --- a/cmd/thanos/sidecar.go +++ b/cmd/thanos/sidecar.go @@ -26,7 +26,6 @@ import ( "github.com/thanos-io/thanos/pkg/shipper" "github.com/thanos-io/thanos/pkg/store" "github.com/thanos-io/thanos/pkg/store/storepb" - "google.golang.org/grpc" "gopkg.in/alecthomas/kingpin.v2" ) @@ -220,12 +219,11 @@ func runSidecar( return errors.Wrap(err, "create Prometheus store") } - opts, err := defaultGRPCServerOpts(logger, reg, tracer, cert, key, clientCA) + opts, err := defaultGRPCServerOpts(logger, cert, key, clientCA) if err != nil { return errors.Wrap(err, "setup gRPC server") } - s := grpc.NewServer(opts...) - storepb.RegisterStoreServer(s, promStore) + s := newStoreGRPCServer(logger, reg, tracer, promStore, opts) g.Add(func() error { level.Info(logger).Log("msg", "Listening for StoreAPI gRPC", "address", grpcBindAddr) diff --git a/cmd/thanos/store.go b/cmd/thanos/store.go index 23a97b344a8..c3f7ccb8710 100644 --- a/cmd/thanos/store.go +++ b/cmd/thanos/store.go @@ -16,8 +16,6 @@ import ( "github.com/thanos-io/thanos/pkg/runutil" "github.com/thanos-io/thanos/pkg/store" storecache "github.com/thanos-io/thanos/pkg/store/cache" - "github.com/thanos-io/thanos/pkg/store/storepb" - "google.golang.org/grpc" kingpin "gopkg.in/alecthomas/kingpin.v2" ) @@ -187,13 +185,11 @@ func runStore( return errors.Wrap(err, "listen API address") } - opts, err := defaultGRPCServerOpts(logger, reg, tracer, cert, key, clientCA) + opts, err := defaultGRPCServerOpts(logger, cert, key, clientCA) if err != nil { return errors.Wrap(err, "grpc server options") } - - s := grpc.NewServer(opts...) - storepb.RegisterStoreServer(s, bs) + s := newStoreGRPCServer(logger, reg, tracer, bs, opts) g.Add(func() error { level.Info(logger).Log("msg", "Listening for StoreAPI gRPC", "address", grpcBindAddr)