Set default conn timeout of 5 sec for storeGateway and alertmanager clients (#6603)

* Set default conn timeout for SG cals

Signed-off-by: Daniel Deluiggi <[email protected]>

* Changelog

Signed-off-by: Daniel Deluiggi <[email protected]>

* Add 5s to alertmanager

Signed-off-by: Daniel Deluiggi <[email protected]>

* Docs

Signed-off-by: Daniel Deluiggi <[email protected]>

* Changelog

Signed-off-by: Daniel Deluiggi <[email protected]>

* Rollback description change

Signed-off-by: Daniel Deluiggi <[email protected]>

---------

Signed-off-by: Daniel Deluiggi <[email protected]>

Author: danielblando

CHANGELOG.md

@@ -1,6 +1,7 @@
 # Changelog
 
 ## master / unreleased
+* [CHANGE] StoreGateway/Alertmanager: Add default 5s connection timeout on client. #6603
 * [FEATURE] Query Frontend: Add dynamic interval size for query splitting. This is enabled by configuring experimental flags `querier.max-shards-per-query` and/or `querier.max-fetched-data-duration-per-query`. The split interval size is dynamically increased to maintain a number of shards and total duration fetched below the configured values. #6458
 * [FEATURE] Querier/Ruler: Add `query_partial_data` and `rules_partial_data` limits to allow queries/rules to be evaluated with data from a single zone, if other zones are not available. #6526
 * [FEATURE] Update prometheus alertmanager version to v0.28.0 and add new integration msteamsv2, jira, and rocketchat. #6590

docs/blocks-storage/querier.md

@@ -222,6 +222,11 @@ querier:
       # CLI flag: -querier.store-gateway-client.healthcheck.timeout
       [timeout: <duration> | default = 1s]
 
+    # The maximum amount of time to establish a connection. A value of 0 means
+    # using default gRPC client connect timeout 5s.
+    # CLI flag: -querier.store-gateway-client.connect-timeout
+    [connect_timeout: <duration> | default = 5s]
+
   # If enabled, store gateway query stats will be logged using `info` log level.
   # CLI flag: -querier.store-gateway-query-stats-enabled
   [store_gateway_query_stats: <boolean> | default = true]

docs/configuration/config-file-reference.md

@@ -503,6 +503,11 @@ alertmanager_client:
   # CLI flag: -alertmanager.alertmanager-client.grpc-max-send-msg-size
   [max_send_msg_size: <int> | default = 4194304]
 
+  # The maximum amount of time to establish a connection. A value of 0 means
+  # using default gRPC client connect timeout 5s.
+  # CLI flag: -alertmanager.alertmanager-client.connect-timeout
+  [connect_timeout: <duration> | default = 5s]
+
 # The interval between persisting the current alertmanager state (notification
 # log and silences) to object storage. This is only used when sharding is
 # enabled. This state is read when all replicas for a shard can not be
@@ -4100,6 +4105,11 @@ store_gateway_client:
     # CLI flag: -querier.store-gateway-client.healthcheck.timeout
     [timeout: <duration> | default = 1s]
 
+  # The maximum amount of time to establish a connection. A value of 0 means
+  # using default gRPC client connect timeout 5s.
+  # CLI flag: -querier.store-gateway-client.connect-timeout
+  [connect_timeout: <duration> | default = 5s]
+
 # If enabled, store gateway query stats will be logged using `info` log level.
 # CLI flag: -querier.store-gateway-query-stats-enabled
 [store_gateway_query_stats: <boolean> | default = true]

pkg/alertmanager/alertmanager_client.go

@@ -40,6 +40,7 @@ type ClientConfig struct {
 	GRPCCompression string           `yaml:"grpc_compression"`
 	MaxRecvMsgSize  int              `yaml:"max_recv_msg_size"`
 	MaxSendMsgSize  int              `yaml:"max_send_msg_size"`
+	ConnectTimeout  time.Duration    `yaml:"connect_timeout"`
 }
 
 // RegisterFlagsWithPrefix registers flags with prefix.
@@ -50,6 +51,7 @@ func (cfg *ClientConfig) RegisterFlagsWithPrefix(prefix string, f *flag.FlagSet)
 	cfg.TLS.RegisterFlagsWithPrefix(prefix, f)
 	f.IntVar(&cfg.MaxRecvMsgSize, prefix+".grpc-max-recv-msg-size", 16*1024*1024, "gRPC client max receive message size (bytes).")
 	f.IntVar(&cfg.MaxSendMsgSize, prefix+".grpc-max-send-msg-size", 4*1024*1024, "gRPC client max send message size (bytes).")
+	f.DurationVar(&cfg.ConnectTimeout, prefix+".connect-timeout", 5*time.Second, "The maximum amount of time to establish a connection. A value of 0 means using default gRPC client connect timeout 5s.")
 }
 
 type alertmanagerClientsPool struct {
@@ -67,6 +69,7 @@ func newAlertmanagerClientsPool(discovery client.PoolServiceDiscovery, amClientC
 		BackoffOnRatelimits: false,
 		TLSEnabled:          amClientCfg.TLSEnabled,
 		TLS:                 amClientCfg.TLS,
+		ConnectTimeout:      amClientCfg.ConnectTimeout,
 	}
 
 	requestDuration := promauto.With(reg).NewHistogramVec(prometheus.HistogramOpts{

pkg/querier/store_gateway_client.go

@@ -79,6 +79,7 @@ func newStoreGatewayClientPool(discovery client.PoolServiceDiscovery, clientConf
 			BackoffOnRatelimits: false,
 			TLSEnabled:          clientConfig.TLSEnabled,
 			TLS:                 clientConfig.TLS,
+			ConnectTimeout:      clientConfig.ConnectTimeout,
 		},
 		HealthCheckConfig: clientConfig.HealthCheckConfig,
 	}
@@ -103,11 +104,13 @@ type ClientConfig struct {
 	TLS               tls.ClientConfig             `yaml:",inline"`
 	GRPCCompression   string                       `yaml:"grpc_compression"`
 	HealthCheckConfig grpcclient.HealthCheckConfig `yaml:"healthcheck_config" doc:"description=EXPERIMENTAL: If enabled, gRPC clients perform health checks for each target and fail the request if the target is marked as unhealthy."`
+	ConnectTimeout    time.Duration                `yaml:"connect_timeout"`
 }
 
 func (cfg *ClientConfig) RegisterFlagsWithPrefix(prefix string, f *flag.FlagSet) {
 	f.BoolVar(&cfg.TLSEnabled, prefix+".tls-enabled", cfg.TLSEnabled, "Enable TLS for gRPC client connecting to store-gateway.")
 	f.StringVar(&cfg.GRPCCompression, prefix+".grpc-compression", "", "Use compression when sending messages. Supported values are: 'gzip', 'snappy' and '' (disable compression)")
+	f.DurationVar(&cfg.ConnectTimeout, prefix+".connect-timeout", 5*time.Second, "The maximum amount of time to establish a connection. A value of 0 means using default gRPC client connect timeout 5s.")
 	cfg.TLS.RegisterFlagsWithPrefix(prefix, f)
 	cfg.HealthCheckConfig.RegisterFlagsWithPrefix(prefix, f)
 }