vttestserver: make tablet_refresh_interval configurable and reduce default value (vitessio#11918)

* setting up refresh_interval and check_leader_interval

Signed-off-by: Rameez Sajwani <[email protected]>

* setting default value, if not provided

Signed-off-by: Rameez Sajwani <[email protected]>

* fixing vttablet flags

Signed-off-by: Rameez Sajwani <[email protected]>

* fixing vtttestserver flags

Signed-off-by: Rameez Sajwani <[email protected]>

* adding comments and removing unwanted script

Signed-off-by: Rameez Sajwani <[email protected]>

* code review

Signed-off-by: Rameez Sajwani <[email protected]>

* removing leader Check interval

Signed-off-by: Rameez Sajwani <[email protected]>

* change to flags out put file

Signed-off-by: Rameez Sajwani <[email protected]>

* adding summary

Signed-off-by: Rameez Sajwani <[email protected]>

* code review feedback

Signed-off-by: Rameez Sajwani <[email protected]>

Signed-off-by: Rameez Sajwani <[email protected]>

Author: rsajwani

doc/releasenotes/16_0_0_summary.md

@@ -246,4 +246,13 @@ VSchema Example
 
 #### Flag Deprecations
 
-The flag `lock-shard-timeout` has been deprecated. Please use the newly introduced `lock-timeout` instead. More detail [here](#lock-timeout-introduction).
+The flag `lock-shard-timeout` has been deprecated. Please use the newly introduced `lock-timeout` instead. More detail [here](#lock-timeout-introduction).
+
+### VTTestServer
+
+#### Improvement
+
+Creating a database with vttestserver was taking ~45 seconds. This can be problematic in test environments where testcases do a lot of `create` and `drop` database.
+In an effort to minimize the database creation time, we have changed the value of `tablet_refresh_interval` to 10s while instantiating vtcombo during vttestserver initialization. We have also made this configurable so that it can be reduced further if desired.
+For any production cluster the default value of this flag is still [1 minute](https://vitess.io/docs/15.0/reference/programs/vtgate/). Reducing this values might put more stress on Topo Server (since we now read from Topo server more often) but for testing purposes 
+this shouldn't be a concern.

docker/vttestserver/run.sh

@@ -42,5 +42,6 @@ rm -vf "$VTDATAROOT"/"$tablet_dir"/{mysql.sock,mysql.sock.lock}
 	--enable_direct_ddl="${ENABLE_DIRECT_DDL:-true}" \
 	--planner-version="${PLANNER_VERSION:-v3}" \
 	--vschema_ddl_authorized_users=% \
+	--tablet_refresh_interval "$TABLET_REFRESH_INTERVAL"
 	--schema_dir="/vt/schema/"
 

go/cmd/vttestserver/main.go

@@ -26,6 +26,7 @@ import (
 	"strings"
 	"sync"
 	"syscall"
+	"time"
 
 	"github.com/spf13/pflag"
 	"google.golang.org/protobuf/encoding/prototext"
@@ -165,6 +166,7 @@ func registerFlags(fs *pflag.FlagSet) {
 	fs.StringVar(&config.ExternalTopoGlobalServerAddress, "external_topo_global_server_address", "", "the address of the global topology server for vtcombo process")
 	fs.StringVar(&config.ExternalTopoGlobalRoot, "external_topo_global_root", "", "the path of the global topology data in the global topology server for vtcombo process")
 
+	fs.DurationVar(&config.VtgateTabletRefreshInterval, "tablet_refresh_interval", 10*time.Second, "Interval at which vtgate refreshes tablet information from topology server.")
 	acl.RegisterFlags(fs)
 }
 

go/flags/endtoend/vttestserver.txt

@@ -108,6 +108,7 @@ Usage of vttestserver:
       --tablet_manager_grpc_key string                                   the key to use to connect
       --tablet_manager_grpc_server_name string                           the server name to use to validate server certificate
       --tablet_manager_protocol string                                   Protocol to use to make tabletmanager RPCs to vttablets. (default "grpc")
+      --tablet_refresh_interval duration                                 Interval at which vtgate refreshes tablet information from topology server. (default 10s)
       --topo_consul_lock_delay duration                                  LockDelay for consul session. (default 15s)
       --topo_consul_lock_session_checks string                           List of checks for consul session. (default "serfHealth")
       --topo_consul_lock_session_ttl string                              TTL for consul session.

go/vt/vttest/local_cluster.go

@@ -27,6 +27,7 @@ import (
 	"path"
 	"path/filepath"
 	"strings"
+	"time"
 	"unicode"
 
 	"google.golang.org/protobuf/encoding/protojson"
@@ -143,6 +144,8 @@ type Config struct {
 	ExternalTopoGlobalServerAddress string
 
 	ExternalTopoGlobalRoot string
+
+	VtgateTabletRefreshInterval time.Duration
 }
 
 // InitSchemas is a shortcut for tests that just want to setup a single

go/vt/vttest/vtprocess.go

@@ -228,6 +228,15 @@ func VtcomboProcess(environment Environment, args *Config, mysql MySQLManager) (
 		fmt.Sprintf("--enable_system_settings=%t", args.EnableSystemSettings),
 	}...)
 
+	// If topo tablet refresh interval is not defined then we will give it value of 10s. Please note
+	// that the default value is 1 minute, but we are keeping it low to make vttestserver perform faster.
+	// Less value might result in high pressure on topo but for testing purpose that should not be a concern.
+	if args.VtgateTabletRefreshInterval <= 0 {
+		vt.ExtraArgs = append(vt.ExtraArgs, fmt.Sprintf("--tablet_refresh_interval=%v", 10*time.Second))
+	} else {
+		vt.ExtraArgs = append(vt.ExtraArgs, fmt.Sprintf("--tablet_refresh_interval=%v", args.VtgateTabletRefreshInterval))
+	}
+
 	vt.ExtraArgs = append(vt.ExtraArgs, QueryServerArgs...)
 	vt.ExtraArgs = append(vt.ExtraArgs, environment.VtcomboArguments()...)