Merge pull request #18 from segmentio/yolken-usability-improvements4

Misc. usability improvements

cmd/topicctl/subcmd/apply.go

@@ -18,22 +18,26 @@ import (
 )
 
 var applyCmd = &cobra.Command{
-	Use:   "apply [topic configs]",
-	Short: "apply one or more topic configs",
-	Args:  cobra.MinimumNArgs(1),
-	RunE:  applyRun,
+	Use:     "apply [topic configs]",
+	Short:   "apply one or more topic configs",
+	Args:    cobra.MinimumNArgs(1),
+	PreRunE: applyPreRun,
+	RunE:    applyRun,
 }
 
 type applyCmdConfig struct {
-	brokersToRemove            []int
-	brokerThrottleMBsOverride  int
-	clusterConfig              string
-	dryRun                     bool
-	partitionBatchSizeOverride int
-	pathPrefix                 string
-	rebalance                  bool
-	skipConfirm                bool
-	sleepLoopTime              time.Duration
+	brokersToRemove              []int
+	brokerThrottleMBsOverride    int
+	clusterConfig                string
+	dryRun                       bool
+	partitionBatchSizeOverride   int
+	pathPrefix                   string
+	rebalance                    bool
+	retentionDropStepDurationStr string
+	skipConfirm                  bool
+	sleepLoopDuration            time.Duration
+
+	retentionDropStepDuration time.Duration
 }
 
 var applyConfig applyCmdConfig
@@ -69,17 +73,23 @@ func init() {
 		0,
 		"Partition batch size override",
 	)
+	applyCmd.Flags().StringVar(
+		&applyConfig.pathPrefix,
+		"path-prefix",
+		os.Getenv("TOPICCTL_APPLY_PATH_PREFIX"),
+		"Prefix for topic config paths",
+	)
 	applyCmd.Flags().BoolVar(
 		&applyConfig.rebalance,
 		"rebalance",
 		false,
 		"Explicitly rebalance broker partition assignments",
 	)
 	applyCmd.Flags().StringVar(
-		&applyConfig.pathPrefix,
-		"path-prefix",
-		os.Getenv("TOPICCTL_APPLY_PATH_PREFIX"),
-		"Prefix for topic config paths",
+		&applyConfig.retentionDropStepDurationStr,
+		"retention-drop-step-duration",
+		"",
+		"Amount of time to use for retention drop steps",
 	)
 	applyCmd.Flags().BoolVar(
 		&applyConfig.skipConfirm,
@@ -88,15 +98,30 @@ func init() {
 		"Skip confirmation prompts during apply process",
 	)
 	applyCmd.Flags().DurationVar(
-		&applyConfig.sleepLoopTime,
-		"sleep-loop-time",
+		&applyConfig.sleepLoopDuration,
+		"sleep-loop-duration",
 		10*time.Second,
 		"Amount of time to wait between partition checks",
 	)
 
 	RootCmd.AddCommand(applyCmd)
 }
 
+func applyPreRun(cmd *cobra.Command, args []string) error {
+	if applyConfig.retentionDropStepDurationStr != "" {
+		var err error
+		applyConfig.retentionDropStepDuration, err = time.ParseDuration(
+			applyConfig.retentionDropStepDurationStr,
+		)
+
+		if err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
 func applyRun(cmd *cobra.Command, args []string) error {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -189,8 +214,9 @@ func applyTopic(
 		DryRun:                     applyConfig.dryRun,
 		PartitionBatchSizeOverride: applyConfig.partitionBatchSizeOverride,
 		Rebalance:                  applyConfig.rebalance,
+		RetentionDropStepDuration:  applyConfig.retentionDropStepDuration,
 		SkipConfirm:                applyConfig.skipConfirm,
-		SleepLoopTime:              applyConfig.sleepLoopTime,
+		SleepLoopDuration:          applyConfig.sleepLoopDuration,
 		TopicConfig:                topicConfig,
 	}
 

cmd/topicctl/subcmd/get.go

@@ -153,7 +153,7 @@ func getRun(cmd *cobra.Command, args []string) error {
 			return fmt.Errorf("Must provide topic and groupID as additional positional arguments")
 		}
 
-		return cliRunner.GetMemberLags(ctx, args[1], args[2])
+		return cliRunner.GetMemberLags(ctx, args[1], args[2], getConfig.full)
 	case "members":
 		if len(args) != 2 {
 			return fmt.Errorf("Must provide group ID as second positional argument")

pkg/apply/apply.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"path/filepath"
 	"reflect"
+	"strings"
 	"time"
 
 	"github.com/hashicorp/go-multierror"
@@ -29,8 +30,9 @@ type TopicApplierConfig struct {
 	DryRun                     bool
 	PartitionBatchSizeOverride int
 	Rebalance                  bool
+	RetentionDropStepDuration  time.Duration
 	SkipConfirm                bool
-	SleepLoopTime              time.Duration
+	SleepLoopDuration          time.Duration
 	TopicConfig                config.TopicConfig
 }
 
@@ -168,7 +170,7 @@ func (t *TopicApplier) applyNewTopic(ctx context.Context) error {
 	}
 
 	// Just do a short sleep to ensure that zk is updated before we check
-	if err := interruptableSleep(ctx, t.config.SleepLoopTime/5); err != nil {
+	if err := interruptableSleep(ctx, t.config.SleepLoopDuration/5); err != nil {
 		return err
 	}
 
@@ -354,6 +356,25 @@ func (t *TopicApplier) updateSettings(
 		return err
 	}
 
+	var retentionDropStepDuration time.Duration
+	if t.config.RetentionDropStepDuration != 0 {
+		retentionDropStepDuration = t.config.RetentionDropStepDuration
+	} else {
+		var err error
+		retentionDropStepDuration, err = t.config.ClusterConfig.GetDefaultRetentionDropStepDuration()
+		if err != nil {
+			return err
+		}
+	}
+
+	reduced, err := topicSettings.ReduceRetentionDrop(
+		topicInfo.Config,
+		retentionDropStepDuration,
+	)
+	if err != nil {
+		return err
+	}
+
 	if len(diffKeys) > 0 {
 		diffsTable, err := FormatSettingsDiff(topicSettings, topicInfo.Config, diffKeys)
 		if err != nil {
@@ -366,6 +387,18 @@ func (t *TopicApplier) updateSettings(
 			diffsTable,
 		)
 
+		if reduced {
+			log.Infof(
+				strings.Join(
+					[]string{
+						"Note: Retention drop has been reduced to minimize cluster disruption.",
+						"Re-run apply afterwards to keep dropping retention to configured value or run with --retention-drop-step-duration=0 to not do gradual step-down.",
+					},
+					" ",
+				),
+			)
+		}
+
 		if t.config.DryRun {
 			log.Infof("Skipping update because dryRun is set to true")
 			return nil
@@ -887,7 +920,7 @@ func (t *TopicApplier) updatePartitionsIteration(
 		return err
 	}
 
-	checkTimer := time.NewTicker(t.config.SleepLoopTime)
+	checkTimer := time.NewTicker(t.config.SleepLoopDuration)
 	defer checkTimer.Stop()
 
 	log.Info("Sleeping then entering check loop")
@@ -945,7 +978,7 @@ outerLoop:
 				len(assignmentsToUpdate),
 				admin.FormatTopicPartitions(notReady, t.brokers),
 			)
-			log.Infof("Sleeping for %s", t.config.SleepLoopTime.String())
+			log.Infof("Sleeping for %s", t.config.SleepLoopDuration.String())
 		case <-ctx.Done():
 			return ctx.Err()
 		}
@@ -1181,7 +1214,7 @@ func (t *TopicApplier) updateLeadersIteration(
 		return err
 	}
 
-	checkTimer := time.NewTicker(t.config.SleepLoopTime)
+	checkTimer := time.NewTicker(t.config.SleepLoopDuration)
 	defer checkTimer.Stop()
 
 	log.Info("Sleeping then entering check loop")
@@ -1212,7 +1245,7 @@ outerLoop:
 				admin.FormatTopicPartitions(wrongLeaders, t.brokers),
 			)
 
-			log.Infof("Sleeping for %s", t.config.SleepLoopTime.String())
+			log.Infof("Sleeping for %s", t.config.SleepLoopDuration.String())
 		case <-ctx.Done():
 			return ctx.Err()
 		}

pkg/apply/apply_test.go

@@ -45,6 +45,8 @@ func TestApplyBasicUpdates(t *testing.T) {
 	}
 
 	applier := testApplier(ctx, t, topicConfig)
+	applier.config.RetentionDropStepDuration = 50 * time.Minute
+
 	assert.Equal(t, 3, applier.maxBatchSize)
 	assert.Equal(t, int64(2000000), applier.throttleBytes)
 
@@ -62,13 +64,15 @@ func TestApplyBasicUpdates(t *testing.T) {
 	assert.Equal(t, "compact", topicInfo.Config["cleanup.policy"])
 
 	// Update retention and settings
-	applier.topicConfig.Spec.RetentionMinutes = 501
+	applier.topicConfig.Spec.RetentionMinutes = 400
 	applier.topicConfig.Spec.Settings["cleanup.policy"] = "delete"
 	err = applier.Apply(ctx)
 	require.Nil(t, err)
 	topicInfo, err = applier.adminClient.GetTopic(ctx, topicName, true)
 	require.Nil(t, err)
-	assert.Equal(t, "30060000", topicInfo.Config[admin.RetentionKey])
+
+	// Dropped to only 450 because of retention reduction
+	assert.Equal(t, "27000000", topicInfo.Config[admin.RetentionKey])
 	assert.Equal(t, "delete", topicInfo.Config["cleanup.policy"])
 
 	// Updating replication factor not allowed
@@ -868,7 +872,7 @@ func TestApplyOverrides(t *testing.T) {
 			TopicConfig:                topicConfig,
 			DryRun:                     false,
 			SkipConfirm:                true,
-			SleepLoopTime:              500 * time.Millisecond,
+			SleepLoopDuration:          500 * time.Millisecond,
 			PartitionBatchSizeOverride: 8,
 		},
 	)
@@ -906,11 +910,11 @@ func testApplier(
 		ctx,
 		adminClient,
 		TopicApplierConfig{
-			ClusterConfig: clusterConfig,
-			TopicConfig:   topicConfig,
-			DryRun:        false,
-			SkipConfirm:   true,
-			SleepLoopTime: 500 * time.Millisecond,
+			ClusterConfig:     clusterConfig,
+			TopicConfig:       topicConfig,
+			DryRun:            false,
+			SkipConfirm:       true,
+			SleepLoopDuration: 500 * time.Millisecond,
 		},
 	)
 	require.Nil(t, err)

pkg/check/check_test.go

@@ -58,11 +58,11 @@ func TestCheck(t *testing.T) {
 		ctx,
 		adminClient,
 		apply.TopicApplierConfig{
-			ClusterConfig: clusterConfig,
-			TopicConfig:   topicConfig,
-			DryRun:        false,
-			SkipConfirm:   true,
-			SleepLoopTime: 500 * time.Millisecond,
+			ClusterConfig:     clusterConfig,
+			TopicConfig:       topicConfig,
+			DryRun:            false,
+			SkipConfirm:       true,
+			SleepLoopDuration: 500 * time.Millisecond,
 		},
 	)
 	require.Nil(t, err)

pkg/cli/cli.go

@@ -352,7 +352,12 @@ func (c *CLIRunner) GetGroupMembers(ctx context.Context, groupID string, full bo
 
 // GetMemberLags fetches and prints a summary of the consumer group lag for each partition
 // in a single topic.
-func (c *CLIRunner) GetMemberLags(ctx context.Context, topic string, groupID string) error {
+func (c *CLIRunner) GetMemberLags(
+	ctx context.Context,
+	topic string,
+	groupID string,
+	full bool,
+) error {
 	c.startSpinner()
 
 	// Check that topic exists before getting offsets; otherwise, the topic get
@@ -370,7 +375,7 @@ func (c *CLIRunner) GetMemberLags(ctx context.Context, topic string, groupID str
 		return err
 	}
 
-	c.printer("Group member lags:\n%s", groups.FormatMemberLags(memberLags))
+	c.printer("Group member lags:\n%s", groups.FormatMemberLags(memberLags, full))
 	return nil
 }
 

pkg/cli/repl.go

@@ -262,7 +262,7 @@ func (r *Repl) executor(in string) {
 				log.Errorf("Error: %+v", err)
 				return
 			}
-			if err := r.cliRunner.GetMemberLags(ctx, words[2], words[3]); err != nil {
+			if err := r.cliRunner.GetMemberLags(ctx, words[2], words[3], false); err != nil {
 				log.Errorf("Error: %+v", err)
 				return
 			}

pkg/config/cluster.go

@@ -3,6 +3,8 @@ package config
 import (
 	"context"
 	"errors"
+	"fmt"
+	"time"
 
 	"github.com/aws/aws-sdk-go/aws/session"
 	"github.com/hashicorp/go-multierror"
@@ -68,6 +70,10 @@ type ClusterSpec struct {
 	// DefaultThrottleMB is the default broker throttle used for migrations in this
 	// cluster. If unset, then a reasonable default is used instead.
 	DefaultThrottleMB int64 `json:"defaultThrottleMB"`
+
+	// DefaultRetentionDropStepDuration is the default amount of time that retention drops will be
+	// limited by. If unset, no retention drop limiting will be applied.
+	DefaultRetentionDropStepDurationStr string `json:"defaultRetentionDropStepDuration"`
 }
 
 // Validate evaluates whether the cluster config is valid.
@@ -98,9 +104,25 @@ func (c ClusterConfig) Validate() error {
 		multierror.Append(err, errors.New("MajorVersion must be v0.10 or v2"))
 	}
 
+	_, parseErr := c.GetDefaultRetentionDropStepDuration()
+	if parseErr != nil {
+		err = multierror.Append(
+			err,
+			fmt.Errorf("Error parsing retention drop step retention: %+v", parseErr),
+		)
+	}
+
 	return err
 }
 
+func (c ClusterConfig) GetDefaultRetentionDropStepDuration() (time.Duration, error) {
+	if c.Spec.DefaultRetentionDropStepDurationStr == "" {
+		return 0, nil
+	}
+
+	return time.ParseDuration(c.Spec.DefaultRetentionDropStepDurationStr)
+}
+
 // NewAdminClient returns a new admin client using the parameters in the current cluster config.
 func (c ClusterConfig) NewAdminClient(
 	ctx context.Context,