Merge pull request #23395 from redpanda-data/vbotbuildovich/backport-…

…22855-v24.1.x-593

[v24.1.x] rpk: supports triggering on-demand balancer

src/go/rpk/pkg/adminapi/api_partition.go

@@ -204,3 +204,8 @@ func (a *AdminAPI) TransferLeadership(ctx context.Context, fs afero.Fs, p *confi
  path := fmt.Sprintf("/v1/partitions/%s/%s/%d/transfer_leadership?target=%s", ns, topic, partition, target)
  return cl.sendOne(ctx, http.MethodPost, path, nil, nil, false)
 }
+
+// TriggerBalancer Trigger on-demand balancer.
+func (a *AdminAPI) TriggerBalancer(ctx context.Context) error {
+ return a.sendToLeader(ctx, http.MethodPost, "/v1/partitions/rebalance", nil, nil)
+}

src/go/rpk/pkg/cli/cluster/partitions/balancer_run.go

@@ -0,0 +1,67 @@
+// Copyright 2024 Redpanda Data, Inc.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.md
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0
+
+package partitions
+
+import (
+ "fmt"
+
+ "github.com/redpanda-data/redpanda/src/go/rpk/pkg/adminapi"
+ "github.com/redpanda-data/redpanda/src/go/rpk/pkg/config"
+ "github.com/redpanda-data/redpanda/src/go/rpk/pkg/out"
+ "github.com/spf13/afero"
+ "github.com/spf13/cobra"
+)
+
+func newTriggerBalancerCommand(fs afero.Fs, p *config.Params) *cobra.Command {
+ cmd := &cobra.Command{
+ Use: "balance",
+ Short: "Triggers on-demand partition balancing",
+ Long: `Triggers on-demand partition balancing.
+
+This command allows you to trigger on-demand partition balancing.
+
+With Redpanda Community Edition, the partition count on each broker
+can easily become uneven, which leads to data skewing. To distribute
+partitions across brokers, you can run this command to trigger
+on-demand partition balancing.
+
+With Redpanda Enterprise Edition, Continuous Data Balancing monitors
+broker and rack availability, as well as disk usage, to avoid topic
+hotspots. However, there are edge cases where users should manually
+trigger partition balancing (such as a node becoming unavailable for
+a prolonged time and rejoining the cluster thereafter). In such cases,
+you should run this command to trigger partition balancing manually.
+
+After you run this command, monitor the balancer progress using:
+
+ rpk cluster partitions balancer-status
+
+To see more detailed movement status, monitor the progress using:
+
+ rpk cluster partitions move-status
+`,
+
+ Args: cobra.ExactArgs(0),
+ Run: func(cmd *cobra.Command, _ []string) {
+ p, err := p.LoadVirtualProfile(fs)
+ out.MaybeDie(err, "rpk unable to load config: %v", err)
+ config.CheckExitCloudAdmin(p)
+
+ cl, err := adminapi.NewClient(fs, p)
+ out.MaybeDie(err, "unable to initialize admin client: %v", err)
+
+ err = cl.TriggerBalancer(cmd.Context())
+ out.MaybeDie(err, "failed to invoke on-demand partition balancing: %v", err)
+
+ fmt.Println("Successfully triggered on-demand partition balancing.\n\nPlease find the progress with 'rpk cluster partitions balancer-status'.")
+ },
+ }
+ return cmd
+}

src/go/rpk/pkg/cli/cluster/partitions/partitions.go

@@ -33,6 +33,7 @@ func NewPartitionsCommand(fs afero.Fs, p *config.Params) *cobra.Command {
  newPartitionEnableCommand(fs, p),
  newPartitionMovementsStatusCommand(fs, p),
  newTransferLeaderCommand(fs, p),
+ newTriggerBalancerCommand(fs, p),
  newUnsafeRecoveryCommand(fs, p),
  )
  return cmd