From 5d646d06fcc32d3939ff490837d39cbd2de5deed Mon Sep 17 00:00:00 2001
From: nickgooding <73336146+nickgooding@users.noreply.github.com>
Date: Wed, 10 Apr 2024 16:45:25 +0100
Subject: [PATCH] fix: ensure each broker is preferred leader for a partition

The end-to-end monitoring service already includes a process to validate
that each broker is the preferred leader for at least one partition, but
this validation was not run for brokers that were not the preferred
leader for any partitions.

For example, this would mean an additional partition would never be
created after the initial creation of the management topic in the case
of another broker being added to the cluster.

This fix could be a breaking change for users without the required
permissions to create partitions or alter partition assignments if they
have inadvertently relied on this feature not working as documented.
---
 e2e/topic.go | 1 +
 1 file changed, 1 insertion(+)

diff --git a/e2e/topic.go b/e2e/topic.go
index 98ec82d..c1c3b15 100644
--- a/e2e/topic.go
+++ b/e2e/topic.go
@@ -169,6 +169,7 @@ func (s *Service) calculatePartitionReassignments(meta *kmsg.MetadataResponse) (
 	// We want to ensure that each brokerID leads at least one partition permanently. Hence let's iterate over brokers.
 	preferredLeaderPartitionsBrokerID := make(map[int32][]kmsg.MetadataResponseTopicPartition)
 	for _, broker := range brokerByID {
+		preferredLeaderPartitionsBrokerID[broker.NodeID] = make([]kmsg.MetadataResponseTopicPartition, 0)
 		for _, partition := range topicMeta.Partitions {
 			// PreferredLeader = BrokerID of the brokerID that is the desired leader. Regardless who the current leader is
 			preferredLeader := partition.Replicas[0]