From 914029e17bcb2f1540d2d361f3667d03d282db14 Mon Sep 17 00:00:00 2001 From: Staffan Olsson Date: Fri, 10 Nov 2017 09:50:44 +0100 Subject: [PATCH] The first two tasks under Partition Management in the Kafka book --- maintenance/README.md | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) create mode 100644 maintenance/README.md diff --git a/maintenance/README.md b/maintenance/README.md new file mode 100644 index 00000000..e25abec8 --- /dev/null +++ b/maintenance/README.md @@ -0,0 +1,22 @@ + +## Trigger leadership election + +This is one of the cases where this repo begs to differ from traditional Kafka setups. +In Kubernetes the restart of a pod, and subsequent start on a different node, should be a non-event. + +> ”when a broker is stopped and restarted, it does not resume leadership of any partitions automatically” + +_ -- Neha Narkhede, Gwen Shapira, and Todd Palino. ”Kafka: The Definitive Guide” _ + +Create the `preferred-replica-election-job.yml` resource, after deleting any previous one. + +## Changing a Partition's Replicas + +> ”From time to time, it may be necessary to change the replica assignments for a partition. Some examples of when this might be needed are: +> * If a topic’s partitions are not balanced across the cluster, causing uneven load on brokers +> * If a broker is taken offline and the partition is under-replicated +> * If a new broker is added and needs to receive a share of the cluster load” + +_ -- Neha Narkhede, Gwen Shapira, and Todd Palino. ”Kafka: The Definitive Guide” _ + +Use the `reassign-paritions-job.yml`, after editing `TOPICS` and `BROKERS`.