fix(tasks): avoid substring match when deriving node_id from rpk output

[ashishpatelkaseya]

Problem

start-redpanda.yml derives node_id using grep on the broker IP:

rpk cluster status {{ redpanda_rpk_opts }} | sed 's/*//g' | grep {{ hostvars[inventory_hostname].private_ip }} | awk '{ print $1;}'

grep matches substrings, so an IP like 172.31.241.1 also matches 172.31.241.153, returning multiple IDs and breaking subsequent logic.

Reproduction

1. Have two brokers with similar IP prefixes:
   • 172.31.241.1
   • 172.31.241.153
2. Run the task on the host 172.31.241.1.
3. The command outputs two IDs (e.g., 6 and 11) instead of one.

Expected

Match the host column exactly and return a single broker ID for the current host.

Proposed fix

Use awk to match column 2 exactly and strip the * from the ID:

- name: Establish node id
  ansible.builtin.shell:
    cmd: >-
      rpk cluster status {{ redpanda_rpk_opts }} |
      awk '$2=={{ hostvars[inventory_hostname].private_ip | quote }} { gsub(/\*/,"",$1); print $1 }'
  register: node_id_out
  changed_when: false

Alternative (more robust, requires jq)

Parse JSON output instead of text columns:

rpk cluster status {{ redpanda_rpk_opts }} -o json \
| jq -r '.brokers[] | select(.host=={{ hostvars[inventory_hostname].private_ip | tojson }}) | .id'

Impact

Prevents incorrect node_id detection in environments where broker IPs share prefixes, improving reliability of maintenance/decommission tasks that depend on node_id.

Reference

roles/redpanda_broker/tasks/start-redpanda.yml (around the node_id derivation)

[hcoyote]

We'd need to do this without jq. We can't guarantee that's installed on a system.