kafka role: systemd unit 'failed', port 9092 not listening

[yakari007]

Summary

Deployed a VM with the kafka service role attached (default/empty config

• no version/node_id/heap_size override submitted), on the Ubuntu
  24.x "with PlusClouds Service" image. After SSH login succeeded, systemctl is-active kafka reported failed, and port 9092 (the default client port
  for the single combined broker+controller node this role describes)
  refused TCP connections outright. This role was given an extended 450s
  verification budget (vs the usual 300s, since it's a heavier JVM
  download+start) and still never came up.

Repro

POST /leo/iaas/virtual-machines/create-wizard with
service_roles: {"kafka": {"config": {}}} against the Ubuntu 24.x "with
PlusClouds Service" image, then after boot (waited up to 450s):

systemctl is-active kafka   # -> failed
exec 3<>/dev/tcp/127.0.0.1/9092   # -> Connection refused

The VM was deployed and torn down automatically as part of the test run,
so the live VM/journalctl -u kafka output is gone - filed from the
test's captured output, not a live inspection.

Found via leo4.tester's iaas-service-roles automated test scenario.

[yakari007]

Investigated but did not find a reproducible code-level bug, so no change made yet.

Downloaded the actual Kafka 4.3.1 release tarball from downloads.apache.org (confirmed still hosted, matches the role's URL exactly), installed a real JDK 21 (matching the role's openjdk-21-jre-headless prerequisite), and ran the actual kafka-storage.sh format + kafka-server-start.sh sequence with the exact server.properties content the role renders. It formatted and started cleanly — KRaft broker came up, bound :9092, no crash. Kafka's JVM heap (heap_size default 1G) plus its own overhead is fairly heavy for a minimal test VM, so if this keeps reproducing, I'd suspect resource pressure (OOM/slow start) on the test image over a config bug — worth checking actual available RAM on the VM and/or journalctl -u kafka from a live instance.