Github user kwmonroe commented on a diff in the pull request:
https://github.com/apache/bigtop/pull/400#discussion_r222415839
--- Diff: bigtop-packages/src/charm/kafka/layer-kafka/reactive/kafka.py ---
@@ -38,9 +40,10 @@ def configure_kafka(zk):
hookenv.status_set('maintenance', 'setting up kafka')
data_changed( # Prime data changed for network interface
'kafka.network_interface',
hookenv.config().get('network_interface'))
+ log_dir = unitdata.kv().get('kafka.storage.log_dir')
--- End diff --
Let's add another `data_changed('kafka.storage.log_dir', log_dir)` here.
This `configure_kafka` method is the primary way that kafka gets installed --
after that, we constantly check data_changed in `configure_kafka_zookeepers` to
trigger a reinstall.
If we *don't* do a `data_changed` with `log_dir` here, we'll get installed
and started, and then immediately reinstall on the first execution of
`configure_kafka_zookeepers`, because the check for
`data_changed('kafka.storage.log_dir', log_dir)` in that method will be new
(and hence, True).
---
