You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently the extension supports ZooKeeper mode (using @ZooKeeperCluster as the constraint annotation) and KRaft in combined mode (using @KraftCluster). It would be good to also support KRaft using separate controllers.
This could be expressed in the API in several ways. Perhaps the simplest is just to add a combinedMode attribute to @KraftCluster.
Although the API is simple the knock-on consequences are not.
KafkaClusterConfig has a kraftMode boolean flag that is perhaps better treated as a 3 element MetadataMode enum (of ZK, Kraft combined and Kraft separate).
KafkaClusterConfig has a getBrokerConfigs() that would now be more clearly called getNodeConfigs(), since not all nodes are brokers.
ConfigHolder has getBrokerNum() that would now be more clearly called getNodeId(), again because not all nodes are brokers.
KafkaCluster.getNumOfBrokers() could remain a count of the brokers (not of all the nodes), but it might also be useful to have a count of all nodes.
The test coverage, which already has to cope with (ZK, combined) × (in-VM, containers) would need to also cover the separate controller cases.
Currently the extension supports ZooKeeper mode (using
@ZooKeeperCluster
as the constraint annotation) and KRaft in combined mode (using@KraftCluster
). It would be good to also support KRaft using separate controllers.This could be expressed in the API in several ways. Perhaps the simplest is just to add a
combinedMode
attribute to@KraftCluster
.Although the API is simple the knock-on consequences are not.
KafkaClusterConfig
has akraftMode
boolean flag that is perhaps better treated as a 3 elementMetadataMode
enum (of ZK, Kraft combined and Kraft separate).KafkaClusterConfig
has agetBrokerConfigs()
that would now be more clearly calledgetNodeConfigs()
, since not all nodes are brokers.ConfigHolder
hasgetBrokerNum()
that would now be more clearly calledgetNodeId()
, again because not all nodes are brokers.KafkaCluster.getNumOfBrokers()
could remain a count of the brokers (not of all the nodes), but it might also be useful to have a count of all nodes.See discussion in #174 which led to this issue.
The text was updated successfully, but these errors were encountered: