KafkaBrokerLoggingConfigurationDiff - Invalid logging level

[armisz]

Hello,

I have an issue setting the log configuration for the Cluster Operator on Strimzi 0.21.0 on OpenShift 3.11. Maybe it's a bug, maybe it's me, I have double checked everything but can't see any problem with my setup.

The error message is flooding my log on every reconciliation:
2021-01-27 15:29:48 WARN KafkaBrokerLoggingConfigurationDiff:252 - Invalid logging level: . Using WARN as a failover.

My log configuration is this:

sh-4.2$ cat custom-config/log4j2.properties 
name = COConfig
monitorInterval = 30

appender.console.type = Console
appender.console.name = STDOUT
appender.console.layout.type = PatternLayout
appender.console.layout.pattern = %d{yyyy-MM-dd HH:mm:ss} %-5p %c{1}:%L - %m%n

rootLogger.level = INFO
rootLogger.appenderRefs = stdout
rootLogger.appenderRef.console.ref = STDOUT
rootLogger.additivity = false

# Kafka AdminClient logging is a bit noisy at INFO level
logger.kafka.name = org.apache.kafka
logger.kafka.level = WARN
logger.kafka.additivity = false

# Zookeeper is very verbose even on INFO level -> We set it to WARN by default
logger.zookeepertrustmanager.name = org.apache.zookeeper
logger.zookeepertrustmanager.level = WARN
logger.zookeepertrustmanager.additivity = false

# Keeps separate level for Netty logging -> to not be changed by the root logger
logger.netty.name = io.netty
logger.netty.level = INFO
logger.netty.additivity = false

All log levels look good for me. Any hint is very much appreciated.

It would be nice if there was a hint to the location of the problem log level in future.

Thanks in advance
Michael

[scholzj]

That looks like the log4j2 properties for the Cluster Operator. Are you trying to use it with Kafka broker? Kafka is using Log4j 1 => so this would not work. Or what exactly are you trying to do?

[scholzj]

Can you also share the custom resources you deployed? From the operator log it should be Kafka CR called abc-strimzi-kafka in namespace abc-test. As I said, I do not think this is related to your cluster operator logging but to your Kafka cluster.

[armisz]

kafka-logging.yml.txt
kafka.yml.txt

I have expected that the log4j2 properties have only an effect on the Cluster Operator and not on the other managed Kafka and ZK instances.

[scholzj]

I do not think this is related to the log4j2.proepties file. It should be all about our log4j.proeprties file for the KAfka cluster. In particular it does not like the log4j.rootLogger=${kafka.root.logger.level}, CONSOLE from it. If you replace it with log4j.rootLogger=INFO, CONSOLE directly instead of using the variable the error should disappear.

@sknot-rh Is this expected? Can this be somehow fixed or improved?

[sknot-rh]

Let me investigate.

[sknot-rh]

Ok, so it is a bug. The configuration uses the kafka.root.logger.level before it is defined. Fixed in #4340