Observation
When you are observing high heap and memory consumption and it grows rapidly over a number of days to its max value. As soon as it reaches the max value, the Hivemq service gets restarted
...
Steadily growing memory consumption of brokers over multiple days until nodes ultimately go out of memory. Once OOM the broker is restarted.
Reason
s3Client was not closed correctly in HiveMQS3Client, which resulted lead to a memory leak affecting the extension.
This is a known issue and addressed in the S3 Cluster Discovery Extension 4.1.1
Affected versions
This affects HiveMQ S3 Cluster Discovery Extension version 4.1.0 version.
Solution/Workaround
A fix is included in S3 Cluster Discovery Extension 4.1.1 and later.