Hello,   We are currently in the process of upgrading our cluster from CDH 5.14 to 6.3.4. We are assessing how this task could be done without causing downtime on the cluster.   We have already performed the steps on the "Full Cluster Restart" approach which lead to downtime while the CDH components were stopped and in the queue to be upgraded.   We were wondering if by applying the "Manual Upgrade" approach this downtime could be avoided, but had some doubts, as the documentation is not very clear on how to perform the steps manually and seems to point to the "Manual Upgrade" as a way to revert a failed upgrade.   The main issue was that from the order the steps are laid out it assumes the services are currently down, which is not the case if we are starting the upgrade process from scratch. We are running Kafka, Zookeeper, Flume Forwarder, Flume Avro, HDFS and Sentry. Can anyone that has done the Manual Upgrade provide some guidelines and advice on how to proceed? Thank you in advance! Best Regards, Diogo ... View more

Hello, My team will be upgrading our cluster from 5.x to 6.x soon and we have some doubts regarding the consumer and producer compatibility after the cluster has been upgraded.   This upgrade will lead to a change on Kafka and Flume versions as shown below: Kafka 1.0.1 2.2.1 Flume 1.6.0 1.9.0   It was not very clear to me which should be the steps to perform on the consumers and producers to avoid them not being able to communicate with Kafka after the upgrade.   From the information I have gathered, it seems that the producers are forward compatible, meaning that initially it would not be required for them to perform any change, contrariwise the consumers must be updated for them to get information from Kafka after the upgrade.   To summarize: No changes required on the producer side; Code update required on the consumer side.   Are the assumptions above correct?   Thanks in advance! ... View more