ZK future?
What is everyone's thoughts of ZK dependency in apigee-edge? Looks like Kafka is moving way from ZK, see link below. I realize Apigee-Edge is a monolith product - ZK, C* (& other services), and we can only follow along.
https://www.confluent.io/blog/removing-zookeeper-dependency-in-kafka/
How does this effect me? Numerous times during a RMP add/remove I've witnessed stale config data in ZK.
Example: rapidly add 6 RMPs in a pod, and wait for the data-sync to occur. On a production-system with 1000s proxies and developer-apps, this can take
Issue1 - MGMT commands only tell partial picture. Only when executing "zk-tree", then I can see the full picture - which RMP is stored in ZK
After running these two commands, you can compare the UUID to expected values. zk-tree # This will dump entire ZK dataset zk-tree | grep -E Router zk-tree | grep -E Message If I find stale data, I will execute a "rmr" command against the stale UUID