I’ve used Ambari for several months now and minimal problems, but a new cluster that we just set up has had strange problems.
First, all would go well until the deployment was due to begin, then Ambari would throw this error:
{ “status” : 404, “message” : “org.apache.ambari.server.controller.spi.NoSuchParentResourceException: Parent Cluster resource doesn’t exist. Cluster not found, clusterName=staging. Cluster not found, clusterName=staging” }
After several ambari-server resets and running into the same problem, I was making a detailed record of everything when suddenly the trouble disappeared (I do not believe I did anything differently) and the deploy proceeded. I do not know if that’s relevant to my current issue, and I’ve been unable to reproduce it since.
Now the cluster is up and seeming to work, but if I attempt to go to the Hosts screen and filter by installed component, it will always show the number of boxes with the components installed, but the same usually wrong machines, which are also the first ones by hostname in alphabetical order.
For instance, let’s say in a simplified manner that
box1 = YARN Resource Manager
box2 = Spark History Server
box3 = Namenode, Zookeeper
box4 = Kafka Broker, Zookeeper
box5 = Kafka Broker, Zookeeper
if I filter for Kafka Brokers it will show me box1 and box2. If I filter for the Namenode, it will show me box1. Zookeeper will give me box1, box2, and box3. Other filters (like for memory, cores, or IP) work fine, it’s only the components that’s off.
How can I go about diagnosing this? I’m wondering if I should be filing a bug report.