which services can bear two failure on nodes?
a.node managerb.cldbc.zookeeperd.resource managere.file server
Recommend you to check out Services Please! Planning and Installing Services on Your Nodes. It outlines how each service works and which one to install.
Thanks Cathy Liu for the link but I am bit confused about the resources manager as they are also two in a cluster one is working and other is on stand by so would it be also answer. As zookeeper is the answer but other than that confused in between CLDB or resource Manager.
Looks like the question is which services can survive two failures. If you only have two ResourceManagers (one active, one standby), and have two failures, there are none running after that, so ResourceManager is no longer available.
ZooKeeper can only survive two failures if you start with 5 ZooKeepers installed. ZooKeeper needs to maintain a quorum, or majority - so if you start with only 3 ZooKeepers, you've lost more than half if two fail (so you no longer have a quorum).
I want answer other than Zookeeper. Is there any other service that can bear 2 failure?
Sure - though you will have a performance impact.
If NodeManager fails on a node, you will not be able to process any data on that node for YARN applications - but NodeManagers on other nodes will continue to work. Suppose you have 5 nodes with NodeManager. If two of them fail, you've lost 40% of your ability to process data, so performance will be impacted - but things can still run.
If you have 3 or more CLDB nodes, two can fail (though if you start with 3 and 2 fail, the last remaining CLDB might go into "read only" mode while it waits for another CLDB).
Same with fileserver - though if the fileserver goes down on a node it can cause other services on that node to fail.
So, yes - you can lose other services on nodes and still run. But it will impact the overall performance of the cluster.
Retrieving data ...