at the moment i reinstall all Datanodes of one Cluster with new OS.
So we has 9 Datnodes.
4 Datanodes with 8 TB * 12 Disks
2 Datanodes with 4TB * 24 Disks
2 Datanodes with 6 TB * 12 Disks
So should we Partion the 8 TB Disks to 4 TB * 24 DiskPartions ?
Or should we leave it as it is ?
What is the best Solution ?
Hi @wernermarcel ,
You need at least 3 data nodes to give you fault tolerance in your architecture as HDFS triple replicates the data across the infrastructure. Using more and smaller disks will give you better throughput and better performance.
Here is a recent Cloudera blog that I recommend for you: