they are acting as a coodinator. But using master nodes is the wrong approach. If you have dedicated master nodes, try not to route any user traffic to those nodes, but use the data nodes instead.
Hey, really clear.
Anyway, since my scenario is write/update oriented, is there any useful tip (regarding role configuration)?
I saw that using Coordinating only-role I am not able to relieve DATA resources consumption and I am already working on application's side to change some logical mechanisms.
If it is write oriented, just send the data to the data nodes and you should be good in the beginning. I would not go with a coordinating node, unless the additional operational complexity is justified.
Hope that helps. Not sure that clears up all your questions?
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.