Open skonto opened 7 years ago
I believe that the TM does download the Hadoop configuration files from the JM, via the JM's artifact server. The JM itself is provided the configuration files in the normal way, via the Mesos fetcher. @skonto are you seeing a problem related to this?
Shouldnt core-site, hdfs-site files be downloaded to the taksmanagers as well provided a docker image is available for them (do we have one?), for example to use then the file in hdfs client code? If the files are updated should we restart the flink job? Same concept applies to Spark: https://docs.mesosphere.com/service-docs/spark/hdfs/