We are trying to setup a tribe node that connects to a remote single master multi-node cluster. The constraints we have right now is that the tribe node has connectivty to the master node only because of issues like firewall configuration, VPN set up and other issues. Since a tribe node creates a node client and tries to connect to all the nodes in the cluster, there are transport exceptions because of the connectivity issues with the cluster data nodes but we observed that
Elasticsearch reads(search/gets) from the tribe node are working fine.
Writes to the indices on the cluster mostly work fine but I have seen random failures.
The question is it okay to have this setup (tribe node connectivity to only the master node of the cluster) as our problem domain involves almost all reads or do you see any other potential fatal issues?
For example, is this the same thing as sniffing, as mentioned here?
"...can you try and remove sniffing please? I am afraid that it will make
the transport client point to the tribe node plus all of the different
nodes in the two clusters, which would explain your mixed results."
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.