Does this configuration (no multicast and only two unicast hosts) make
sense? What would happen if one of these fails?
At the same time I've created a client to index/search data in a ES
cluster, based on configuration it initiates a Node or a Transport
client, with a list of hosts also configured for the latter.
Hosting the client in 172.16.137.156, if I start a Node client, with
the cluster name correctly set, it can't discover the cluster
(MasterNotDiscoveredException), but using the Transport one, with the
list of hosts explicitly declared, it works ok, as expected.
So, how the Node client would find the master if no multicast nor list
of hosts was explicitly defined?
How did you configure the node client? It uses the same discovery as the
other nodes in the cluster. You need to configure it with the same settings
as the non client nodes for discovery.
Does this configuration (no multicast and only two unicast hosts) make
sense? What would happen if one of these fails?
At the same time I've created a client to index/search data in a ES
cluster, based on configuration it initiates a Node or a Transport
client, with a list of hosts also configured for the latter.
Hosting the client in 172.16.137.156, if I start a Node client, with
the cluster name correctly set, it can't discover the cluster
(MasterNotDiscoveredException), but using the Transport one, with the
list of hosts explicitly declared, it works ok, as expected.
So, how the Node client would find the master if no multicast nor list
of hosts was explicitly defined?
Oh, yes that makes sense. I just configured the cluster name for the
node client following the Java API client page, but forgot to set the
other cluster settings.
One last one, in which scenarios would you recommend to configure
cluster discovery using unicast over multicast? (thinking on server
failures, new added nodes, etc)
How did you configure the node client? It uses the same discovery as the
other nodes in the cluster. You need to configure it with the same settings
as the non client nodes for discovery.
Does this configuration (no multicast and only two unicast hosts) make
sense? What would happen if one of these fails?
At the same time I've created a client to index/search data in a ES
cluster, based on configuration it initiates a Node or a Transport
client, with a list of hosts also configured for the latter.
Hosting the client in 172.16.137.156, if I start a Node client, with
the cluster name correctly set, it can't discover the cluster
(MasterNotDiscoveredException), but using the Transport one, with the
list of hosts explicitly declared, it works ok, as expected.
So, how the Node client would find the master if no multicast nor list
of hosts was explicitly defined?
Unicast allows for simpler fine grained control over what talks to what.
Multicast allows for that as well, but you need to configure different
multicast IPs / ports. Both are good.
Oh, yes that makes sense. I just configured the cluster name for the
node client following the Java API client page, but forgot to set the
other cluster settings.
One last one, in which scenarios would you recommend to configure
cluster discovery using unicast over multicast? (thinking on server
failures, new added nodes, etc)
How did you configure the node client? It uses the same discovery as the
other nodes in the cluster. You need to configure it with the same
settings
as the non client nodes for discovery.
Does this configuration (no multicast and only two unicast hosts) make
sense? What would happen if one of these fails?
At the same time I've created a client to index/search data in a ES
cluster, based on configuration it initiates a Node or a Transport
client, with a list of hosts also configured for the latter.
Hosting the client in 172.16.137.156, if I start a Node client, with
the cluster name correctly set, it can't discover the cluster
(MasterNotDiscoveredException), but using the Transport one, with the
list of hosts explicitly declared, it works ok, as expected.
So, how the Node client would find the master if no multicast nor list
of hosts was explicitly defined?
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.