I'm not entirely sure how discovery in ES works. Are there resources out there that detail it? How can I dynamically add a new node (on a different physical machine) to an existing cluster. I tried looking at some previous posts in this forum. I get the idea that one's supposed to list nodes in the config file and they discover each other. But that doesn't quite explain dynamic addition of nodes and load balancing.
On the same lines, is the master elected or configured? Does the master use more resources than data nodes? Because in the latter case, I would like to have a machine with more physical resources be the master.
On Sat, Aug 13, 2011 at 11:02 PM, optimusprime usmasood@gmail.com wrote:
I'm not entirely sure how discovery in ES works. Are there resources out
there that detail it? How can I dynamically add a new node (on a different
physical machine) to an existing cluster. I tried looking at some previous
posts in this forum. I get the idea that one's supposed to list nodes in
the
config file and they discover each other. But that doesn't quite explain
dynamic addition of nodes and load balancing.
On the same lines, is the master elected or configured? Does the master use
more resources than data nodes? Because in the latter case, I would like to
have a machine with more physical resources be the master.
Master is automatically elected, but it does not require more resources. You
can control which nodes will can become master nodes though.
So if I understand Zen unicast discovery right, we have elasticsearch
running on a set of nodes with fixed IPs (gossip routers). Any new node
talks to these nodes first to "register" itself as part of the cluster.
Client nodes can obtain the master node (or data nodes) from the cluster by
talking to one of these routers. Needless to say at least one of the gossip
routers must be up at all times, otherwise discovery fails. Can we
dynamically add more gossip routers or would that require us to shutdown the
node, update its config file and restart it?
I'm not entirely sure how discovery in ES works. Are there resources out
there that detail it? How can I dynamically add a new node (on a different
physical machine) to an existing cluster. I tried looking at some previous
posts in this forum. I get the idea that one's supposed to list nodes in
the
config file and they discover each other. But that doesn't quite explain
dynamic addition of nodes and load balancing.
On the same lines, is the master elected or configured? Does the master
use
more resources than data nodes? Because in the latter case, I would like
to
have a machine with more physical resources be the master.
Master is automatically elected, but it does not require more resources.
You can control which nodes will can become master nodes though.
Currently, you have to bring down a node in order to update its list of
"seed" IPs. Though you can add new machines to new nodes you bring into the
cluster.
So if I understand Zen unicast discovery right, we have elasticsearch
running on a set of nodes with fixed IPs (gossip routers). Any new node
talks to these nodes first to "register" itself as part of the cluster.
Client nodes can obtain the master node (or data nodes) from the cluster by
talking to one of these routers. Needless to say at least one of the gossip
routers must be up at all times, otherwise discovery fails. Can we
dynamically add more gossip routers or would that require us to shutdown the
node, update its config file and restart it?
I'm not entirely sure how discovery in ES works. Are there resources out
there that detail it? How can I dynamically add a new node (on a
different
physical machine) to an existing cluster. I tried looking at some
previous
posts in this forum. I get the idea that one's supposed to list nodes in
the
config file and they discover each other. But that doesn't quite explain
dynamic addition of nodes and load balancing.
On the same lines, is the master elected or configured? Does the master
use
more resources than data nodes? Because in the latter case, I would like
to
have a machine with more physical resources be the master.
Master is automatically elected, but it does not require more resources.
You can control which nodes will can become master nodes though.
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.