Hi guys, I start a client node with the parameter "node.data" turned off,
why is the data directory automatically created in my working directory?
Can I avoid creating the "data" directory since it's a pure client node?
Hi guys, I start a client node with the parameter "node.data" turned off, why is the data directory automatically created in my working directory? Can I avoid creating the "data" directory since it's a pure client node?
Using a transport client would be slightly slower due to the two-hops,
that's why I prefer a client node. I guess a pure client node just need
some cluster info in memory, I don't see a reason why client need to save
cluster info to disk.
Now I have a interesting question, if I start multiple client nodes
(multiple VMs with the same working directory), they will write to the same
temporary data directory, will the data directory be messed up?
On Monday, February 2, 2015 at 10:31:34 PM UTC+8, David Pilato wrote:
As it's a node, it's storing some cluster info in it.
You could use a Transport Client which will act as a pure Client without
joining the cluster.
--
David
Twitter : @dadoonet / @elasticsearchfr / @scrutmydocs
Le 2 févr. 2015 à 15:08, Daniel Woo <daniel...@gmail.com <javascript:>> a
écrit :
Hi guys, I start a client node with the parameter "node.data" turned off,
why is the data directory automatically created in my working directory?
Can I avoid creating the "data" directory since it's a pure client node?
Using a transport client would be slightly slower due to the two-hops, that's why I prefer a client node. I guess a pure client node just need some cluster info in memory, I don't see a reason why client need to save cluster info to disk.
Now I have a interesting question, if I start multiple client nodes (multiple VMs with the same working directory), they will write to the same temporary data directory, will the data directory be messed up?
On Monday, February 2, 2015 at 10:31:34 PM UTC+8, David Pilato wrote:
As it's a node, it's storing some cluster info in it.
You could use a Transport Client which will act as a pure Client without joining the cluster.
--
David
Twitter : @dadoonet / @elasticsearchfr / @scrutmydocs
Le 2 févr. 2015 à 15:08, Daniel Woo <daniel...@gmail.com <javascript:>> a écrit :
Hi guys, I start a client node with the parameter "node.data" turned off, why is the data directory automatically created in my working directory? Can I avoid creating the "data" directory since it's a pure client node?
Cluster state must be persisted, that is the reason why NodeClient needs to
write data to files.
It is a myth that TransportClient is slower because of two-hop. You will
not notice the difference in reality because the total time of an ES
operation takes usually longer than the extra hop. And, for the overhead of
managing cluster state, TransportClient does not need to care about that.
Using a transport client would be slightly slower due to the two-hops,
that's why I prefer a client node. I guess a pure client node just need
some cluster info in memory, I don't see a reason why client need to save
cluster info to disk.
Now I have a interesting question, if I start multiple client nodes
(multiple VMs with the same working directory), they will write to the same
temporary data directory, will the data directory be messed up?
On Monday, February 2, 2015 at 10:31:34 PM UTC+8, David Pilato wrote:
As it's a node, it's storing some cluster info in it.
You could use a Transport Client which will act as a pure Client without
joining the cluster.
--
David
Twitter : @dadoonet / @elasticsearchfr / @scrutmydocs
Hi guys, I start a client node with the parameter "node.data" turned off,
why is the data directory automatically created in my working directory?
Can I avoid creating the "data" directory since it's a pure client node?
Using a transport client would be slightly slower due to the two-hops,
that's why I prefer a client node. I guess a pure client node just need
some cluster info in memory, I don't see a reason why client need to save
cluster info to disk.
Now I have a interesting question, if I start multiple client nodes
(multiple VMs with the same working directory), they will write to the same
temporary data directory, will the data directory be messed up?
On Monday, February 2, 2015 at 10:31:34 PM UTC+8, David Pilato wrote:
As it's a node, it's storing some cluster info in it.
You could use a Transport Client which will act as a pure Client without
joining the cluster.
--
David
Twitter : @dadoonet / @elasticsearchfr / @scrutmydocs
Hi guys, I start a client node with the parameter "node.data" turned off,
why is the data directory automatically created in my working directory?
Can I avoid creating the "data" directory since it's a pure client node?
Yes, I guess the extra hop is not a big deal compared to the disk IO ops on
the data node. I kind of prefer transport client now, the data directory
really drives me crazy
Cluster state must be persisted, that is the reason why NodeClient needs
to write data to files.
It is a myth that TransportClient is slower because of two-hop. You will
not notice the difference in reality because the total time of an ES
operation takes usually longer than the extra hop. And, for the overhead of
managing cluster state, TransportClient does not need to care about that.
Using a transport client would be slightly slower due to the two-hops,
that's why I prefer a client node. I guess a pure client node just need
some cluster info in memory, I don't see a reason why client need to save
cluster info to disk.
Now I have a interesting question, if I start multiple client nodes
(multiple VMs with the same working directory), they will write to the same
temporary data directory, will the data directory be messed up?
On Monday, February 2, 2015 at 10:31:34 PM UTC+8, David Pilato wrote:
As it's a node, it's storing some cluster info in it.
You could use a Transport Client which will act as a pure Client without
joining the cluster.
--
David
Twitter : @dadoonet / @elasticsearchfr / @scrutmydocs
Hi guys, I start a client node with the parameter "node.data" turned
off, why is the data directory automatically created in my working
directory? Can I avoid creating the "data" directory since it's a pure
client node?
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.