Hi -- I'll be looking to migrate elasticsearch off our application servers,
soon, onto its own server or servers.
I was wondering if there's any considered best practice for doing this?
Would one instance on one server be fine, or should I run at least two
servers, with the data replicated?
And, if I did run more than one, would that be best setup as a
master/child, or load balanced between two or more equal partners?
& if there's any documentation suggesting how to handle any of this around,
I'd be really grateful for a pointer to it, so's I can get reading.
Cheers,
Doug.
You probably want to run at least two servers, but it depends on your
requirements and if you can deal with a loss of service. Elasticsearch can
do master/slave, but by default runs hot/hot and is really the way to go
99.9% of the time.
You should connect your new nodes to your existing cluster and drain the
data off the current instances by using the allocation APIs or by bumping
your replica counts, so a replica is pushed to every node (auto expand
replicas).
Best Regards,
Paul
On Monday, January 28, 2013 2:26:40 AM UTC-7, biot023 wrote:
Hi -- I'll be looking to migrate elasticsearch off our application
servers, soon, onto its own server or servers.
I was wondering if there's any considered best practice for doing this?
Would one instance on one server be fine, or should I run at least two
servers, with the data replicated?
And, if I did run more than one, would that be best setup as a
master/child, or load balanced between two or more equal partners?
& if there's any documentation suggesting how to handle any of this
around, I'd be really grateful for a pointer to it, so's I can get reading.
Cheers,
Doug.
--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
Brilliant, cheers for that.
How would I identify my cluster from my app if I were running elasticsearch
on multiple servers?
I use the HTTP API, which obviously uses an IP address to find the service.
I currently have an ES instance running on each app server, but that's
becoming unscalable.
Cheers,
Doug.
On 29 January 2013 05:21, ppearcy ppearcy@gmail.com wrote:
You probably want to run at least two servers, but it depends on your
requirements and if you can deal with a loss of service. Elasticsearch can
do master/slave, but by default runs hot/hot and is really the way to go
99.9% of the time.You should connect your new nodes to your existing cluster and drain the
data off the current instances by using the allocation APIs or by bumping
your replica counts, so a replica is pushed to every node (auto expand
replicas).Best Regards,
PaulOn Monday, January 28, 2013 2:26:40 AM UTC-7, biot023 wrote:
Hi -- I'll be looking to migrate elasticsearch off our application
servers, soon, onto its own server or servers.
I was wondering if there's any considered best practice for doing this?
Would one instance on one server be fine, or should I run at least two
servers, with the data replicated?
And, if I did run more than one, would that be best setup as a
master/child, or load balanced between two or more equal partners?
& if there's any documentation suggesting how to handle any of this
around, I'd be really grateful for a pointer to it, so's I can get reading.
Cheers,
Doug.--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.