Replica path


(dkoum) #1

Hi,

I have elasticsearch set up on one ec2 vm with ebs storage, with 5
shards and 0 replicas.

I am trying to increase index availability beyond the ebs levels
without using more servers.

My first thought was to set replicas=1,attach a second ebs drive to my
server and configure elasticsearch to store the replicas in the second
drive. Unfortunately I cannot find any config option doing that. I
guess this is not how es replication works. Or am I missing something?

This leaves me with the option to introduce downtime to the service so
that I can shut down ES to flush data to disk, copy index to s3 and
start again.

Any other options?

thx in advance,
Dionysis


(Shay Banon) #2

Why do you need to shutdown the node in order to do backups? Simply backup
EBS to s3, if you can, send a disable flush to the translog using the index
update settings API before the backup, and then enable it.

On Thu, May 31, 2012 at 4:20 PM, dkoum kumdio@gmail.com wrote:

Hi,

I have elasticsearch set up on one ec2 vm with ebs storage, with 5
shards and 0 replicas.

I am trying to increase index availability beyond the ebs levels
without using more servers.

My first thought was to set replicas=1,attach a second ebs drive to my
server and configure elasticsearch to store the replicas in the second
drive. Unfortunately I cannot find any config option doing that. I
guess this is not how es replication works. Or am I missing something?

This leaves me with the option to introduce downtime to the service so
that I can shut down ES to flush data to disk, copy index to s3 and
start again.

Any other options?

thx in advance,
Dionysis


(dkoum) #3

thanks for answering.
I had totally missed the disable-flush functionality, I'm checking it right
now.
I just wanted to make sure I don't backup an inconsistent index due to an
ongoing index merge or other index operations.

On Sun, Jun 3, 2012 at 12:26 PM, Shay Banon kimchy@gmail.com wrote:

Why do you need to shutdown the node in order to do backups? Simply backup
EBS to s3, if you can, send a disable flush to the translog using the index
update settings API before the backup, and then enable it.

On Thu, May 31, 2012 at 4:20 PM, dkoum kumdio@gmail.com wrote:

Hi,

I have elasticsearch set up on one ec2 vm with ebs storage, with 5
shards and 0 replicas.

I am trying to increase index availability beyond the ebs levels
without using more servers.

My first thought was to set replicas=1,attach a second ebs drive to my
server and configure elasticsearch to store the replicas in the second
drive. Unfortunately I cannot find any config option doing that. I
guess this is not how es replication works. Or am I missing something?

This leaves me with the option to introduce downtime to the service so
that I can shut down ES to flush data to disk, copy index to s3 and
start again.

Any other options?

thx in advance,
Dionysis


(system) #4