Adding another disk to an ES node

Hi,

Is there a way to change path.dir
(see http://www.elasticsearch.org/guide/reference/setup/dir-layout.html
) on a live cluster, say when one wants to add new disks to ES servers and
tell ES to start using them?

Also, once a new disk is added to an ES node (even if it requires ES node
restart), how will ES start using it? It sounds like it would immediately
switch writing to just that disk because ES would see that new disk has
more free space than the old one(s)? If so, does that mean it would start
writing to all disks only once the new one is equally full as the old
one(s)?

Thanks,
Otis

ELASTICSEARCH Performance Monitoring - http://sematext.com/spm/index.html

--

Is there a way to change path.dir (see
Elasticsearch Platform — Find real-time answers at scale | Elastic ) on a
live cluster, say when one wants to add new disks to ES servers and tell ES
to start using them?

I think the only way to do this is by doing rolling restart.

Also, once a new disk is added to an ES node (even if it requires ES node
restart), how will ES start using it? It sounds like it would immediately
switch writing to just that disk because ES would see that new disk has
more free space than the old one(s)?

Correct. All new files will be created on the disc with the most available
free space.

If so, does that mean it would start writing to all disks only once the
new one is equally full as the old one(s)?

Correct.

--

Hi Igor,

On Monday, January 21, 2013 2:57:18 PM UTC-5, Igor Motov wrote:

Is there a way to change path.dir (see

Elasticsearch Platform — Find real-time answers at scale | Elastic ) on
a live cluster, say when one wants to add new disks to ES servers and tell
ES to start using them?

I think the only way to do this is by doing rolling restart.

Why rolling if you need to add a disk to just 1 server? Then just that 1
server would need to be restarted, right?

Would it make sense to expose an API that tells ES "Hey, go look at the
config and reload it, I added something to it"?

Also, once a new disk is added to an ES node (even if it requires ES node

restart), how will ES start using it? It sounds like it would immediately
switch writing to just that disk because ES would see that new disk has
more free space than the old one(s)?

Correct. All new files will be created on the disc with the most available
free space.

If so, does that mean it would start writing to all disks only once the
new one is equally full as the old one(s)?

Correct.

I don't think I've ever seen anyone on this list mention using multiple
disks in this way.
Anything one should be aware of when this writing across all disks starts
happening?
Is there any parallelization involved that would actually improve write
throughput as more disks are added?

Thanks,
Otis

ELASTICSEARCH Performance Monitoring - Sematext Monitoring | Infrastructure Monitoring Service

--