You want to shut down one instance at a time.
Each of your shards is on two nodes (the original on one and the
replica on the other).
If you shut down two nodes, you may take out both copies of a shard
before it has a chance to regenerate.
On Fri, Jun 24, 2011 at 10:57 PM, Karl Baum firstname.lastname@example.org wrote:
Interesting, in my case I have 20 shards but only one replica. Is that a
replica per shard? If so, then can I safely shut down 20 instance?
Thanks for your help!
Sent from my iPhone
On Jun 24, 2011, at 9:50 AM, Shay Banon email@example.com
Thats not a problem, those shards will get rebalanced into the remaining
nodes. When you scale down, make sure to shutdown nodes one at a time, and
wait for a green status between nodes being shutdown (we, you can shutdown
nodes up to the replica count of the index without the chance of missing
shards, note missing, and not loosing data, the data can still be recovered
if you start the node back up).
On Friday, June 24, 2011 at 4:48 PM, Karl Baum wrote:
In an attempt to index 200 million documents as quickly as possible, i
scaled my elasticsearch cluster up to 31 instances and my index consists of
20 shards and 1 replica. Everything is going well but my question is can I
scale the cluster down once i am done indexing? What happens if i drop down
to let's say 5-10 elasticsearch instances within the cluster?