Local Gateway: What's the difference between gateway.recover_from_nodes and gateway.recover_from_master_nodes?


(Pablo Borges) #1

It's unclear to me. The docs didn't help much. :slight_smile:


(Shay Banon) #2

Yea, I am working on a complete high level docs of the gateway concepts, to
tie things together. I plan to explain all of that there.

In general, a (yet undocumented) feature of elasticsearch is the ability to
start specific master nodes (node.master=true, node.data=false) and data
nodes (node.master=false, node.data=true). Cluster level state recovery
(indices created, mappings and so on) only matter for master nodes count.
Thats why that setting is there.

Data recovery (actual index and transaction log) differs a bit between
shared gateway and local gateway. With local gateway, there is
a quorum based recovery, and once a quorum of shard + replica is found for
each shard, then the index is recovered.

-shay.banon

On Fri, Oct 1, 2010 at 5:45 PM, Pablo Borges pablort@gmail.com wrote:

It's unclear to me. The docs didn't help much. :slight_smile:


(system) #3