Details
-
New Feature
-
Resolution: Unresolved
-
Major
-
7.0.2, Neo
-
None
-
DOC-2022-S12
-
1
Description
Feature Context - Conventionally the Search service re-assigns the index partitions across nodes during a rebalance operation. And the partitions which get freshly reassigned would get rebuild from scratch in the newly alloted nodes. This partition rebuild process was both time and resource consuming.
Hence Search introduced file transfer based rebalance operation which essentially moves the index partition files from the former partition node to the new node. This proved to be a faster and resource friendly approach to rebalance Search service.
-For the file-based rebalance feature to work, all the nodes in the cluster have to be supporting this feature. (ie on 7.0.2 at least)
-The file transfer-based rebalance feature is available under a feature flag in release 7.0.2. Meaning, the user has to explicitly enable the feature using the manager options endpoint as shown below. On 7.0.2, the customer has to trigger the below REST endpoint to enable the feature explicitly prior to starting the rebalance.
| curl -XPUT -H "Content-type:application/json" http://<Administrator>:<pwd>@<node>:8094/api/managerOptions -d '{"disableFileTransferRebalance": "false"}' |
-This(file transfer based) would become the default rebalance behavior in release 7.1.0. (Neo).
-This is an Enterprise only feature, and it moves the partition files across nodes (as against rebuilding it from scratch) to make the rebalance faster.
-This feature would be functional if the customer is upgrading from a CE version to an EE version of the feature supported server software. (ie, 7.0.2 onwards)
-In case of any unforeseen errors, the rebalance would fallback to fresh rebuild of partitions based DCP as earlier.
-Using the above endpoint, users can disable this feature too.