Full text search replication on CE

query

#1

Hi, for what I see replication is not allowed on CE. Is there a way to have the index replicated even with other name? I could not find where the index is located when created on full text search. Are the index for full text search located in only one node? When the node goes out are they automatically recreated in other index or is this a manual process? Can i create the same index with different name in multiple nodes?

Thanks


#2

Is there a way to have the index replicated even with other name?

Hmm, that might not quite work as intended…

Let’s say you create two full-text indexes A and B… and, they each have their own partitions underneath the hood (like A1, A2, …, A6… B1, B2 … B6)

What you’d like to have happen with an index replica (and this happens with the Enterprise Edition) is that partition A1 has a replica partition A1’ that’s assigned to a different node that wherever A1 is living (say, A1 lives on node X and A1’ lives on node Y). That way, if node X fails, then node Y (which has partition A1’) can take over rapidly.

But, the partitions B1, B2 … B6 don’t have any relationship (from the system’s perspective) to the partitions from index A.

I could not find where the index is located when created on full text search. Are the index for full text search located in only one node?

The index is partitioned into partitions and those partitions are assigned across the nodes in your cluster. Of course, if you only have a cluster of a single node, then they’ll all be assigned to that single node.

When the node goes out are they automatically recreated in other index or is this a manual process?

They underlying partitions will be automatically reassigned and “transferred” during cluster Rebalance and node Failover operations, so you don’t have to do anything manual.

Can i create the same index with different name in multiple nodes?

Yes, you can create multiple indexes using the exact same index definition. The system will think of them as separate independent indexes, though, so it doesn’t quite have that “replica” story that I think you’re looking for.

cheers,
steve


#3

@steve thanks for your feedback, Now I understand how it works. So a full text search index is distributed in the cluster, if a node goes down, rebalance takes place and migrate/recreate when need. Nothing to worry on my side. Global secondary index if a node goes down, index needs to be manually recreated. So big difference.

Thanks!


#4

“if a node goes down, rebalance takes place and migrate/recreate when need” - just to be clear here, the rebalance or the node fail over has to be triggered manually which in turn will take care of all the index/partition migrations/creations.

-sreekanth