Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
Description
Hi,
If we create a region with indexes and we insert million rows inside a full restart of the product is very long, the cluster come back quickly but the region is back after minutes (Ex : 16 minutes for a 30 nodes with 200 000 000 rows with 3 indexes) because indexes are rebuild at restart before servicing the region.
If we create and serve a non indexed region we can stop/start quickly, if we create indexes after creating/servicing a region with a lot of rows there is no service freeze.
Could it be possible to have a parameter that say to nodes we do not want anymore indexes on a specific region at restart, just keep their definition, and then index back again after region is serviced like when we create indexes ?
On our need it is not a big problem if indexes are not there at restart, it won't help but it could be acceptable, because in our solution we need to be UP 24/24 to serve DATA acquisition/writing, reads/selects are for few on demand consumption which can be long in case of DRP (The time indexes are back).