We're facing issues with loading data on the database.
With BULK COMMIT the operation is getting terminated after 2-3 hours informing not able to connect tp database.
Hence we're doing row lovel commits with 25 parallel threads (RAM=83G, CPU=14, shared_buffer=21G, work_mem = 15MB
maintenance_work_mem = 4275MB
effective_cache_size = 65GB)
We've 1 primary and 1 synchronous standby in our system.
With 25 parallel thread the dataload is happening fine for first 2 hours and we're able to see 10+ lakh records getting inserted nicely, and post that the data load activity is getting slowed down.with only 50k-1 lacs records inserting .
Can anyone help on this.
Without knowing much detail, My first guess is that you have a synchronous replica setup. Can that be disabled during your load and set up replication later ?