This problem has been affected by online use, but when the program reads data in batches, the scanner is prone to timeout or "java.io.IOException: Couldn't get scan data" exception. Can someone help answer and help optimize this question, thank you.
Because I have 12 storage disks, I have previously set up 12 maintenance manager threads, but found that the disk IO is too high, causing the disk to be fully loaded. It was later changed to 4 threads, which can alleviate it, but it still takes up a lot.The average read and write of each disk is above 70m/s.
# This is my tserver configuration
--fs_wal_dir=/data1/kudu/tserver
--fs_data_dirs=/data2/kudu/tserver,/data3/kudu/tserver,/data4/kudu/tserver,/data5/kudu/tserver,/data6/kudu/tserver,/data7/kudu/tserver,/data8/kudu/tserver,/data9/kudu/tserver,/data10/
kudu/tserver,/data11/kudu/tserver,/data12/kudu/tserver,/data13/kudu/tserver
--rpc_service_queue_length=30000
--scanner_ttl_ms=600000
--scanner_batch_size_rows=10000
--scanner_default_batch_size_bytes=10485760
--scanner_max_batch_size_bytes=83886080
--rpc_num_service_threads=32
--consensus_rpc_timeout_ms=600000
--consensus_max_batch_size_bytes=10485760
--maintenance_manager_num_threads=4
--block_cache_capacity_mb=5120
