From mapd logs, we can see that its taking about 3-4 mins to get the catalog data and more minutes for other activities. We have single node with 4 p100 GPUs with 60 cpu cores. Following is snapshot of the log
I1221 00:37:20.778688 45929 MapDHandler.cpp:568] sql_execute :A1o2n4PoHEkJ7ApmVgGWIA4Aqi2dHXUH:query_str:drop table store_scan_dly;
I1221 00:37:20.778934 45929 MapDHandler.cpp:3256] passing query to legacy processor
I1221 00:40:24.928164 45929 FileMgr.cpp:173] Completed Reading table’s file metadata, Elasped time : 184129ms Epoch: 1035247 files read: 10130 table location: '/var/lib/mapd/data/mapd_data/table_3_12/'
I1221 00:42:32.940171 45929 Catalog.cpp:1525] Instantiating Fragmenter for table store_scan_dly_shard_#1 took 312141ms
I1221 00:52:47.649708 45929 Calcite.cpp:179] Time to updateMetadata 1 (ms)
I1221 00:53:10.694839 45929 FileMgr.cpp:173] Completed Reading table’s file metadata, Elasped time : 23044ms Epoch: 1035813 files read: 10135 table location: '/var/lib/mapd/data/mapd_data/table_3_13/'
I1221 00:55:16.718389 45929 Catalog.cpp:1525] Instantiating Fragmenter for table store_scan_dly_shard_#2 took 149068ms
Drop table is still in progress as of this posting…
There is nothing suspicious in ERROR/WARNING logs
We also saw similar issue when we try and truncate the table
Do anyone have an insight into the issue/behaviour?