It would be useful in reproducing workloads to be able to schedule a large table scan that happens occasionally. Especially in mutli-table workloads.
The configuration string I'm thinking is:
threads=((count=20,reads=1),(count=5,updates=1,throttle=100),(count=1,scan=1000000,pause=300,table_name=test00003))
The new options are:
- scan Number of items to iterate over
- pause How many seconds to wait between scans
- table_name The name of the table (in multi-table workloads) that will be scanned
The scan operation would start at a random point in the specified tree (or the first/only tree if no explicit value is specified) and walk until it has either visited the count of values or reaches the end of the tree.
- is depended on by
-
SERVER-27700 WT secondary performance drops to near-zero with cache full
- Closed