# Dml Operations Cause Oom (Continue...)

+ Question:
  + Coprocessor CPU usage spikes —> CPU runs out (full) —> duration of cluster goes up —> large amount of data is deleted —> loads into memory —> full TiDB of memory —> causing OOM

------------------------------------------------------------
When we have a large amount of data that needs to be cleaned up, we may start a too large transaction exception, or OOM may be caused by insufficient memory on the TiDB node. In this case, although there is an index on the imp_date field, TiDB tends to request the TiKV coprocessor to read a large amount of data because the index scan range is too large, whether the optimizer chooses index scan or table scan.

There are two implications for large read data request TiKV nodes:
- Coprocessor CPU usage of TiKV nodes spikes and CPU resources are full, resulting in high cluster duration
- When data is deleted, it is loaded into memory, filling up the memory resources of TiDB, resulting in OOM