# Coprocessor（E）

+ Analysis conclusion:
  + A TiKV has a higher Unified Read Pool CPU than other TiKV nodes for nearly 1 hour from 21:20 to 22:20
  + The coprocessor wait duration of each TiKV is not high, and the time to get snapshots and build the Handler is very short, not a bottleneck
  + A TiKV takes a lot of time to select and index from a Coprocessor Handle duration of almost 1 hour from 21:20 to 22:20
+ Further confirm the direction:
  + Read hot phenomenon
------------------------------------------------------------
- The Unified Read Pool CPU utilization for each TiKV is not high, and the CPU is not a resource bottleneck. However, it is obvious that at the same time, the resource utilization of each TiKV is uneven, such as 18:20181 TiKV from 21:20 to 22:20 for nearly 1 hour from 21:20 to 22:20 Unified read pool CPU is higher than other TiKV nodes, and there is a suspicion that there is a read hot phenomenon to be investigated
- The coprocessor wait duration for each TiKV is not high, indicating that the request is scheduled, and the time to get the snapshot and build the Handler is very short, not a bottleneck
- The coprocessor handle duration takes a lot of time for select and index, but the corresponding time for each TiKV is high. At the same time, a TiKV will be higher, for example, 18:20181 TiKV takes almost 1 hour from 21:20 to 22:20. Other TiKV nodes suspected to have read hot spots to be checked