# monitoring limitations

Some monitoring data is missing or hard to come by.

## Does TiKV measure io size somewhere?

We currently (2020-09-23) don't have IO size metrics. The rocksdb compaction flow metrics can be an approximation, but it doesn't include WAL write IO size and read IO size. TiKV Summary > Cluster > MBps is the size of actual read/write data size from RocksDB. It does not directly related to IO size (MBps includes those cached reads and is after compression. it also ignore IO read amp of RocksDB). EBS also get cached by page cache, so even if TiKV have IO metrics, it doesn't necessary match the EBS console.