If sqm_cache_enable and cmd_direct_replicate are on, and sqm_cmd_cache_size and sqm_max_cmd_in_block are set to nonzero values, you can use several counters to monitor replication performance, as the Executor and Distributor threads interact with the parsed data.
| Counter | Description |
|---|---|
| RACmdsDirectRepSend | The number of commands sent from the Executor thread associated with parsed data. |
| DISTCmdsDirectRepRecv | Number of commands received by Distributor that have parsed data associated with the statement directly from Executor, and where the parsing processing can be skipped. |
| SQMNoDirectReplicateInCache | The number of commands that have parsed data sent from the Executor thread, but the parsed data cannot be sent further along the replication pathway towards the Distributor because the command cache exceeds sqm_cmd_cache_size |
| SQMNoDirectReplicateInSQMCache | The number of commands that have parsed data sent from the Executor thread, but the parsed data cannot be sent further along the replication pathway towards the Distributor because these commands were overwritten in SQM cache before being read |
| SQMNoDirectReplicateInBlock | The number of commands that have parsed data sent from the Executor thread, but the parsed data cannot be sent further along the replication pathway towards the Distributor because the number of parsed data entries for the current SQM block exceeds sqm_max_cmd_in_block |