I used MemSQL Studio’s resource usage feature to to profile our cluster. A query that was running spent >90% of its time on “miscellaneous internal waits”. Is there a way to find out what these actually are?
“miscellaneous internal wait” time is technically the unaccounted time.
In practice, the unaccounted time is likely to be dominated by “inherent” waits for events and a good rule of thumb is to assume it’s not an indication of a problem. For example, time spent waiting for work dispatched to another thread to complete. We often do such dispatching for read queries, so it’s normal to see output like:
The row with high misc wait time is reporting on a thread which was waiting for per-partition threads to do real work.