Same query executes 2 times in a row
I'm experiencing a strange issue that I can't explain. Sometimes when an INSERT ... SELECT query gets executed from our ETL tool (Pentaho) I can see in v_monitor.query_requests that it for some reason tried to be executed 2 times in a row. First time memory_acquired_mb is usually very large (in this case success=False in v_monitor.query_requests) and the second time memory_acquired_mb is often significantly smaller (in this case success=True in v_monitor.query_requests).
My first guess was that engine tries to execute the query in memory, fails to do so and proceeds with the execution on disk. However, in this case I would expect to see events like GROUP_BY_SPILLED or JOIN_SPILLED, but I don't see these events.
Could there be other reasons for this behaviour?
I can't share an SQL query, because this issue appears for many queries we have in our ETL process - therefore I don't think it's query-specific. What I can say is that we're on Vertica 9.3.1 and we have a cluster of 3 nodes (if this information is useful for this question).