Uncontrolled ExeMgr memory utilization (other than hash-join memory) ?

Build tested: 1.0.7-1

This issue is related to the 1TB test and issue ported in MCOL-558. Please refer to that ticket for additional details.

With totalUMMemory se at 50%, I noticed that ExeMgr's memory utilization went up to 70% before ColumnStore self-restarted due to swapping issue. Since I did not get the hash join memory exhausted error so I assume that we had not hit that 50% yet. That means ExeMgr used over 20% for processing other than hash operations. If the system did not get restarted, I don't know how high this memory utilization will go.

As far as I know, there is no parameter for controlling this memory utilization.

Is this memory utilization strictly based on dataset size?
Is there any mechanism built-in to cap/control ever increasing memory usage?

For this test, disk based join was not enabled. If you are think about that, please check the another ticket that I will be writing right after this. I will post the ticket number in a follow up comment.