Details
-
Bug
-
Status: Closed (View Workflow)
-
Critical
-
Resolution: Fixed
-
None
-
None
-
None
-
None
Description
Customer is hitting this error when running a UMTS query thats self joining 5 times, with an exponential output the more joins are added.
Error:
FATAL ERROR: ExeMgr has allocated too much memory! Percent allocation-96, allowed-95. ExeMgr is restarting |
The goal of this ticket is to do whatever it takes to enable this workload in production
Notes:
Command wc -l out.txt Peak Ram
|
time mariadb mariadb_actixone_owner < sub2.sql > out.txt 61,276 |
real 0m27.297s
|
time mariadb mariadb_actixone_owner -qs < h0.sql > out.txt 83,165 ~46GB (primproc 42) |
real 0m31.205s
|
time mariadb mariadb_actixone_owner -qs < h1.sql > out.txt 209,272 ~46GB (primproc 42) |
real 0m45.975s
|
time mariadb mariadb_actixone_owner -qs < h2.sql > out.txt 1,185,876 ~48GB (primproc 44) |
real 1m11.682s
|
time mariadb mariadb_actixone_owner -qs < h3.sql > out.txt 11,796,689 ~80GB (primproc 75) |
real 5m27.624s
|
time mariadb mariadb_actixone_owner -qs < h4.sql > out.txt 158,300,163 250->293GB (primproc 282) |
real 63m49.800s
|
time mariadb mariadb_actixone_owner -qs < h5.sql > out.txt (w/ configuration change)
|
real 1073m2.118s 2,516,858,066 (1.2TB) |
Attachments
Issue Links
- relates to
-
MCOL-5797 New memory consumption accounting mechanism to prevent OOM events-related crashes.
-
- Closed
-