I have a large postgresql advanced server database (v11.2), 9+ Tb, OLTP system on RHEL 6 which injest 1M+ rows of inserts/updates per minute which becomes sluggish after about 10-20 minutes of connecting to injesting application reducing injestion to maybe 300K per minute. We have tried to modify vacuuming times, checkpoint times, and adjusting other parameters in the postgresql.conf file but always get the same result. Our recent attempt we modified some of the OS kernal paging which no luck. This database schema has 15K tables with about half paritions tables with parent tables. I have tried the same load with non-partitioned tables but still get the same sluggish outcome in about the same time frame. Any input would be welcome.
... View more