ok, running too many connections without recycling has an overhead. no wonder pgbouncer is all over as a sidecar :)
In case anyone is interested, this blog is a great read. Analyzing the Limits of Connection Scalability in Postgres - Microsoft Tech Community <https://techcommunity.microsoft.com/t5/azure-database-for-postgresql/analyzing-the-limits-of-connection-scalability-in-postgres/ba-p/1757266> <https://techcommunity.microsoft.com/t5/azure-database-for-postgresql/analyzing-the-limits-of-connection-scalability-in-postgres/ba-p/1757266> On Sun, 30 May 2021 at 20:19, Vijaykumar Jain < vijaykumarjain.git...@gmail.com> wrote: > I have a two dumb questions. > > 1) > I know the max_connections value change requires a restart. > > I also read a thread, which says why it is the case, assuming it still > holds true. > > Jean Arnaud <Jean(dot)Arnaud(at)inrialpes(dot)fr> writes: > > I'm looking for a way to change the "max_connections" parameter without > > restarting the PostGreSQL database. > > There is none. That's one of the parameters that determines shared > memory array sizes, and we can't change those on-the-fly. > > regards, tom lane > > Does that mean, if I set max_connections to 1000 users, but only intend to > use 10 at max (via conn limit per role settings), what would be the perf > difference compared to max_connections = 100 , but still using only 10 at > max. > > 2) > can i for test purposes.compile src with a xid max value to 2^16 etc ? > like with uint16 > > Basically, I was trying to simulate xid wraparound, by creating an open > transaction and an inactive replication slot, and then running a huge > parallel loops to bump txid_current() to see what happens at 2B crossover. > it takes forever on my slow laptop. > I do not see these in test scenarios, or could not find any. > > as always, ignore if not relevant, or already discussed. > I am preparing for interviews, hence asking myself dumb questions :) > > -- > Thanks, > Vijay > Mumbai, India > -- Thanks, Vijay Mumbai, India