Unfortunately I’ve seen the same thing in a few projects. The JIT seems a better fit for OLAP queries but the default cost settings are too aggressive, causing JIT to kick in for smaller OLTP-type queries where it’s not a win. I have a project where I get to benchmark each Postgres release every year, comparing both the old vs new Postgres and JIT vs no JIT in the new version, and I’ve never seen an improvement from JIT although the gap has narrowed in recent years. If there was a straightforward way to optimize the query optimizer params I’d consider doing it for this workload but barring that I’ve just had JIT off.
Unfortunately I’ve seen the same thing in a few projects. The JIT seems a better fit for OLAP queries but the default cost settings are too aggressive, causing JIT to kick in for smaller OLTP-type queries where it’s not a win. I have a project where I get to benchmark each Postgres release every year, comparing both the old vs new Postgres and JIT vs no JIT in the new version, and I’ve never seen an improvement from JIT although the gap has narrowed in recent years. If there was a straightforward way to optimize the query optimizer params I’d consider doing it for this workload but barring that I’ve just had JIT off.