> On May 12, 2023, at 9:52 PM, Tom Lane <tgl@xxxxxxxxxxxxx> wrote: > > Alex Balashov <abalashov@xxxxxxxxxxxxxxx> writes: >> Occasionally, about once every 2 months, we'll get a slow-motion implosion over a period of about 24 hours, where the resident memory size of some of the backends shoots up from a reasonable few hundred MB to several gigs -- and will keep growing, e.g. > > I'm a little suspicious that the excess memory consumption is a > consequence of JIT-compiled code not getting recycled promptly. > Do you have JIT enabled, and if so are you relying on it too much > to turn it off? > > I don't have any concrete idea of why JIT code management would work > fine and then stop working fine, but we do have other reports of > memory leaks from that. Hi Tom, thanks for your response. JIT is enabled, but I'm not familiar with this feature and not sure what we are using which would touch JIT somehow. Also, am I wrong to imagine that a memory leak would leak memory gradually over time, rather than to be set off at some arbitrary point by the same exact workload that has been visited upon the database for the previous 1-2 months hitherto? -- Alex -- Alex Balashov Principal Consultant Evariste Systems LLC Web: https://evaristesys.com Tel: +1-706-510-6800