On Tue, Jul 9, 2019 at 4:44 AM Justin Pryzby wrote:
> [...many partial GroupAggregate/Sort repeated
> here for ~200 child tables...]
>
> This isn't the first time I've had to reduce work_mem on a parallel query to
> avoid OOM, but it seems unreasonable to be concerned
One of our instances was killed:
| Out of memory: Kill process 27670 (postmaster) score 381 or sacrifice child
| Killed process 27670 (postmaster) total-vm:13207796kB, anon-rss:5736kB,
file-rss:0kB, shmem-rss:12421696kB
The VM has: 32GB RAM, shared_buffers=12GB, work_mem=128MB
Running only pg11
