On Tue, Aug 27, 2019 at 12:06 PM Holtgrewe, Manuel <manuel.holtgrewe@xxxxxxxxxxx> wrote: > iotop tells me that walwriter does not go beyond ~35MB/s so maybe this is the culprit? Is there a way to tune walwriter I/O performance? As far as I know, walwriter is there to help background processes, so in the case it cannot keep up with WALs the backends will write on their own. If my understanding is correct, I don't think that could be the bootleneck. I've seen you have checkpoints at 4h, that's quite huge to me. Do you have any hint that checkpoints are happening too frequently? Any chance you can turn fsync off (only if this is a testbed)? Also this <https://www.postgresql.org/list/pgsql-performance/> could be a better place to ask for help. Luca
