Обсуждение: Huge Pages - setting the right value
Hi, I'm currently testing performance with and without huge pages. Documentation says that in order to estimate the number of huge pages needed one should check the postmaster's VmPeak value. I wonder if it's only postmaster memory usage what's matters? Or I could get better estimation from the most memory intensive postgres process - not necessarly postmaster? I'm using following command to check it: for i in $(ps -ef | grep postgres|awk '{print $2}'); do grep ^VmPeak /proc/${i}/status|awk '{print $2}' >> log; done; sort -n -r log | head -1 I'm asking because some other process takes 606788kB while postmaster only 280444kB. -- View this message in context: http://www.postgresql-archive.org/Huge-Pages-setting-the-right-value-tp5952972.html Sent from the PostgreSQL - general mailing list archive at Nabble.com.
W dniu 2017-03-30 11:45:55 użytkownik pinker <pinker@onet.eu> napisał: > Hi, > I'm currently testing performance with and without huge pages. Documentation > says that in order to estimate the number of huge pages needed one should > check the postmaster's VmPeak value. I wonder if it's only postmaster memory > usage what's matters? Or I could get better estimation from the most memory > intensive postgres process - not necessarly postmaster? I'm using following > command to check it: > for i in $(ps -ef | grep postgres|awk '{print $2}'); do grep ^VmPeak > /proc/${i}/status|awk '{print $2}' >> log; done; sort -n -r log | head -1 > > I'm asking because some other process takes 606788kB while postmaster only > 280444kB. > > > > -- > View this message in context: http://www.postgresql-archive.org/Huge-Pages-setting-the-right-value-tp5952972.html > Sent from the PostgreSQL - general mailing list archive at Nabble.com. > > > -- > Sent via pgsql-general mailing list (pgsql-general@postgresql.org) > To make changes to your subscription: > http://www.postgresql.org/mailpref/pgsql-general > or maybe sum of all processes? I assume that memory allocated by postmaster means shared buffers, so if one wants to huge pages beeing used for sortingas well then should set some bigger number of huge pages in the kernel? Is it a right assumption?