Sivaraman K.G wrote:
> Hi,
>
> Has any one have the statistics that how long the postmaster takes for
> graceful shutdown
> (when SIGTERM is send) ?
that would be EXTREMELY dependent on a large number of variables.
an idle instance with no pending writes on a reasonably fast modern 'nix system
should exit nearly instantly. If there's 50 open connections with
transactions-in-progress, and a lot of write buffers to flush, its gonna take
longer.
ok, here's a linux box with RHEL 3.5 X86_64, postgres 7.4.7 built for 32bit...
# time service postgresql stop
Stopping postgresql service: [ OK ]
real 0m2.100s
user 0m0.000s
sys 0m0.020s
so, 2 seconds on a opteron 2.4Ghz for pgsql 7.4.7
now, the 'service postgresql' command runs /etc/rc.d/init.d, which in turn is
using pg_ctl to stop rather than SIGTERM, but the net effect is the same.
here's a RHEL3.5 x86 (32bit) postgres 8.0.1 system..,
# time service postgresql stop
Stopping PostgreSQL: postmaster stopped
ok
real 0m1.086s
user 0m0.030s
sys 0m0.040s
1 second for a xeon 2.8Ghz w/ pgsql 8.0.1
