Hello, hackers.
Today I was doing some aggregates over pg_stat_statements in order to
find types of queries consuming most of the CPU. Aggregates were made
on two pg_state_statement snapshots within 30 sec delay.
The sum(total_time) had the biggest value for a very frequent query
with about 10ms execution. I was thinking it is the biggest CPU
consumer.
But after reducing the frequency of queries a lot I was unable to see
any significant difference in server CPU usage...
So, looks like clock_gettime is not so accurate to measure real CPU
usage for some OLTP workloads. I suppose it is caused by the wall time
vs CPU time difference (IO, thread switch, etc).
But what do you think about adding cpu_time (by calling getrusage) to
pg_stat_statements? Seems it could be very useful for CPU profiling.
I am probably able to prepare the patch, but it is always better to
get some feedback on the idea first :)
Best regards,
Michail.