On 4/22/2016 1:30 AM, raghu vineel wrote:
I have a 4 core CPU for postgres and I have submitted 6 queries parallely in 6 different sessions. But I am seeing only 1 query per CPU and the remainig queries are queued. I am not sure why is it happening. Do you have any special scheduler for postgres? I can see that CPU has been shared with other non postgres processes but postgres processes are not sharing any CPU within a core. The following is the process states.
lms 2868 2209 98 07:20 ? 00:02:30 postgres: lms controller [local] SELECT
lms 2869 2209 98 07:20 ? 00:02:30 postgres: lms controller [local] SELECT lms 2870 2209 0 07:20 ? 00:00:00 postgres: lms controller [local] SELECT
lms 2871 2209 0 07:20 ? 00:00:00 postgres: lms controller [local] SELECT
lms 2872 2209 0 07:20 ? 00:00:00 postgres: lms controller [local] SELECT
lms 2873 2209 98 07:20 ? 00:02:30 postgres: lms controller [local] SELECTIf you see above, three Select queries have been queued. They are acquiring CPU only when there is a core without any postgres query running on it.
I know the postgres limitation of one query using only one core. But is there any limitation of one CPU running only one query at a time?
no such limitation. any chance there's locking involved here? take a look at pg_locks and pg_stat_activity the next time you run such a test.
--
john r pierce, recycling bits in santa cruz
