On 30-Aug-06, at 7:35 AM, Willo van der Merwe wrote:
> Luke Lonergan wrote:
>>> Currently the load looks like this:
>>> Cpu0 : 96.8% us, 1.9% sy, 0.0% ni, 0.3% id, 0.0% wa, 0.0%
>>> hi, 1.0% si
>>> Cpu1 : 97.8% us, 1.6% sy, 0.0% ni, 0.3% id, 0.0% wa, 0.0%
>>> hi, 0.3% si
>>> Cpu2 : 96.8% us, 2.6% sy, 0.0% ni, 0.3% id, 0.0% wa, 0.0%
>>> hi, 0.3% si
>>> Cpu3 : 96.2% us, 3.2% sy, 0.0% ni, 0.3% id, 0.0% wa, 0.0%
>>> hi, 0.3% si
>>>
>>
>> All four CPUs are hammered busy - check "top" and look for runaway
>> processes.
>>
>> - Luke
>>
>>
>>
> Yes, the first 463 process are all postgres. In the meanwhile I've
> done:
> Dropped max_connections from 500 to 250 and
> Upped shared_buffers = 50000
With 4G of memory you can push shared buffers to double that.
effective_cache should be 3/4 of available memory.
Can you also check vmstat 1 for high context switches during this
query, high being over 100k
Dave
>
> Without any apparent effect.
>
> ---------------------------(end of
> broadcast)---------------------------
> TIP 9: In versions below 8.0, the planner will ignore your desire to
> choose an index scan if your joining column's datatypes do not
> match
>