On Tue, Sep 6, 2016 at 1:18 PM, Steve Atkins <steve@blighty.com> wrote:
>
>> On Sep 6, 2016, at 12:08 PM, Scott Marlowe <scott.marlowe@gmail.com> wrote:
>>
>> On Fri, Sep 2, 2016 at 9:38 PM, Pradeep <pgundala@avineonindia.com> wrote:
>>>
>>> max_connections = 100
>>> shared_buffers = 512MB
>>> effective_cache_size = 24GB
>>> work_mem = 110100kB
>>
>> This is WAY too high for work_mem. Work_mem is how much memory a
>> single sort can grab at once. Each query may run > 1 sort, and you
>> could have 100 queries running at once.
>>
>> This setting is 110GB. That's about 109.9GB too high for safety. When
>> things go wrong with this too big, they go very wrong, sending the
>> machine into a swap storm from which it may not return.
>
> It's an oddly spelled 110MB, which doesn't seem unreasonable.
oh yeah. still kind biggish but not as big as I had thought.