Re: DB load balancer

So the reads happen only at: 4 p.m. and 6 p.m. Looks more like batch jobs
then reads! Can you post the "Number of Reads" during these times too...

Deepak
"The greatness of a nation can be judged by the way its animals are treated
- Mahatma Gandhi"

+91 73500 12833
deicool(at)gmail(dot)com

LinkedIn: www.linkedin.com/in/deicool

"Plant a Tree, Go Green"

Make In India : http://www.makeinindia.com/home

On Tue, Aug 5, 2025 at 11:42 AM Siraj G <tosiraj(dot)g(at)gmail(dot)com> wrote:

> Hello Deepak
>
> CPU utilization is high, below is just a 1 day utilization graph. We have
> been having these more often between 3:30 and 5:30 in the afternoon. This
> is a daily pattern and during this the application response goes very slow.
>
> [image: image.png]
>
> Regards
> Siraj
>
> On Tue, Aug 5, 2025 at 10:51 AM Deepak Goel <deicool(at)gmail(dot)com> wrote:
>
>> Please share what you mean by slow (CPU usage of secondary over time,
>> response times of the transactions over time).
>>
>>
>> Deepak
>> "The greatness of a nation can be judged by the way its animals are
>> treated - Mahatma Gandhi"
>>
>> +91 73500 12833
>> deicool(at)gmail(dot)com
>>
>> LinkedIn: www.linkedin.com/in/deicool
>>
>> "Plant a Tree, Go Green"
>>
>> Make In India : http://www.makeinindia.com/home
>>
>>
>> On Tue, Aug 5, 2025 at 10:18 AM Siraj G <tosiraj(dot)g(at)gmail(dot)com> wrote:
>>
>>> Hello Experts!
>>>
>>> We have a multi tenant model DB with a Primary and a Secondary. There
>>> are 3x or 4x connections flowing to Secondary as the application is read
>>> intensive. Accordingly we have allocated more number of resources to the VM
>>> hosting the secondary/replica DB.
>>>
>>> Spec:-
>>> Primary: 48 CPUs, 48GB memory
>>> Secondary: 80 CPUs, 128GB memory
>>> PG version: 12.22 (we have already started the upgrade process)
>>> OS: Ubuntu
>>>
>>> The problem very frequently we are experiencing is that the CPU load on
>>> the SECONDARY is going very very high causing the app slowness. Also,
>>> frequently the postgres is crashing due to OOM.
>>>
>>> I was thinking of these (please give feedback):
>>> 1. Have a DB level balancer for the load balancing in the DB level,
>>> meaning primary will also receive the READ traffic
>>> 2. Add a second REPLICA and configure load balancer for the connection
>>> load balancing
>>>
>>> What do we do for load balancing at the DB level, in postgres?
>>>
>>> If there are any suggestions, please let me know.
>>>
>>> Regards
>>> Siraj
>>>
>>