BufferIO and BufferMapping are different types of events than previous ones. Your server is probably suffering because of intensive IO operations. There can be big queries who are causing big read or write operations and/or a significant number of active connections. I would check it at the operating system level by using top command to see what is the percentage for IO Wait (wa), system (sy), and software interrupts (si).
Judging by PgBouncer configuration, I suspect that shared_buffers would be set to a huge number, proportional to the available memory. I would check this configuration too. You would want to check swap usage too.
Thank you for the information you provided about LWLocks. When I look at the type of these locks I see "BufferIO" and "BufferMapping" it really slows down the system. However, I have not yet fully understood what action I should take in this situation, especially why this situation is happening? and what should i do?
