database connection pooling for microservices, is it overkill or essential

hey folks, been thinkin bout how we handle database connections in our microservices. were seein some contention under heavy load, and im wonderin if implementin connection poolin at the service level is the right move. some argue it adds unnecessary complexity, others say its crucial for performance. whats your take? do you pool per service, or rely on the database itself? any painful lessons learned