A resource request tells the scheduler how much capacity a container needs. A resource limit bounds how much it may use at runtime. CPU values can use millicores, while memory quantities use units such as Mi. A Pod can remain Pending when its requests exceed every available node, even when the application would use less after startup. A low memory limit can instead allow scheduling and later cause OOMKilled. Treat those failures separately. These tasks set complete fictional resource budgets and retain a known replica count. Check each field rather than assuming a lower limit also lowers the scheduling reservation.