Every container can state two amounts of CPU and memory. The request is what the scheduler reserves: a Pod lands only on a node whose free requests cover it, and a Pod whose request fits nowhere stays Pending with Insufficient cpu or Insufficient memory. The limit is the most the container may use. Above its CPU limit it is slowed down; above its memory limit it is killed (OOMKilled).
Units matter. CPU is in cores or thousandths (500m is half a core). Memory uses Mi and Gi (powers of two) or M and G (powers of ten), which are not the same amounts. A request above its limit is rejected, and a limit with no request makes the request equal to the limit.
Requests and limits set a Pod's QoS class, which decides who is evicted first when a node runs short: BestEffort (nothing set), Burstable, then Guaranteed (requests equal limits for CPU and memory in every container). kubectl top shows actual usage.
Set, diagnose a Pod that doesn't fit, and measure.