[CI] Bump memory for k8s discovery e2e tests

[ruizhang0101]

Router resources 2G→4G; engine requestMemory 30Gi→40Gi across e2e values files.

[gemini-code-assist]

Code Review

This pull request increases the memory allocations across several configuration files, raising the serving engine memory request from 30Gi to 40Gi and the router memory requests and limits from 2G to 4G. The feedback advises using binary SI units (e.g., '4Gi' instead of '4G') for the router configurations to prevent potential Out-Of-Memory (OOM) issues and maintain consistency with the serving engine's memory specifications.

[gemini-code-assist]

In Kubernetes, it is standard practice to specify memory requests and limits using binary SI units (e.g., Gi) rather than decimal SI units (e.g., G). Since operating systems and runtimes measure memory in powers of two, using 4G (decimal) instead of 4Gi (binary) allocates approximately 7.3% less memory than expected, which can lead to unexpected Out-Of-Memory (OOM) kills. It is recommended to use 4Gi to ensure consistency with the serving engine's memory specification (40Gi).

      memory: "4Gi"
    limits:
      cpu: "1"
      memory: "4Gi"

[gemini-code-assist]

In Kubernetes, it is standard practice to specify memory requests and limits using binary SI units (e.g., Gi) rather than decimal SI units (e.g., G). Since operating systems and runtimes measure memory in powers of two, using 4G (decimal) instead of 4Gi (binary) allocates approximately 7.3% less memory than expected, which can lead to unexpected Out-Of-Memory (OOM) kills. It is recommended to use 4Gi to ensure consistency with the serving engine's memory specification (40Gi).

      memory: "4Gi"
    limits:
      cpu: "1"
      memory: "4Gi"

[gemini-code-assist]

In Kubernetes, it is standard practice to specify memory requests and limits using binary SI units (e.g., Gi) rather than decimal SI units (e.g., G). Since operating systems and runtimes measure memory in powers of two, using 4G (decimal) instead of 4Gi (binary) allocates approximately 7.3% less memory than expected, which can lead to unexpected Out-Of-Memory (OOM) kills. It is recommended to use 4Gi to ensure consistency with the serving engine's memory specification (40Gi).

      memory: "4Gi"
    limits:
      cpu: "1"
      memory: "4Gi"

[gemini-code-assist]

In Kubernetes, it is standard practice to specify memory requests and limits using binary SI units (e.g., Gi) rather than decimal SI units (e.g., G). Since operating systems and runtimes measure memory in powers of two, using 4G (decimal) instead of 4Gi (binary) allocates approximately 7.3% less memory than expected, which can lead to unexpected Out-Of-Memory (OOM) kills. It is recommended to use 4Gi to ensure consistency with the serving engine's memory specification (40Gi).

      memory: "4Gi"
    limits:
      cpu: "1"
      memory: "4Gi"