Correctness Checking of F5 Values

No source files or tests are identified. Start by reproducing the reported comparisons between tmctl snapshots, the bash endpoint, and the named Prometheus metrics using the stated five-second collection interval. Done means explaining whether the observed differences come from rounding, scaling, or incorrect collection, and identifying any required correction.

Hi, I was conducting an analysis into seeing how the metric values from AST compare with what we get directly from curling to the load balancer via the bash endpoint. Here is some info on our configuration for AST:

Collection Interval: 5 seconds

Enabled data-types:

• f5.cgnat
• f5.dns
• f5.dos
• f5.firewall
• f5.gtm
• f5.policy.firewall
• f5.policy.ip_intelligence
• f5.policy.nat
• f5.profile.dos

Comparison Between tmctl of "plane_cpu_stat" snapshot vs f5_plane_cpu_utilization_5s: generally a small difference is observed. Previous iterations showed larger discrepancies.

Comparison Between tmctl of "virtual_server_cpu_stat" snapshot vs f5_virtual_server_cpu_utilization_5: generally a small difference is observed (this is with the assumption that the decimal value is the percentage value but divided by 100). Appears to be rounding applied on AST. Unsure if the value in Prometheus given is decimal representation of the percentage due to what looks like inconsistent rounding - if that's what is being applied (info on this would be great)

Comparison Between tmctl of "profile_http_stat" vs f5_profile_http_requests_total: generally a very small difference is observed. Appears to be rounding applied on AST.

Comparison Between tmctl of "cpu_info_stat" vs f5_system_cpu_utilization_ratio generally a moderate difference is observed.

Comparison Between tcmtl of "systemcpu_info_stat" vs f5_system_cpu_utilization_ratio generally a moderate to large difference is observed.

If more info is needed, please let me know