Question 680
At Riverton Utilities, the server engineering team maintains field operations Linux telemetry nodes under requirements for controlled upgrades and retained system events. The operations lead requests proof of the fix for: A service uses excessive memory and the kernel begins killing processes, while the incident handler misreads cached pages as unreclaimable application RAM. Which observed result BEST demonstrates that the correction works?
Explanation
Correct Answer
A representative load stays within the intended memory limit and no new OOM kill occurs during the test window. This measurable outcome verifies correction of: A service uses excessive memory and the kernel begins killing processes, while the incident handler misreads cached pages as unreclaimable application RAM. The associated technical action is: Inspect free and process RSS/cgroup memory metrics and OOM logs, then correct the identified process or memory limit. The Riverton Utilities review checks actual behavior rather than an untested change claim.
Incorrect Answers
A controlled test notification reaches the intended manager and normal polling continues without opening public SNMP access. This evidence pertains to an unrelated remediation: Verify the supported SNMP agent trap target, credentials and transport through the approved management network. The alternative issue is: An SNMP monitoring system can poll a Linux device but receives no event notifications because the configured trap destination is incorrect. The current Riverton Utilities problem is: A service uses excessive memory and the kernel begins killing processes, while the incident handler misreads cached pages as unreclaimable application RAM. The decisive expected evidence is instead: A representative load stays within the intended memory limit and no new OOM kill occurs during the test window.
A controlled alert identifies the responsible server and affected filesystem and routes to the correct on-call group. This evidence pertains to an unrelated remediation: Include stable host, instance and environment labels in the monitoring data and notifications without exposing unnecessary secrets. The alternative issue is: An on-call team cannot distinguish which Linux host produced a disk alarm because the monitoring events omit environment and host labels. The current Riverton Utilities problem is: A service uses excessive memory and the kernel begins killing processes, while the incident handler misreads cached pages as unreclaimable application RAM. The decisive expected evidence is instead: A representative load stays within the intended memory limit and no new OOM kill occurs during the test window.
A short healthy burst stays below the alert condition while sustained overload triggers one actionable notification. This evidence pertains to an unrelated remediation: Set sustained threshold windows and recovery conditions based on measured service impact rather than one noisy sample. The alternative issue is: A monitoring alert floods operators every minute because it fires for any single transient CPU spike without duration or hysteresis. The current Riverton Utilities problem is: A service uses excessive memory and the kernel begins killing processes, while the incident handler misreads cached pages as unreclaimable application RAM. The decisive expected evidence is instead: A representative load stays within the intended memory limit and no new OOM kill occurs during the test window.