Professional Cloud Architect

Question

You have a Compute Engine application that should autoscale when total memory utilization exceeds 80%. You installed the Cloud Monitoring agent and configured the autoscaling policy as follows:

Metric identifier: agent.googleapis.com/memory/percent_used
Filter: metric.label.state = 'used'
Target utilization level: 80
Target type: GAUGE

The application does not scale during high load. You want to fix this. What should you do?

Explanation

The agent.googleapis.com/memory/percent_used metric is broken out by state (used, buffered, cached, slab, free) whose values sum to 100%. On Linux much RAM is reported as buffered/cached, so the strictly 'used' state can stay well below 80% even under heavy load, meaning a filter of state='used' with target 80 never triggers scaling. Filtering state='free' with a target of 20 correctly scales out when free memory drops below 20% (equivalent to total utilization exceeding 80%). Option C is invalid because a single metric label cannot equal several values simultaneously (the AND filter matches nothing), and DELTA_PER_MINUTE (A) is wrong for a gauge metric.

Learn more

Scale based on Monitoring metrics

Professional Cloud Architect

AstroTutor

Notes

Readiness

QuestionQ5

Community Discussion