Hypervisor Load Imbalance Detection
warningscalingUpdated Oct 24, 2022
Identify uneven VM distribution across compute nodes leading to performance degradation on overloaded hypervisors while others remain underutilized.
How to detect:
Compare running_vms, memory_mb_used, vcpus_used, and load averages (load_1, load_5, load_15) across all hypervisors. Alert when standard deviation exceeds thresholds or when specific nodes show high load while others are idle.
Recommended action:
Review nova-scheduler configuration and filters. Consider live migration to rebalance workloads. Investigate scheduler weights and host aggregates for proper distribution.