OpenStack

Storage I/O Bottleneck Identification

warning
latencyUpdated Jan 12, 2015

Detect storage performance issues by correlating instance disk metrics with hypervisor disk availability and Cinder response times.

How to detect:

Monitor per-instance disk read/write rates, errors, and requested vs actual I/O. Correlate with hypervisor disk_available_least and Cinder API response times. Alert on elevated error rates or significant gaps between requested and actual I/O.

Recommended action:

Investigate Cinder backend storage health, check for competing workloads on storage nodes, review QoS policies. Verify hypervisor local storage isn't exhausted (disk_available_least).