Troubleshoot Persistent Disk Performance Status Healthy

What the Healthy Performance Status Indicates

In Google Cloud, the compute.googleapis.com/instance/disk/performance_status metric reports the operational state of a Persistent Disk from Google's infrastructure perspective. A status value of Healthy (value 1 / 100%) indicates that the underlying Google Cloud storage subsystem is operating without internal infrastructure degradation, storage host failures, or ongoing maintenance interference.

How This Diagnoses and Resolves the Bottleneck

When a disk reports Healthy while simultaneously experiencing flatlined IOPS/throughput, elevated disk queue depth, and write latency spikes:

• Persistent Disk scaling model: Standard Persistent Disk (pd-standard), Balanced Persistent Disk (pd-balanced), and SSD Persistent Disk (pd-ssd) scale their maximum IOPS and throughput limits linearly with the provisioned capacity (disk size).
• Saturating throughput ceilings: If an application requires 400 MiB/s of write throughput or 15,000 IOPS, but the attached disk is sized such that its calculated baseline cap is much lower, the storage driver throttles I/O operations at the VM-to-storage interface.
• Remediation: The engineer can dynamically resize the existing Persistent Disk without downtime to scale up its baseline throughput and IOPS limits, or transition the storage architecture to Hyperdisk / Extreme Persistent Disk where IOPS and throughput can be provisioned independently of storage capacity.

Key Capabilities and Best Practices

• Non-disruptive capacity expansion: Resizing a Persistent Disk in Compute Engine can be done on the fly while the VM is running, allowing immediate throughput increases after resizing the partition/filesystem.
• Observability alignment: Pairing queue depth metrics with read/write throughput charts helps pinpoint whether an issue is caused by external hardware faults or provisioned quota exhaustion.
• Service Health correlation: Validating against Personalized Service Health (PSH) confirms that no storage platform incidents are artificially degrading volume responsiveness.

Why This Is Optimal

This approach directly targets the root cause of the storage bottleneck. Rather than treating the symptom as an infrastructure failure or misdirected OS error, resizing or re-tiering the disk provides the required I/O bandwidth to clear the elevated queue depth.