Volume Metrics Dashboard

The Volume Metrics dashboard provides cloud administrators with a real-time overview of Cinder block storage resources across the deployment. It is accessible under Admin > Telemetry > Volume Metrics and requires the openstack.roles.admin permission.

All metrics originate from Ceilometer volume pollster data, stored in Prometheus via the sg-core telemetry pipeline, and queried through the Aetos observability proxy (service_type: metric-storage).

Ceilometer meters follow a naming convention when exposed as Prometheus metrics: dots are replaced with underscores and a ceilometer_ prefix is added (e.g. volume.size becomes ceilometer_volume_size).

Panels

Summary Cards

Four metric cards displayed at the top of the dashboard, auto-refreshing every 10 seconds.

Card	Ceilometer Meter	PromQL
Total Storage (GiB)	volume.provider.pool.capacity.total	sum(ceilometer_volume_provider_pool_capacity_total)
Allocated Storage (GiB)	volume.provider.pool.capacity.allocated	sum(ceilometer_volume_provider_pool_capacity_allocated)
Total Volumes	volume.size	count(ceilometer_volume_size)
Total Snapshots	volume.snapshot.size	count(ceilometer_volume_snapshot_size)

Total Storage and Allocated Storage aggregate the capacity reported by the Cinder volume backend(s) (e.g. LVM, Ceph) across all storage pools. Total Volumes and Total Snapshots count the number of distinct Ceilometer samples, which corresponds to the number of active volumes and snapshots in the deployment.

Volume and Snapshot Distribution (Doughnut Chart)

A Chart.js doughnut chart showing the proportion of volumes to snapshots. Rendered on initial page load from the same count() queries used by the summary cards.

Segment	Ceilometer Meter	PromQL
Volumes	volume.size	count(ceilometer_volume_size)
Snapshots	volume.snapshot.size	count(ceilometer_volume_snapshot_size)

This chart gives a quick visual ratio of how storage objects are distributed between volumes and snapshots.

Storage Pools

A per-pool breakdown with capacity bars and a detailed table. Rendered on initial page load.

The dashboard first discovers all pools by querying the raw (un-aggregated) capacity total metric, then fetches per-pool details by filtering on the resource label, which contains the Cinder pool name.

Metric	Ceilometer Meter	PromQL
Total capacity	volume.provider.pool.capacity.total	ceilometer_volume_provider_pool_capacity_total
Allocated capacity	volume.provider.pool.capacity.allocated	ceilometer_volume_provider_pool_capacity_allocated{resource="<pool>"}
Free capacity	volume.provider.pool.capacity.free	ceilometer_volume_provider_pool_capacity_free{resource="<pool>"}
Provisioned capacity	volume.provider.pool.capacity.provisioned	ceilometer_volume_provider_pool_capacity_provisioned{resource="<pool>"}

Each pool is displayed with:

• A capacity bar showing the percentage of allocated vs. total capacity.

• A detail row with free and provisioned values.

• A collapsible table with all numeric values and a color-coded usage percentage label (green < 75%, yellow 75-89%, red >= 90%).

All values are in GiB, matching the units reported by Cinder’s volume backend statistics.

Capacity Forecast

A prediction table that forecasts when each storage pool will run out of free space, auto-refreshing every 60 seconds.

Metric	Ceilometer Meter	PromQL
Current free capacity	volume.provider.pool.capacity.free	ceilometer_volume_provider_pool_capacity_free{resource="<pool>"}
Predicted free in 30 days	volume.provider.pool.capacity.free	predict_linear(ceilometer_volume_provider_pool_capacity_free{resource="<pool>"}[7d], 30 * 86400)

The predict_linear() function uses linear regression over the last 7 days of free capacity samples to project the value 30 days into the future. If the predicted value is negative, the dashboard computes the estimated number of days until the pool reaches zero free capacity using linear interpolation:

decline_per_day = (current_free - predicted_value) / 30
days_until_full = current_free / decline_per_day

Status labels:

• Full in N days (red) — predicted free capacity drops below zero within 30 days.

• Declining (yellow) — free capacity is decreasing but will not reach zero within 30 days.

• Healthy (green) — free capacity is stable or increasing.

• Insufficient data (grey) — not enough historical data for predict_linear() to produce a result (requires at least 2 data points over the 7-day window).

Note

Predictions require at least 7 days of continuous Ceilometer data collection. On a fresh deployment, this panel will show “Insufficient data” until enough history accumulates.

Data Pipeline

The metrics flow through the following pipeline:

Cinder → Ceilometer pollster → sg-core (STF) → Prometheus → Aetos → Horizon

1. Ceilometer polls Cinder every 600 seconds (default) for volume and pool statistics.

2. sg-core receives Ceilometer samples over AMQP/TCP and writes them as Prometheus metrics.

3. Prometheus stores the time-series data and serves PromQL queries.

4. Aetos acts as a Keystone-authenticated proxy to Prometheus, registered in the service catalog as metric-storage.

5. Horizon (grian-ui) queries Aetos using python-observabilityclient and renders the results.