Skip to main content

Storage

EmberNet's storage layer, Cinder, gives your workloads resilient, replicated block storage at the edge, and gives you a single EmberCORE view of capacity, volume health, and backups across every cluster you operate.

You don't need to know how Cinder works underneath to use it. The Storage view surfaces what matters: how full you are, which volumes are healthy, and whether your backups are landing.

The Volumes tab under Data, saying Cinder is not running on this cluster and that volumes, disks, backups, and storage settings appear once Cinder runs on it

In the console, storage is under Data, then Volumes. On a cluster that has no Cinder service, the tab says so, and shows no volumes, disks, backups, or settings until Cinder runs there. On a hosted cluster, a tenant Admin who is refused the cluster-wide storage settings sees that Fireball manages them, not a red failure.

The Storage view​

Open Storage from EmberCORE. For the selected cluster or tenant it shows:

  • Capacity: how much of your storage is actually used versus provisioned, sourced from real on-disk usage rather than nominal volume sizes.
  • Volume health: a live count of healthy volumes versus those degraded or rebuilding.
  • Scheduling guardrail: whether any disk has run out of room to place new or rebuilt replicas.
  • Backups: recent backups and their targets.

Volumes and replicas​

Every volume keeps multiple replicas on different nodes, so the loss of a single node or disk doesn't lose your data. A volume's health reflects how many of its configured replicas are currently good:

StateMeaning
HealthyAll configured replicas are up.
DegradedServing data, but with fewer healthy replicas than configured, so another failure would risk data loss. Cinder rebuilds automatically once there's somewhere to place the replica.
FaultedNo usable replicas; the volume cannot serve data. Restore from backup.

A degraded volume raises a Storage alert. The most common cause is a volume configured for more replicas than there are storage nodes to hold them. A three-replica volume on a two-node cluster can never be fully healthy. Match the replica count to your node count, or add a node.

Capacity and the scheduling guardrail​

Cinder will keep accepting writes right up until a disk is full, at which point it can no longer place a rebuilt replica, which is the exact moment you most need the headroom. EmberNet watches for this and raises a capacity guardrail alert before you get there, based on two thresholds:

  • Over-provisioning: how far total requested storage may exceed physical capacity.
  • Minimum available: how much free space each disk must keep in reserve.

When a disk breaches the guardrail, add capacity, raise the over-provisioning allowance, or free space. See expanding capacity.

Backups​

Cinder can back volumes up to an external backup target:

TargetNotes
S3 / S3-compatibleAWS S3 or any S3-compatible object store (endpoint, bucket, region, credentials).
Azure BlobAn Azure storage account and container.
NFSAn NFS export.

After you configure a target, use Test connection to confirm EmberNet can reach it: the test checks that the bucket, container, or export exists and that the credentials are accepted. Saving a target does not run the test for you. Configure targets under Storage → Backup Target. CinderSilo, EmberNet's own backup store, shows in the picker as not available yet. See Cinder.

Expanding capacity​

When a node's disk is the limiting factor, add storage without downtime by attaching an additional data disk to the node and registering it as a second Cinder disk, never by resizing the OS disk, which requires a reboot. New and rebuilt replicas can then schedule onto the added space, and degraded volumes recover on their own.

Next steps​