Usage and service-level reports

Ferry measures what each project protects and stores, and what each plan achieves against its policy. Projects see the figures for their own resources in the console and the CLI. A cloud management platform reads them through the API, to bill for backup and to report service levels.

Measurements that are not yet available appear under unavailable, with a reason. Treat these as missing data when calculating usage or charges.

A plan's usage

Open a plan's What it takes up section to compare provisioned capacity, written data and backup storage usage:

Figure	What it measures
Declared	The size of the disks the plan protects, as created.
Written	How much of those disks the guests had actually used at the last backup.
Stored	Storage used by the newest backup of each resource, after compression and deduplication.
Read, all runs	How much has been read from the disks in total.

A project's usage

curl -H "X-Auth-Token: $OS_TOKEN" "https://backup.mycompany.com/v1/usage?project_id=<project-id>"
openstack backup usage show --project <project-id>

Both need a system-scoped token. With a token scoped to the project, GET /v1/usage and openstack backup usage show return the project's own usage, as its members see it.

Field	Meaning
protected_bytes	The written size of the protected resources, which is the data under protection. The size the volumes were created with is provisioned_bytes.
provisioned_bytes	The size the protected resources were created with.
stored_bytes	What the project's backups occupy in the repository, compressed and deduplicated within the project's own data.
soft_deleted_bytes	Data of backups the project deleted that is still kept until it is erased. Present only when there is some.
point_count	Backups held.
protected_resources	Resources under protection.
repositories	The same figures for each repository holding the project's data, with the vault it is offered as.
computed_at	When stored_bytes was last computed.
live_as_of	When the other figures were read. They are read on each request.

Project usage is calculated from that project's data. Identical data in two projects counts toward each project's usage. Charges based on protected_bytes or stored_bytes therefore remain independent of other projects' activity.

Each repository sweep recalculates stored_bytes, so it can lag behind backups and deletions by one sweep interval. Use computed_at as the measurement date when billing.

A repository's footprint per project

curl -H "X-Auth-Token: $OS_TOKEN" https://backup.mycompany.com/v1/repositories/<repository-id>/usage
ferry repository usage <repository-id>

Operator only. Each project with data in the repository is listed under tenants:

Field	Meaning
referenced_bytes	Everything the project's backups refer to, compressed, counted within its own data.
exclusive_bytes	What deleting all of the project's backups would free: data no other project refers to.
logical_allocated_bytes, logical_provisioned_bytes	The written and provisioned size of what its backups protect.
point_count, resource_count	Backups and resources.

And for the repository as a whole:

Field	Meaning
physical_bytes	What the repository holds.
sum_tenant_referenced	The sum of every project's referenced_bytes.
cross_tenant_margin	sum_tenant_referenced minus physical_bytes. Positive when deduplication across projects saves space. Negative while data that no backup refers to any more waits for the sweep, or for its lock to end.
locked_bytes, release_schedule	Data the storage's lock still holds, and how much of it is released each month.
held_bytes, held_release	Data of expired backups that the lock still holds, and the days it is released.
reclaimable_pending_bytes	What the next sweep frees.
verified_coverage_pct_30d	The share of the repository read back and checked over the last 30 days.
gc_last_completed_at	When the last sweep completed. Every footprint figure is as of that sweep.
ceph_trash_pinned_bytes	Volumes Cinder deleted into the Ceph trash that the service still keeps, because a backup needs them.
leaked_disk_bytes	Disks whose server or volume OpenStack deleted while the service still held them.

exclusive_bytes and the repository-wide figures depend on every project's data, so they are never shown to a project.

A plan's service level

curl -H "X-Auth-Token: $OS_TOKEN" https://backup.mycompany.com/v1/plans/<plan-id>/sla
openstack backup plan sla show <plan-id>

Any reader of the plan's project can read it. It compares the policy's targets with the plan's results over the last 30 days:

Field	Meaning
sla.actual_rpo	The largest age among the newest backups of the plan's resources, in nanoseconds.
sla.rpo_target, sla.rpo_compliance_pct	The policy's recovery point target, and the share of the window the plan met it.
sla.success_rate_30d	The share of backups that succeeded.
sla.consecutive_failures	Failed backups in a row, now.
sla.oldest_restorable_point	The oldest backup that can still be restored.
sla.immutably_protected_since	The oldest backup in a locked vault. It shows since when the plan has been protected against deletion, which is separate from since when it has been backed up.
consistency_promised, consistency_achieved	The consistency requested by the plan and achieved by its backups.
consistency_downgrade, points_downgraded	Why backups fell back to crash-consistent, and how many.

One backup's figures

openstack backup point metrics show <backup-id>

It shows the sizes of one backup, the compression and deduplication ratios, how long each phase took and how long the guest was frozen. A project sees the figures computed within its own data. An operator also sees what the backup transferred, which depends on what the repository already held.

Not protected

GET /v1/coverage and ferry coverage show list, per project, what is not protected and why. See Monitoring and coverage.