The Capacity tab¶

New in the next release

This is on the dev and qa builds now, and in the next release.

The Capacity tab appears once Ferrule watches a hypervisor or storage source. Without a license it explains what it does and that history is already being kept.

From the top, the tab has:

the Capacity table, the heart of it;
Plan: how many more fit, and what if (Planning);
the Monthly summary bar (Reports);
Reclaim and right-size (Reclaim);
Where storage growth comes from (Storage).

The table¶

Each row is one resource of one thing, soonest to fill at the top. A cluster has four rows (CPU, memory, and the two allocations), a datastore two, a pool, backup store or volume one.

Column	What it means
What	The cluster, host, datastore, pool, backup store or volume, its kind and source. `N+1` means the cluster keeps a host spare. A backup store also says what it keeps and how full its retention is.
Resource	See the table below.
Now	Use as a share of usable capacity, and the amount.
Usable	Capacity less anything kept back: spare hosts, Ceph's healing room. Hover it for how it was worked out.
Reaches	The threshold and when it's reached, with the likely range: 80% around 2027-03-04 (2027-02-12 to 2027-04-02).
In	Days until then. Amber inside the warning runway (90 days unless set), red inside the critical one (30).
History	Days of history behind it, and how many were read back from old inventories.

Resources¶

Resource	On	Measures	Against
CPU	Clusters, hosts	Cores in use, each day's 95th percentile	The cores there are
Memory	Clusters, hosts	Memory in use, each day's 95th percentile	The memory there is
vCPUs allocated	Clusters	vCPUs given to VMs	Cores × the vCPU ratio (4 unless set)
Memory allocated	Clusters	Memory given to VMs	Host memory × the memory ratio (1 unless set)
Disk	Datastores, pools, backup stores, volumes	Space used, each day's average	The usable space
Disk provisioned	Datastores	The size of every VM disk on it, thin or not	Capacity × the provisioning ratio (1.5 unless set)

CPU is always counted in cores, whichever hypervisor reports it: vCenter reports megahertz, which Ferrule turns into cores using each host's clock.

A cluster can be 40% busy and fully allocated at once. That's why both are shown: busy says whether VMs will run well, allocated says whether you can create more under the rules you've set.

What "Reaches" can say¶

It says	Meaning
80% around 2027-03-04 (02-12 to 04-02)	The forecast date, and the likely earliest and latest
80% around 2027-03-04 (from 02-12)	At the slow end of the range it may never get there
now	It's at or over the threshold already, counting its busiest weekday
not rising	Flat or falling: no date
not within a year	Rising, but slower than would reach it in 12 months
needs N more days of history	Under three weeks of history since the last change of capacity

Thresholds¶

Each resource has its own threshold: 80% for CPU and memory, 85% for disk and Ceph pools, 100% for the allocations, unless changed on Settings.

Dates for reaching __%: type a percentage to re-date every row for it, such as "when does anything reach 70%?". It doesn't change any setting. Leave it empty to go back to each resource's own threshold.

Hosts¶

Show hosts adds each host's own CPU and memory. Hosts are shown for standalone servers and lopsided clusters, but never alert: a cluster balances its VMs across hosts, so one busy host isn't a capacity problem. A host that isn't in a cluster is listed as a one-host cluster as well, and that row does alert.

The chart¶

Click a row to open its chart:

history as a share of usable capacity, a dot a day;
the threshold, as a dashed line;
the forecast running on from today, with its range shaded, out to the crossing or 90 days, whichever is further.

Click the row again to close it. Charts are drawn from Ferrule's own daily history, so they work without VictoriaMetrics.

The count on the tab¶

The tab's count is how many resources are inside the warning runway or over their threshold, the same ones that raise runway alerts.