Settings

New in the next release

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

Capacity planning works with no settings at all. Change these when your rules differ from the defaults.

Settings can be made in two places:

• Settings → Capacity planning, for admins, with a license and the settings store (SETTINGS_KEY). Saving re-works every forecast at once, and is in the audit log.
• The stack, as variables. Anything set in the stack shows locked on Settings, so each setting has one source.

Thresholds

The share of usable capacity each forecast is for.

Resource	Default	Notes
cpu	80%	Clusters and hosts
mem	80%	Clusters and hosts
disk	85%	Datastores, backup stores, volumes
pool	85%	Ceph and other replicated storage: Ceph's own nearfull
alloc_cpu	100%	Of cores × the vCPU ratio
alloc_mem	100%	Of host memory × the memory ratio
alloc_disk	100%	Of capacity × the provisioning ratio

On Settings, the Everything line sets defaults, and a line per cluster, datastore or pool overrides them. A blank keeps the value from the line above. Each must be 10% to 100%.

Which wins, for one resource of one thing:

1. a value set for that cluster, datastore or pool on Settings;
2. CAPACITY_THRESHOLDS, where it differs from the built-in default;
3. the Everything line on Settings;
4. the built-in default above.

Spare hosts (N+1)

How many hosts a cluster keeps spare, 0 to 4. With one, usable CPU and memory are the cluster less its largest host, and the forecast dates when it can no longer absorb a host failure. Set it per cluster on Settings, or with CAPACITY_HA, which wins. See forecasts.

Allocation ratios

CAPACITY_RATIOS, in the stack only:

Ratio	Default	Means
vcpu	4	vCPUs allowed per physical core
vmem	1	VM memory per byte of host memory (above 1 is overcommit)
disk	1.5	Provisioned disk per byte of datastore capacity (thin provisioning)

The VM size

CAPACITY_VM_SIZE is the VM that How many more fit counts in (4 vCPU, 16 GB, 100 GB unless set). Anyone can change it on the tab for a look without saving.

Monthly summary

On Settings → Capacity planning → Monthly summary:

Field	Variable	What it does
Send to	CAPACITY_SUMMARY_TO	Who gets it on the 1st, comma-separated. Its own list, not the alert email list
Prepared by	CAPACITY_SUMMARY_BY	A name on the summary, such as your company's
What to buy plans in a host of n cores and n GB	CAPACITY_HOST	The host "What to buy" adds. Blank: each cluster's largest

Email last month's now sends one at once: a good way to check it arrives. It needs the mail server set up for alerts (SMTP_HOST, SMTP_FROM). See Reports.

Racks

On Settings → Capacity planning → Racks, for the rack check. One line a rack:

Field	What it's for
Rack	Its name, such as R1
Free U	Rack units free now, 0 to 60
Host U	How many units a new host takes, 0 to 12. 0 is "don't know", and the check says so
Holds	The hosts and switches in it, comma-separated, by the names Ferrule shows. A host is matched by its full name or the part before the first dot
Fed by	The UPS that feeds it, from the UPSes Ferrule reads

Save racks saves them all, checked as a whole: two racks of one name, or numbers outside the ranges, are refused with the reason and nothing changes. When NetBox is connected, its racks are used for rack units instead; the UPS still comes from here.

Racks and reserved ports are kept in the settings store, so they're in its export and survive updates.

Every variable

All are optional. In the variable reference they're under Capacity.

Variable	Default	Format
CAPACITY_THRESHOLDS	cpu=80,mem=80,disk=85,pool=85	resource=percent, comma-separated. Also alloc_cpu, alloc_mem, alloc_disk
CAPACITY_RUNWAY	warn=90,bad=30	Days left at which a forecast warns, and goes critical
CAPACITY_RATIOS	vcpu=4,vmem=1,disk=1.5	See Allocation ratios
CAPACITY_VM_SIZE	vcpu=4,mem=16,disk=100	Memory and disk in GB
CAPACITY_HA	empty	cluster=hosts, such as Prod=1,lab=0. Use source/cluster=1, such as vcsa/Prod=1, where two sources have a cluster of the same name
CAPACITY_SUMMARY_TO	empty	Email addresses, comma-separated
CAPACITY_SUMMARY_BY	empty	A name
CAPACITY_HOST	empty	cores=64,mem=1024, memory in GB

A malformed value stops Ferrule starting, with the variable and the reason in the log, so a typo never silently plans against the wrong numbers.