Planning: how many fit, and what if¶
New in the next release
This is on the dev and qa builds now, and in the next release.
Open Plan: how many more fit, and what if above the Capacity table. All of it works from the first poll; the dates in it need the three weeks of history the forecasts do.
The VM size¶
A VM of n vCPU, n GB memory, n GB disk is the size the plan counts
in. It starts at CAPACITY_VM_SIZE (4 vCPU, 16 GB, 100 GB unless set);
change it here to try another. It isn't saved.
How many more fit¶
For each cluster and datastore: how many more VMs of that size fit before anything reaches its threshold, and what stops the next one: CPU, memory, vCPU allocation, memory allocation, disk, or provisioned disk.
- A new VM's memory counts in full.
- Its CPU counts at what each allocated vCPU in that cluster actually uses today. If the cluster's VMs use a quarter of a core per vCPU, a 4-vCPU VM counts as one core. That's how its neighbours behave.
- Its disk counts in full on a datastore, so thin provisioning leaves more room than shown.
A 0 in red means the next one doesn't fit; amber means fewer than five.
What if, for a cluster¶
Pick a cluster, then a change. Each resource shows its use now and after, and when it reaches its threshold now and after.
Add VMs of that size¶
N VMs of the size above, counted the way "how many more fit" counts them.
Add a host like its largest¶
The cluster grows by one host the size of its largest. Below the table, the rack check says whether that host would actually fit: switch ports, power, rack units, addresses and licenses.
Lose its largest host¶
The cluster shrinks by its largest host. For a cluster already keeping one spare, this takes one more away, and it says so.
Move VMs from it to another cluster¶
Tick VMs in one cluster and see both clusters after: the one they leave and the one they land on. It works between any two clusters, even on different hypervisors.
- The list is the cluster's VMs, busiest first, with each one's vCPUs and memory, its peak use and its disk. A filter finds them by name.
- CPU and memory move at what each VM used on its busiest day in the last 30, not its average: you plan for its worst day.
- Allocation moves at what it's given.
- A VM that was off all month moves allocation only.
- Storage stays where it is unless the target cluster counts disk.
VMs are listed from the first poll after this release, so a newly watched cluster shows its VMs once the collector has seen them.
What if: add disks¶
Below the cluster what-if: pick a pool or volume, a number of disks and their size in TB. Ferrule works out the usable space they add from the pool's layout, and shows its disk use and runway after.
| Layout | Grows by | Example |
|---|---|---|
| ZFS RAIDZ (TrueNAS) | Another vdev like the ones it has. A RAIDZn vdev of w disks needs w at once and gives w − n of them. | 6 × 8 TB to a pool of RAIDZ2 vdevs six wide: +32 TB. |
| ZFS mirrors | A mirror at a time, each giving one disk's worth. | 3 × 4 TB to two-way mirrors: +4 TB, one disk unused. |
| Synology SHR | Everything but the largest disk's worth. | A 12 TB disk into 8 + 8 + 4 TB: +8 TB, not 12, until a second large disk arrives. |
| Synology SHR-2 | Everything but the two largest disks' worth. | |
| RAID 5 / RAID 6 (QNAP, Synology) | Every disk at the smallest disk's size. | 8 TB disks into a RAID 5 of 4 TB: +4 TB each. |
| RAID 10 | A pair at a time. | |
| RAID 1 | Nothing: more disks are more copies. It says to replace with bigger disks instead. | |
| Ceph | Raw space, as Ceph's capacity is planned. The note says how much more data that holds at three replicas, and to spread the new OSDs evenly across hosts. | 3 × 4 TB OSDs: +12 TB raw, about 4 TB of data. |
It also says:
- when fewer disks are offered than a vdev needs, that they can't make one, and that OpenZFS 2.3 (TrueNAS 24.10 on) can widen a RAIDZ vdev one disk at a time instead;
- how many empty bays there are, where the NAS lists them (QNAP), and if the disks need more bays than that, that it takes an expansion shelf or bigger disks;
- how many disks are already installed but in no pool (TrueNAS).
Usable space is before the file system's own overhead, a few percent.
A pool whose layout its NAS doesn't report isn't offered. See Storage for what each NAS reports.
The API¶
Each what-if is worked out in the browser from the same figures as the table, except moving VMs, adding disks and the rack check, which ask Ferrule. Those endpoints are on the reference page.