Skip to content

Prefixes

A prefix is a block of address space written in CIDR form - for example 10.0.10.0/24. This page covers creating, editing, and deleting prefixes, and the time-saving inheritance that fills in context for you.

Add a prefix

There are three ways into the prefix form, and each carries different context with it:

You start from… The new prefix inherits…
The Add prefix button on the prefix list Nothing - it starts as a top-level block.
More ▸ Add child prefix on a prefix's detail page The site, VLAN, and VRF of the parent prefix.
A free slot in the Children tab, or a free chip in the Space map The site, VLAN, and VRF of the smallest existing prefix that contains the new block.

When inheritance applies, the form shows a small green chip confirming what's being carried over, so you can see at a glance where the values came from:

✓ Inheriting site + VLAN from parent 10.0.10.0/24

Fill in the remaining fields - status, role, description, gateway, tags - and save.

Let the children inherit

Building out a network under a parent? Use Add child prefix instead of the global Add prefix button. You'll skip re-picking the site, VLAN, and VRF every time, and you won't accidentally split a block across the wrong VRF.

A site's address scope

A site can own one or more prefix ranges - together they're the site's address scope. Site-scoped editors (see Permissions → Site roles) can only carve child prefixes inside those ranges, so the scope doubles as a guardrail.

Open a site and use its Prefixes tab (the address scope is also summarised on the site's Edit form). Two ways to populate it:

  • Add prefix range - create a brand-new prefix already assigned to this site.
  • Assign prefix - pull an existing prefix into the site. Search, pick one, and it's moved into the site's scope. You can assign several in a row.

To remove a range from a site, edit that prefix and clear its site.

Edit a prefix

Open the prefix and click Edit. The form opens populated with the prefix's current values, including its tags. Change what you need and save.

Delete a prefix

Open the prefix and click Delete. You'll be asked to confirm.

The addresses recorded inside it are handed to the prefix that still contains them - the next one up, longest match - and keep their assignments; the confirmation says how many move and where. Only addresses no other prefix covers are removed with it, and the confirmation counts those too.

To delete several, tick them in the Prefixes list and press Delete in the bar. Their addresses move the same way, onto prefixes that stay. More than 1000 go 1000 at a time (see Large selections); an address on a prefix a later batch deletes moves on again, so it ends up where one call would have put it.

Renumbering a prefix

When a block you don't control changes - a provider's DHCPv6-PD /60, a re-issued /29 - everything under it has to move by the same offset. Open the prefix and pick More ▸ Renumber…, type the new network, and press Preview. The new prefix must be the same family and length, written as a network address; the dialog says so as you type. Renumbering stays in the prefix's VRF.

The preview lists every change, grouped by kind, as old → new; badges count each kind, and the linked badges count records that point at a moved prefix or address and follow it without an edit. Left unchanged lists everything that carries an address inside the prefix but is not rewritten, with the reason. Renumber asks once more, then writes it all in one change: if any part is refused, nothing moves.

What moves, by the difference between the new and old network:

Kind What changes
Prefixes The prefix and every prefix inside it in the same VRF; a prefix's gateway.
IP addresses Every address inside it in the VRF. Mask length, interface or VM assignment, role (gateway, anycast, VIP), status and DNS name stay. An address recorded on a prefix above the tree moves onto the tree prefix that holds it.
IP ranges Start and end.
DHCP Local scopes on a moved prefix: scope id, pool, options that hold an address, exclusions and reservations.
DNS A/AAAA records in local zones, linked to a moved address.
Routing Static route destinations and next hops, BGP neighbour addresses and EIGRP summary addresses in the prefix's VRF; prefix-list lines that name a moved prefix.
Global table only Prefix-list lines by text, policy-rule selectors, route-map next hops, the BGP VPN export next hop, the LDP transport address, RPKI cache addresses, SNMP binding targets, BMC (Redfish) hosts and watched endpoints - they have no VRF to match on.
Monitoring Deny subnets in the prefix's VRF.

FHRP virtual IPs, NAT rules, services, VTEP addresses, BGP local addresses, device primary IPs and anything else holding a moved address or prefix by reference follow it: they are counted as linked, not rewritten.

Left unchanged, and listed: scopes, records and reservations on a Windows DHCP or DNS server (change them there; the next sync follows), reverse (PTR) records, observed leases, ARP sightings and DNS drift, Oxidized and PRTG addresses, an Outpost's own connection address, ROAs (issue one for the new prefix), aggregates, a gateway outside its prefix, and an address or range on a moved prefix that lies outside it. With the prefix in a VRF, the global-table-only references above are listed instead of rewritten.

It is refused when a prefix, address or range already sits in the new network in that VRF, or an IP range crosses the edge of the old one. A prefix that contains the new network - a covering /48 - is fine; site-scoped editors may not move into another site's or the shared space. You need change on every kind the preview lists, within your sites; the refusal names what is missing.

The preview shows only what you may view, each row by the rule of its own page: an ARP entry or BMC address of a device at another site is not listed. Those rows are counted together under Not shown, with nothing else about them; they stay as they are. When something the renumber would have to rewrite is one you can't view, it is refused and says how many such objects there are, so another site's records are never changed on your behalf.

Every changed object gets a change-log entry with the cause Renumbered X to Y, and the prefix's journal records Renumbered from X to Y. The API is POST /api/prefixes/<id>/renumber/ - see REST API.

Browse by tree

The left rail of the prefix list shows either Filters (the facets) or a Tree - switch at the top of the rail. The choice is remembered per user for this list and kept in the URL (?rail=tree), so a link opens the same rail. On a narrow window the rail opens from the Filters / Tree button in the header.

The tree shows one VRF at a time: pick Global or a VRF at the top. Below it, the VRF's top-level prefixes; a chevron opens a prefix to its direct children, and the number beside a prefix is how many direct children it has. Each level loads when you open it, 200 prefixes at a time - a /16 with 2,000 /29s shows the first 200 and Show 200 more.

Type in Filter prefixes to search the VRF on the server: the start of a CIDR (10.1 finds 10.1.0.0/16, 10.10.0.0/16, …) or a whole address or network (10.1.0.0/16, 10.1.2.7), which finds what is inside it, or else the smallest prefix holding it. The tree then shows only the matches and the prefixes above them; open a match to see its children.

Click a prefix (or press Enter) to select it. The table then lists that prefix and everything inside it, in its VRF; with nothing selected and the tree open, the table lists the whole VRF. The selection is ?within=<id> and works with the other filters. Switch back to Filters and it stays, as a Within 10.1.0.0/16 chip over the table; its ✕ (or clear in the tree) goes back to the whole list.

The tree is a keyboard tree: Up / Down move, Home / End jump, Right opens a prefix or steps into it, Left closes it or steps out to its parent, Enter or Space selects.

On a prefix's page

A prefix's IPs and Child prefixes tabs have the same Filters / Tree switch. There the tree is for getting around: it shows the branch the prefix sits in, from the top prefix above it in its VRF, opened down to the prefix. Its parent's other children - its siblings - sit around it, the prefix itself is marked This prefix and opens to its own children. A prefix with nothing above it is the top of its tree. There is no VRF picker (the prefix's own VRF applies), and Filter prefixes searches the whole branch.

When the prefix is deep in a long list (a /16 with 2,000 /29s), its level opens at the 200 that hold it, with Show 200 earlier above and Show 200 more below.

Click a prefix in the tree (or press Enter on it) to open its page, on the same tab with the tree still open - where the tree centres on that prefix. That goes for every prefix shown: one above, a sibling, a sibling's child, one inside this prefix, or a search match. Clicking this prefix does nothing. The tabs always list this prefix's own rows: the tree does not scope them, and a ?within= in a prefix page's link is ignored.

The mode is ?rail=tree, beside the tab's ?tab=, so a link opens the same view; both tabs share it. Tree or Filters is remembered per user for prefix pages, apart from the two lists - a tree you browse the lists by doesn't change how a prefix page opens. On a narrow window the rail opens from the Filters / Tree button beside the row count.

On a VRF's page

A VRF's Prefixes and IPs tabs have the same switch. Filters shows the list pages' facets for the VRF's rows (the prefix list's, less VRF, and the IP list's). Tree is the VRF's own tree: its top-level prefixes on top, opening and paging as above, with no VRF picker, and Filter prefixes searches the VRF.

Click a prefix to scope the tab to it, on the server: Prefixes lists it and everything inside it, IPs the addresses on it or on any prefix below it. A pick from another VRF scopes nothing (the chip reads unknown prefix). The pick (?within=) and the mode (?rail=tree) sit beside ?tab= and are shared by both tabs, with the Within … chip and clear as above. Tree or Filters is remembered per user for VRF pages, apart from the lists and prefix pages. On a narrow window the rail opens from the button beside the row count.

Hidden prefixes

The tree holds only prefixes you may view. A prefix whose parent you can't view hangs from the nearest prefix above it that you can - or sits at the top of the VRF - so a hidden prefix is never named, shown or counted; on a prefix's page the branch starts at the top prefix you can view. A selection naming a prefix you can't view lists nothing.

Free space in the list

Tick Show free space at the top of the prefix list's filter sidebar. Each container, and each prefix that holds child prefixes, then gets free rows between its children: the gaps, as the largest aligned blocks. A /16 holding 10.20.0.0/24, 10.20.1.0/24 and 10.20.4.0/22 shows 10.20.2.0/23, 10.20.8.0/21, 10.20.16.0/20 and so on.

A free row shows its block, a dashed Free badge and its size (512 addresses; IPv6 counts /64s). Create opens the prefix form with the block, the parent's VRF and the parent's site filled in; the form's Size picker carves a smaller block at the same start, so a /23 gap can become a /24, /25 and so on. A parent with more than 8 free blocks shows the first 8 and + N more free, which loads the rest (up to 256; past that it opens the prefix, whose Space map shows them all).

Free rows are never selected, exported or counted: the row count, the VRF group counts, Download and the bulk bar see only real prefixes. The switch lives in the URL (?free=1), so a link or a saved view keeps it, along with the size, Next available and scope below.

IPv6 free space is offered in whole /64s under a /64 or shorter: a /64 holding a /127 link is not free, and blocks start on /64 boundaries. Under a longer parent (a /120), any size is shown.

Only prefixes count as used. Stray addresses and IP ranges don't take space here, as on the Space map: a new prefix over them adopts them.

Find a size

With free space on, type a length into Free size (/29, or pick one of the sizes that fit). The list then shows only the parents that have a free block of that size, each followed by its first 16, lowest address first, and + N more free for the rest.

Tick Next available to see just one: the lowest free block of that size among the listed parents (the first VRF in the list that has one), highlighted, with Create. The sidebar's other filters (VRF, site, tags, status, search) narrow which parents are looked in.

Scope

Scope limits every search to a block inside the parent: type any CIDR (10.20.0.0/19 out of a /15) or pick one of the listed parents. Free rows, Free size, Next available and the planner then offer only space inside it.

Who sees what

Free space is worked out on the server against every prefix in the VRF, including ones you may not view, so a free row is always really free. Space a prefix you can't see takes is simply not offered - it is never named and never counted in the totals. Free rows appear only under prefixes you can view. Create, allocate and the planner check again on the server: they need add on prefixes, and with site-limited access the new prefix must sit in your own site's space, as for any new prefix.

With site-limited access, a prefix that is shared (no site) or another site's - or sits inside one - takes no children from you, so its Add child prefix and Plan subnets… buttons, the Space map's New child prefix here and its free rows' Create are not shown. Each prefix in the API says so in permissions.add_child.

A manual create takes the same lock as allocate and the planner: a row lock on every prefix holding the new block, taken before it is checked. So does every other write that puts a prefix somewhere new: a row of a CSV or JSON import that creates one or changes its CIDR or VRF, an edit that changes a prefix's CIDR or VRF, and a bulk edit that moves prefixes to another VRF (the lock on what holds each of them there). Two such writes running at once wait for each other, so they never land on the same block. An edit that leaves the CIDR and VRF alone takes no lock.

Plan subnets

Plan subnets… on a container or parent prefix (and in the list toolbar when free space is on and one parent is in scope) places a set of subnets at once. Each row is a count and a size - 2 × /24, 5 × /29 - with an optional description pattern (servers-{n} numbers them), status, VLAN and site (the parent's site by default). Keep together places a row's subnets side by side in one block.

The planner places the largest first, each aligned to its size, inside the parent or the Scope:

  • Lowest address first fills from the bottom of the free space;
  • Best fit uses the smallest free block that holds each subnet, which keeps big blocks whole;
  • Leave room to grow keeps the next block of the same size free after each one (not created, just left alone).

Preview shows the proposed prefixes in a table and on a strip of the parent in the Space map's colours, with the proposed blocks outlined. A row that doesn't fit says why: No free /24 left in 10.0.0.0/22; largest free block is /25. Create all is offered only when everything fits, asks once more, and creates the lot in one change - all or nothing. If the free space changed since the preview, it refuses and asks for a new preview. Every new prefix gets a change-log entry with the cause Planned in <parent>.

Moving existing prefixes to make room is not part of planning; use Renumber for that.

The API: GET /api/prefixes/?free=1, GET /api/prefixes/available/, POST /api/prefixes/<id>/allocate/ and POST /api/prefixes/<id>/plan/ - see REST API.

What gets validated

Danbyte checks a few things when you save, so your data stays clean:

Rule What it means
Valid CIDR The block must be a real network (for example 10.0.0.0/24). It's stored in its canonical form.
Valid gateway If you set a gateway, it must be a valid IP address.
No duplicates in a VRF The same block can't exist twice in the same VRF. The same block can exist in two different VRFs - that's expected and supported.

Carving a subnet re-homes its IPs

Each IP belongs to exactly one prefix. When you create a child prefix (from the detail page's Add child prefix, the space map, or the API), Danbyte automatically moves the IP addresses it now most-specifically contains onto the new prefix - pulling them down from the broader parent so they aren't stranded. It never steals IPs that belong to an existing more-specific child, and only affects the same tenant + VRF.

Add a pool of addresses

Rather than adding hosts one at a time, the Add pool button on a prefix's IPs tab bulk-creates every address across a range in one step. It appears on enumerable leaf prefixes (not containers or prefixes with child prefixes).

The dialog takes a start and end address, an optional status, role, and description applied to every address, and a Populate from shortcut that fills the range from:

Addresses that already exist are skipped, and the prefix's network and broadcast addresses are never created. A single pool is capped at 1024 addresses - add a wider span in smaller ranges. Each address is created individually, so it lands in the change log and history like any other IP.

Edit or delete many addresses

The IP addresses list (and a prefix's IPs tab) has a checkbox per row and a select-all in the header. Once something is ticked a bar appears with the count and the actions that take a selection: Edit opens the bulk editor for the chosen addresses (status, role, VRF, tags, description and the rest, one field at a time), Delete asks once, naming how many rows go, Export writes just those rows, and Check now re-runs their monitoring checks. Select-all covers the rows on the page in front of you, after the filters you have set.

Gateway autospawn

If you leave the gateway field blank and the prefix's site has a gateway policy set, Danbyte creates the gateway address for you automatically when you save. See Gateway autospawn for how to set that up.

See also