Terraform provider
fiberax/fiberax is the Terraform and OpenTofu provider for the Fiberax API: servers, SSH
keys, public addresses, private networks, disks, network cards, the server firewall, backup policies and,
since 0.2.0, router configuration: access rules, port forwards, WireGuard and OpenVPN tunnels with their devices;
and since 0.3.0, server types with graphics cards - the cards you pick yourself, the GPU package that defines a whole
machine, and the storage class of the system disk. Since 0.4.0 you can also choose the host a GPU Pro server
lands on (GPU Pro is opened per project on request - see the GPU Pro guide), and since 0.4.1 the provider refuses at plan time what the platform would refuse anyway - a server
with no network, no name or no credentials, an empty public_ipv4 block, two spellings of one label.
Since 0.5.0 the provider follows the API rules for the server firewall and network cards at plan time
(see Upgrading to 0.5) and paces itself by the rate limit the API reports in every
response. 0.5.1 needs no change in your configuration: a change of ssh_keys no longer
recreates the server, a server left over from an interrupted run is named on the next plan, and an
unchanged firewall no longer blocks terraform destroy. 0.6.0 refuses at plan time a server
order the platform would refuse (a type, image or plan that is not in the catalogue), no longer stops a server
for a plan change the platform refuses anyway, and needs Terraform or OpenTofu 1.11 or later (see
What is new in 0.6.0). 0.7.0 tells the platform on every terraform destroy
what to do with the server's public addresses: an address taken with public_ipv4 = { new = true }
goes back to the shared pool, any other stays with the project (see What is new in 0.7.0);
0.7.1 corrects the recipe for keeping such an address (see What is new in 0.7.1),
0.7.2 explains how to release only the address of a server (see What is new in 0.7.2),
0.7.3 removes false warnings from that recipe (see What is new in 0.7.3),
0.7.4 corrects the price reference (see What is new in 0.7.4),
0.7.5 gives a way out when terraform destroy stops on a server that is already gone (see
What is new in 0.7.5), and 0.8.0 takes up to four backups a day, refuses at
plan time more server orders the platform would refuse (limits of the type, a GPU package the type does not
sell, Windows without a password, a name that is not a host name, a password with the word Administrator) and
warns before a replacement that could destroy the old server and only then refuse the new one (see
What is new in 0.8.0); 0.9.0 manages object storage (S3) - buckets, S3 access
keys, service accounts and the storage reservation (see What is new in 0.9.0), 0.9.1
warns on the plan that replacing a server erases its data disks (see What is new in 0.9.1),
0.10.0 manages Virtual Data Center projects with an account key - ordering and closing projects, keys of one
project, and working inside any project with project_id (see What is new in 0.10.0),
and 0.11.0 installs SSH keys given as text on a server without adding them to the account's key pool (see
What is new in 0.11.0).
The current version is 0.11.0. It is in a trial period with the first customers and is not in
the public Terraform registry yet, so it is served from this mirror.
What is new in 0.11.0
A new minor version: SSH keys as text come to fiberax_server. Nothing changes for an existing
configuration, but ~> 0.10.0 keeps getting 0.10.0 until you change the constraint to
~> 0.11.0 (see Upgrading to 0.11).
ssh_public_keysinfiberax_server. OpenSSH public keys as text, one per element - the line of a.pubfile. They are installed in the guest when the server is created and are NOT added to the account's key pool, which every project of the account can list and delete from: use them for keys that must stay with one server, for example the keys of your own customers. Together withssh_keysno more than twenty. The field is sensitive, so the plan does not print it; a server read shows the installed keys as fingerprints only.- Checked on the plan the way the API checks them. DSA keys (
ssh-dss) are refused, and so is a line with a control character anywhere (a tab between the key and the comment too), a line separator or U+FFFE/U+FFFF, or a comment with:or a trailing:- they would break the server's cloud-init data. - Changing the set never recreates the server. The new set is recorded in state and the plan warns: to
put a key on a running server, call
POST /v1/servers/{id}/actions/add-ssh-keys; to give it exactly a new set, reinstall it withssh_public_keys(POST /v1/servers/{id}/actions/rebuild).
The guide Reseller: your customers' SSH keys explains when to use text keys instead of the pool.
Upgrading to 0.11
- No state migration. A configuration without
ssh_public_keyskeeps working unchanged. - Change the constraint to
~> 0.11.0and runterraform init -upgrade.
What is new in 0.10.0
A new minor version: account keys and projects come to the provider. Nothing changes for an existing
configuration, but ~> 0.9.0 keeps getting 0.9.1 until you change the constraint to
~> 0.10.0 (see Upgrading to 0.10).
- An account key in the provider. An account key (client area, API tokens, Account keys) works in
every Virtual Data Center project of the account. The new provider argument
project_id(or theFIBERAX_PROJECT_IDenvironment variable) names the project it works in; the provider sends it in theFiberax-Projectheader. Several projects - one provider configuration per project, with an alias. A project key needs nothing new. fiberax_project. Orders a project (free; you pay for what you create inside), renames it and sets its labels, and closes it - only an empty one: otherwiseterraform destroynames what is left, and nothing is closed.fiberax_project_token. Issues a key of one project - for a customer of yours or a pipeline. The secret is kept in state (sensitive) and set only when the key is issued; destroy revokes it.- Data source
fiberax_projects. The projects of the account, optionally by label - the way to feedproject_id, which must be known when the plan starts.
The guide Projects with an account key walks through a configuration end to end.
Upgrading to 0.10
- No state migration. A configuration with a project key keeps working unchanged.
- Change the constraint to
~> 0.10.0and runterraform init -upgrade.
What is new in 0.9.1
A patch version: ~> 0.9.0 picks it up, and nothing changes in configurations.
- Replacing a server warns that its data disks lose their data. A change that recreates a server
(such as
image),-replaceortaintreplaces everyfiberax_diskthat takesserver_idfrom it: the disk is destroyed together with its data and created empty on the new server. The plan showed only "must be replaced" for the disk; now each such disk warns "The disk will be destroyed together with its data, and an empty one created" and says why. A disk moved to another server in the configuration warns the same way. Replacing the disk itself (-replaceortaintof the disk) still gets no such warning. - No false warning or refusal on such a replacement. A disk replaced together with its server no longer
warns "The server will be restarted to grow the disk", and a smaller
size_gbin the same change is no longer refused with "A disk can only grow": the new disk is created at the size the configuration asks for.
What is new in 0.9.0
A new minor version: object storage (S3) comes to the provider. A configuration pinned to
~> 0.8.0 keeps getting 0.8.0 until you change the constraint to ~> 0.9.0
(see Upgrading to 0.9).
- Object storage resources.
fiberax_object_storage_bucket(a bucket: name and Object Lock),fiberax_object_storage_access_key(an S3 key to every bucket of the project),fiberax_object_storage_service_account(an S3 key limited to one bucket with read, write or readwrite access) andfiberax_object_storage_reservation(the storage reservation the project is billed for); the data sourcefiberax_object_storagegives the endpoint and region for S3 tools, the bounds of the reservation and the buckets with their size. - A token of its own. An object storage project has its own token, so its resources take a second
provider configuration with an alias and its own
token. A token of the wrong product is now named as such instead of "The project is unavailable". - Secrets live in state. The secret of a key or a service account is returned once, when it is
created, and is kept in the sensitive attribute
secret_access_key; after an import it is empty. - The reservation is money and only grows. The plan names the growth before you approve it; a value
below the reservation is refused;
terraform destroyleaves the reservation as it is.
The guide Object storage (S3) walks through a configuration end to end, the aws provider for the objects included.
Upgrading to 0.9
- No state migration. A configuration of a Virtual Data Center project keeps working unchanged.
- Change the constraint to
~> 0.9.0and runterraform init -upgrade.
What is new in 0.8.0
A new minor version: the plan refuses more configurations that the platform would refuse anyway. A
configuration pinned to ~> 0.7.0 keeps getting 0.7.5 until you change the constraint to
~> 0.8.0 (see Upgrading to 0.8).
- Several backups a day.
fiberax_server_backup_policytakestimes: 1 to 4 runs a day,"HH:MM"with leading zeros, at least 120 minutes apart.hourandminutestay and read the first of the times; write eithertimesorhourandminute, not both. One run a day:times = ["03:00"]. Removingtimesdoes not fold the schedule back to one time; on a server with several times,hourandminutemay only repeat the first one. - More is refused at plan time, only when a server is created (the new server of a replacement
included): cores, memory, disk or graphics cards above the limits of the server type; a
gpu_packagethe type does not sell; a Windows image withoutpassword(SSH keys do not work on Windows); anamethat is not a host name (a..b,a.-b); apasswordwith the wordAdministratorin any letter case. A change of an existing server is not affected. Every such refusal names the way out forterraform destroyand says not to addpublic_ipv4ornetworksjust to get past it: that orders a new, billed server. - A replacement that could lose the old server warns. When a field the rules for a new server check
is known only during apply (an expression of a resource created in the same run), the old server, with its
disks, was destroyed first and the new one refused after. The plan now warns and names the fields; apply the
resource the value comes from first (
terraform apply -target=...).create_before_destroyon the server alone is not enough - see the guide A project end to end. - Disks. A data disk written by 0.5.1 or imported now gets the record of its size at the first refresh, and the plan refuses shrinking it, as for a disk the provider created.
- Power and plans. A server that is starting after a disk grew reads as started, and another apply
does not power it on again. A change of
plan_idto a smaller plan, or one the type does not sell, now warns that the platform refuses it without stopping the server. - Texts. The
ssh_keyswarning namesreset-passwordon Windows;terraform destroy -refresh=falseno longer warns that an address is held by a server that is already gone; the firewall refusal names the way out forterraform destroy.
Every change is in the reference.
Upgrading to 0.8
- No state migration. A configuration that worked on 0.7.x and creates servers the platform accepts keeps working.
- One exception: a data disk taken over with
terraform importor written by 0.5.1, grown outside Terraform and with that size already in state, gets its record at the current size; a configuration with the smaller size then gets a plan error instead of a warning. Writesize_gbas the platform has it. - Change the constraint to
~> 0.8.0and runterraform init -upgrade.
What is new in 0.7.5
A patch version: ~> 0.7.0 picks it up.
terraform destroyno longer stops without a way out on a server that is gone. When the configuration describes a server withoutpublic_ipv4and withoutnetworks(the recipe for keeping the address of a server created withnew = true) and that server is no longer in state - after a firstdestroywithoutdepends_onstopped onip_in_use, afterterraform destroy -target=fiberax_server.<name>, or after the server was deleted outside Terraform -destroyfailed with "The server has no network", and the address imported as afiberax_ipstayed with the project and was billed. Every such refusal now names the way out:terraform destroy -refresh=false(tofu destroy -refresh=falsein OpenTofu), or remove the server from the configuration.- Replacing a server that took its address with
new = truenow warns with the order that keeps THIS address: do not apply the plan; remove thepublic_ipv4block, changing nothing else, and apply; import the address as afiberax_ip; then make the change together withpublic_ipv4 = { id = fiberax_ip.<name>.id }- the new server takes the same address. - Checksums verify. The checksum file of every version names a
_manifest.json; it is now on the mirror for all versions, sosha256sum -con a downloaded set ends with success.
What is new in 0.7.4
A patch version: ~> 0.7.0 picks it up. Only texts changed - documentation; what the
provider does is the same as in 0.7.3.
- The price reference matches the API. The
fiberax_server_type_pricingreference and its example no longer describeplan_<key>lines inmetered: the API stopped returning them on 2026-10-05 (see the changelog), and the price of a plan is inplans.backup_gbis the only line ofmeteredthat is not a price; on a type that sells GPU packages backups cost the package's flatbackup_price, not 20% of what the server costs. The rate of a GPU package is keyed by the package key with hyphens as underscores (gpu-l4-24-gpu_l4_24) and can belong to a package that is no longer sold: servers already on it keep paying that rate. - Changelog in the menu. The reference pages carry the Changelog item of the portal menu.
What is new in 0.7.3
A patch version: ~> 0.7.0 picks it up.
- No false warnings on the recipe's destroy.
terraform destroyof the recipe for keeping the address of a server created withnew = true(the server without the block, the address imported as afiberax_ip,depends_onon the server) no longer warns that the address stayed with the project or that it must be taken offnet0first: thefiberax_ipreleases it in the same run. Addresses named withid = ...(afiberax_ipreference or a literal number) are not listed, as before: you chose them; they stay with the project and are billed until released. When the server and thefiberax_ipare removed from the configuration at once, the warnings may still appear; the address is released all the same. - Replacing a server with
new = true. The old address goes back to the shared pool first, and the new server takes a free address from that pool, which may or may not be the same one - do not rely on keeping it.-replaceandtaintreplace without the plan warning: the provider cannot tell these replacements apart when it plans them. - Advice to run
terraform importalso namestofu importfor OpenTofu; the refusal ofnew->reusenames the address ID. - Replacing a tainted server no longer reads as "another server with this name already exists" without explanation.
What is new in 0.7.2
A patch version: only texts changed - documentation and error messages; what the provider does is the same as
in 0.7.1, and ~> 0.7.0 picks it up.
- Releasing only the address of a server. Terraform cannot take an address off
net0, the card the server was created with: it is not a resource, and removing thepublic_ipv4block changes state only. The guide Public addresses and what they cost has a new section with the steps: remove the dependency and apply (until an apply records that,terraform destroy -targetstill deletes the server), take the address offnet0withPATCH /v1/servers/{id}/network-interfaces/net0and{"ips": []}(or{"network_id": N}, or in the client area) - a running server restarts once and is left without a public IPv4 address - then remove thefiberax_ipfrom the configuration and apply; a server withnew = truehas nofiberax_ipfor its address: release it withDELETE /v1/ips/{id}, or import it first (afiberax_ipwritten without import takes a new, billed address). The plan warning and the error for an address still held by a server say the same. - The import allowance, precisely. The first successful apply whose plan shows a change for the imported server ends it, even when that change only records values in state; an apply with no change for the server - a refresh, changes to other resources - does not. The plan does not warn that the allowance is spent: a replacement shows only as the usual "forces replacement".
- The addresses guide names the API's answer when an address is still on a server - 409
ip_in_use- next to Terraform's heading "The address is held by a server".
What is new in 0.7.1
A patch version: only texts changed - documentation and plan error messages; what the provider does is the same
as in 0.7.0, and ~> 0.7.0 picks it up.
- The recipe for keeping a
new = trueaddress no longer replaces the server. The plan errors and the guide Public addresses and what they cost now say: remove thepublic_ipv4block (the address is kept from then on, still billed until you release it) and, to manage it,terraform import fiberax_ip.<name> <id>and give the serverdepends_on = [fiberax_ip.<name>]- the server is not touched, anddestroydeletes the server before releasing the address (withoutdepends_onboth go at once and releasing the address fails until a seconddestroy). Like a reference frompublic_ipv4 = { id = ... }, thatdepends_onmakesterraform destroy -target=fiberax_ip.<name>delete the server too: to release only the address later, first removedepends_onand apply, take the address off the server's cardnet0outside Terraform (PATCH /v1/servers/{id}/network-interfaces/net0with{"ips": []}, or the client area; a running server restarts once), then removefiberax_ipfrom the configuration and apply (withnew = truethere is nofiberax_ip: release the address withDELETE /v1/ips/{id}). Writingpublic_ipv4 = { id = ... }on that server afterwards replaces it and erases its disks; do it only when the server is to be replaced anyway. - Imported or created. On a server taken over with
terraform import, apublic_ipv4block (ornetworks,host,accept_license,gpus) may be added without replacing it only in the first apply that records anything for it after the import - that apply ends the allowance even when it only records values in state. On a server Terraform created - without the block, or after it was removed - adding the block replaces the server. The import guide, the addresses guide and thepublic_ipv4reference say so. - A switch from
new = truetoidof another address returns the old address to the shared pool; fromreuseoridthe old address stays with the project, billed.
What is new in 0.7.0
A new minor version: destroying a server now changes what happens to its public address. A configuration
pinned to ~> 0.6.0 keeps getting 0.6.0 until you change the constraint to ~> 0.7.0
(see Upgrading to 0.7). Upgrade before you destroy or replace a server whose
address you want to keep: the platform now releases a deleted server's public addresses by default, and
0.6.0 and earlier do not tell it otherwise.
- An address taken with the server is released with it. Destroying (or replacing) a server created
with
public_ipv4 = { new = true }releases its public address to the platform's shared pool: it is no longer left in the project and billed. The platform releases every public address attached to the server at that moment, and a released address cannot be taken back. - Any other address stays. A server with
public_ipv4 = { id = ... }or{ reuse = true }, and a server without the block (an imported one included), keeps its addresses with the project, still billed until you release them: the provider asks the platform for that explicitly (release_ips=false) on every destroy. - A replacement that would lose the address is refused at plan time. A plan that replaces a server
created with
new = trueand names the same address with{ id = ... }, or switches it to{ reuse = true }, is refused: the replacement would release that address before the new server could take it. To keep the address, remove the block (the address then stays with the project, still billed) and, to manage it, import it withterraform import fiberax_ip.<name> <id>and give the serverdepends_on = [fiberax_ip.<name>]- the server is not replaced, anddestroydeletes the server before releasing the address; see the guide Public addresses and what they cost. The recipe in the 0.7.0 error text writes{ id = ... }back on the server as a second step: on a server Terraform created that replaces the server and erases its disks - 0.7.1 corrects it. - Warnings tell what happens to the addresses: when a change in the configuration replaces a server
created with
new = true(its address goes back to the pool first, and the new server takes a free address from the pool, which may or may not be the same one;-replaceandtaintreplace without this warning); when the server was already being deleted elsewhere and that deletion's choice about the addresses differs from the configuration; and when an address the platform was to release is still in the project after the server is gone (it stays billed until you release it).
Upgrading to 0.7
- No state migration. A plan of an unchanged configuration is empty after the upgrade.
- Upgrade first, destroy later. With 0.6.0 or earlier, destroying or replacing a server releases
EVERY public address it has, whatever the form - one named with
id(itsfiberax_ipthen points at an address the project no longer holds), one taken withreuse, and the addresses of a server without the block or an imported one. - Change the constraint to
~> 0.7.0and runterraform init -upgrade.
What is new in 0.6.0
A new minor version: a configuration pinned to ~> 0.5.0 keeps getting 0.5.1 until you change
the constraint to ~> 0.6.0 (see Upgrading to 0.6). It is a minor
version because the provider now refuses more at plan time.
- A server order the platform would refuse is refused at plan time, when a server is created or
recreated: a server type the project cannot see, an image that is not in the type's catalogue, an ISO image,
plan_idon a type sized by sliders,cores/memory_mb/disk_gbor noplan_idon a type sold as plans, an unknown plan, a GPU package type withoutgpu_package. Before, these failed onapply, after neighbouring resources of the run (a billed address, a network) had already been created. - A plan change the platform refuses no longer stops the server. A smaller plan or one that does not exist is refused in a second while the server keeps running. The server is stopped only for a real upgrade, and an interrupted stop no longer leaves it powered off.
- Deleting a server that is still being created waits for the creation to finish, up to
timeouts.delete, instead of failing after four retries. - Adding a disk to a running server warns at plan time that the server restarts.
- A disk grown outside Terraform no longer blocks the plan and
terraform destroy. The plan warns and names the value to write; applying that disk fails without touching it. - Warnings tell the truth. Destroying a backup policy together with its server no longer says that
backups keep being billed; a policy of a server created in the same plan no longer names "server 0"; a
network card recreated with its server no longer promises a restart;
gpu_packagetogether withcoresgets one error instead of advice to add two more forbidden fields. - Reference. Router access rules see the port after the port forward and do not filter traffic to
the router itself; the server firewall applies an address to its own family (
0.0.0.0/0is IPv4 only, ping over IPv6 isipv6-icmp);plan_*andbackup_gblines of the pricing are billing multipliers, not prices.
Known limitation. A disk written by 0.5.1 or imported has no record of the size the provider
applied. Until the provider changes that disk once, reducing its size_gb in the configuration
gives a plan-time warning that says the disk was grown outside Terraform (0.5.1 refused the plan); the apply
then fails on that disk and changes nothing on it, while other resources of the run are applied. A disk can
only grow.
Upgrading to 0.6
- No state migration. A plan of an unchanged configuration is empty after the upgrade, and going back to 0.5.1 is safe.
- Minimum version: Terraform 1.11 or OpenTofu 1.11.
fiberax_server.passwordandsetupare write-only arguments, which older versions reject with "WriteOnly Attribute Not Allowed"; a Windows server cannot be described withoutpassword. Putrequired_version = ">= 1.11"in theterraformblock. - Change the constraint to
~> 0.6.0and runterraform init -upgrade.
Upgrading to 0.5
0.5.0 changes what some configurations must say, so it is a new minor version: a configuration
pinned to ~> 0.4.0 keeps getting 0.4.1 until you change the constraint to
~> 0.5.0. Two things to check before you do:
- A switched-on firewall names both policies. A
fiberax_server_firewallwithenabled = trueneedspolicy_inandpolicy_outwritten out - the API itself refuses to switch filtering on with a policy nobody named, and the provider no longer fills indropandacceptfor you. Withenabled = falsean omitted policy is now left as it is on the platform; earlier versions wrotedropthere, which cut the server off the next time the firewall was switched on in the client area. Adropwritten that way stays on the platform - 0.5 does not change a stored policy by itself. For each server whose firewall was switched off under 0.4 without policies, check the policy in the client area or withGET /v1/servers/{id}/firewall, and where you wantaccept, writepolicy_in = "accept"in the configuration. - A card turned public names its address. Moving an existing
fiberax_network_interfaceto the public network withoutipsis refused at plan time: the platform would allocate a new, billed address that the configuration does not hold. Add afiberax_ipand list itsidinips.
Both show up as an error on terraform plan, before anything is changed - the firewall one on
a plan that creates or changes the firewall. Since 0.5.1 a firewall written under 0.4 and left unchanged
plans and destroys as it is; name the policies the next time you edit it.
The normal way: let Terraform download it
Add nine lines to your Terraform or OpenTofu configuration file and terraform init fetches
the provider from here with a verified checksum. The lines, the example and the one warning you will see
are in Getting started, section 7.
Reference
Every resource and data source with its fields, examples and import syntax: Terraform provider reference.
Direct downloads
Each archive contains one file, the provider binary terraform-provider-fiberax_v0.11.0.
- Linux amd64 (x86-64) 7.6 MB
- Linux arm64 6.9 MB
- Linux 386 (x86 32-bit) 7.2 MB
- Linux arm (32-bit) 7.2 MB
- macOS amd64 (x86-64) 7.8 MB
- macOS arm64 7.3 MB
- Windows amd64 (x86-64) 7.9 MB
- Windows arm64 7.0 MB
- Windows 386 (x86 32-bit) 7.6 MB
- FreeBSD amd64 (x86-64) 7.6 MB
- FreeBSD arm64 6.9 MB
- FreeBSD 386 (x86 32-bit) 7.2 MB
- FreeBSD arm (32-bit) 7.2 MB
Checksums of all archives: terraform-provider-fiberax_0.11.0_SHA256SUMS.
Earlier versions
Versions 0.10.0, 0.9.1, 0.9.0, 0.8.0, 0.7.5, 0.7.4, 0.7.3, 0.7.2, 0.7.1, 0.7.0, 0.6.0, 0.5.1, 0.5.0, 0.4.1, 0.4.0, 0.3.0, 0.2.0 and 0.1.0 stay on the mirror for configurations and lock files that
pin them, and Terraform fetches them the usual way; new configurations should use 0.11.0. To download one
by hand, archives are named terraform-provider-fiberax_<version>_<os>_<arch>.zip
under /terraform/mirror/archives/, with the same platforms as above. Checksums:
- terraform-provider-fiberax_0.10.0_SHA256SUMS
- terraform-provider-fiberax_0.9.1_SHA256SUMS
- terraform-provider-fiberax_0.9.0_SHA256SUMS
- terraform-provider-fiberax_0.8.0_SHA256SUMS
- terraform-provider-fiberax_0.7.5_SHA256SUMS
- terraform-provider-fiberax_0.7.4_SHA256SUMS
- terraform-provider-fiberax_0.7.3_SHA256SUMS
- terraform-provider-fiberax_0.7.2_SHA256SUMS
- terraform-provider-fiberax_0.7.1_SHA256SUMS
- terraform-provider-fiberax_0.7.0_SHA256SUMS
- terraform-provider-fiberax_0.6.0_SHA256SUMS
- terraform-provider-fiberax_0.5.1_SHA256SUMS
- terraform-provider-fiberax_0.5.0_SHA256SUMS
- terraform-provider-fiberax_0.4.1_SHA256SUMS
- terraform-provider-fiberax_0.4.0_SHA256SUMS
- terraform-provider-fiberax_0.3.0_SHA256SUMS
- terraform-provider-fiberax_0.2.0_SHA256SUMS
- terraform-provider-fiberax_0.1.0_SHA256SUMS
Manual installation
If you prefer not to use the mirror settings, put the archive for your platform into Terraform's local plugin directory, unchanged, under its registry path - for example on Linux:
~/.terraform.d/plugins/registry.terraform.io/fiberax/fiberax/terraform-provider-fiberax_0.11.0_linux_amd64.zip
On Windows the directory is %APPDATA%\terraform.d\plugins\; OpenTofu reads the same
layout under registry.opentofu.org/. Terraform picks the provider up from there on
init without any further configuration.
Mirror index files for tools live under
registry.terraform.io/fiberax/fiberax/ and
registry.opentofu.org/fiberax/fiberax/.
Something not working? Write to us and quote what terraform init printed.