Silo Management

Silos are designed to provide strict tenancy separation. They have separate UI and API endpoints, separate resource limits (IP address pools, compute and storage quotas), and separate IAM policies. You can choose to set up one or multiple silos depending on your organization resource and access management policies.

There are two types of silos in the Oxide rack: system silos and user silos. The only instance of system silos is the recovery silo which is created during rack setup. The silo hosts a system account which is also named recovery. This built-in account has the fleet admin role that allows the operator to create user-defined silos and grant the fleet admin privilege to other users. After the rack setup process, the recovery silo will only be used for rescue purposes such as remediating identity provider integration configuration problems.

Rack administrators and users are typically onboarded into the user silos via identity provider (IdP) import. User silos can also be configured to have local rack accounts created manually by the silo administrators.

Silos vs Projects as Tenants

Both silos and projects can be used as the abstraction for tenants. While projects provide a level of isolation between different application workloads, when the tenant is in an entirely different security domain (e.g. a subsidiary, a third-party company), it may be more appropriate to place their workload in a separate silo so that they are prevented from accessing other projects altogether.

Preparations for Silo Setup

Generate a TLS Certificate

A TLS certificate is required during silo setup and will be used for TLS termination on the silo-specific UI/API endpoint. The certificate can reference either the fully-qualified domain name (FQDN), i.e.


or a one-level wildcard domain name, i.e.


Configure Identity Provider

The Oxide control plane currently supports only SAML authentication with third-party identity provider systems using just-in-time (JIT) provisioning. This means that user accounts and groups are created or updated immediately after users have successfully authenticated themselves.

If you choose to delegate silo user management to an identity provider, you will need to identify the IdP group, role, or attribute to be used for governing memberships of the new silo and users who are designated as administrators.

The process for configuring silo IdP integration is the same as the one used for initial rack setup. Please refer to Configure Silo Identity Provider for more details about the integration settings.

Determine which IP pools to use

You may configure the new silo to have its own IP pools or use existing pools for external IP address assignment. If you choose to provide the silo with a new IP pool, you will create the pool and allocate IP ranges to it prior to setting up the silo. If the new silo will be using an existing pool, you may need to review the size of that IP pool and add capacity as needed.

You can find out more about IP pools in the IP Pool Management guide.

Determine Resource Limits

Virtual compute and storage resources such as vcpus and virtual disks make use of physical resources in the rack. They are shared among users of different silos. To ensure individual silos can get the resources guaranteed to them, you will establish the appropriate resource limits as part of the silo setup. Currently there are three kinds of resource quotas: vcpu, memory, storage. Their utilization calculations are defined as follows:

  • vcpu, in number of cores: utilization = (aggregated ncpus of powered-on instances owned by the silo) / (silo vcpu quota)

  • memory, in Gibibytes: utilization = (aggregated memory size of powered-on instances owned by the silo) / (silo memory quota)

  • storage, in Gibibytes: utilization = (aggregated size of disks owned by the silo) / (silo storage quota)

Powered-on states include running, starting, stopping, and rebooting.
Snapshots and images are not counted towards the storage quota enforcement currently. This may change in the future when usage metering is enabled for these objects.

The sum of the silo quotas should ideally be below the total usable rack capacity to avoid over-provisioning. If silo quotas are set above the rack capacity and an instance or disk request cannot be accommodated by any sleds in the rack, the rack-level capacity check will kick in to reject the request.

It is a best practice to always maintain capacity utilization of an Oxide rack below 70%. Apart from resources consumed by the systems software, there are other conditions that require a 15-30% buffer of compute resources:

  1. Instance placement currently follows a random sled selection strategy. This reduces the risk of application outage when instances are spread across different sleds. The strategy may, however, result in the inability to place a large instance when all sleds are relatively full. If you have large instances that are dynamically brought up and down, you may want to reserve a larger percentage of unallocated compute resources.

  2. Usable capacity may also be reduced when some of the sleds are taken out of service due to planned maintenance (e.g. software update) or unplanned outage.

Storage resources may be subject to contention to a lesser extent. The usable capacity represented in the system utilization view takes into account the 3x data replication factor and a fixed percentage overhead for snapshots and images.

Setting up a Silo

The procedures for setting up new silos and identity provider integration are similar to those during rack initialization. Operators who have the fleet-admin role can use the Oxide Console, API, or CLI to create a new silo and configure its resource and integration settings.

1. Create Silo and Upload TLS Certificate

The first step of silo setup is creating the silo record and importing a matching TLS certificate. Virtual resource quotas are required during silo creation and can be modified afterwards. Please refer to the Create a User Silo section for all the attributes involved and an example API request.

2. Create identity provider (optional)

Perform this step only if your silo uses the saml_jit authentication mode.

In this step, you’ll create the identity provider (IdP) configuration to integrate Oxide rack user authentication and authorization with the IdP. An example of the API request can be found in Configure Silo Identity Provider.

You can verify the IdP setup by logging into the Oxide console as any of the users in the silo admin group specified in the IdP configuration. The new silo URL will be at: https://$siloName.sys.$oxideDomainName/.

3. Create silo administrator account (optional)

Perform this step only if your silo uses the local_user authentication mode. An example of the API requests for creating the user account and granting administrator access can be found at Create Local Users.

4. Configure IAM policies (optional)

If the silo is set up for IdP integration, the admin role is automatically granted to members of the admin group specified in the IdP configuration. You may grant other access levels via the collaborator and viewer roles to other users via the Console, API, or CLI.

5. Link silo to IP pools

Each silo can be linked to one or more IP pools and one of them should be marked as the default. Only linked IP pools can be used for automatic address assignment when ephemeral or floating IPs are created for VM instances.

Silo Resource Quota Management

The vcpu, memory and disk resources allowed for a silo can be adjusted up or down any time. Modifying quotas to levels below the current utilization does not impact resources already in use. The new limits only take effect when users attempt to make new instances, disks, or snapshots. Here are a few CLI examples for modifying resource quotas:

oxide silo quotas update --silo $siloName --cpus $vcpuCount --memory $memoryInBytes --storage $storageInBytes

# To view the current quota and utilization of a certain silo
oxide silo quotas view --silo $siloName

# To view the current quota and utilization of all silos
oxide silo utilization list

Related API Reference

The following APIs are accessible to fleet administrator only:

Silo Administrator Capabilities

Silo Resource Access

Silo administrators can access and act on all the child resources within the silo, including projects, instances, images, disks, snapshots, and VPCs. They can also modify the IAM policies of the projects to grant or revoke access permissions.

Silo administrators can view the capacity allocated to the silo as well as the current resources utilization on the console

Silo Capacity Utilization

They can also get the available resources, including IP pools, via API:

When users provision more disks or start up more instances than permitted by the resource quotas or IP pools, they will receive a capacity validation exception. Silo administrators can either request the fleet administrator to increase the resources allocated to the silo, or advise users in the silo to remove unused disks or running instances to free up compute, storage, and network resources.

Silo Certificate Replacement

Silo administrators have the permission to manage TLS certificates for their silos.

To replace the existing TLS certificate for a silo:

  1. If the administrator has access to more than one silo, first configure the OXIDE_HOST and OXIDE_TOKEN environment variables that correspond to the target silo.

  2. Import the new certificate using certificate_create. Pass the --insecure or -k argument if the existing certificate already expired.

  3. Remove the old certificate using certificate_delete.

If a wildcard certificate is used, ensure that .sys is included in the CN before the subdomain name, i.e. *.sys.$oxideDomainName, to keep the wildcard to a single level.
There should be at least one valid TLS certificate associated with each silo for the silo API/Console endpoint to be accessible. Always create a new certificate before removing the old one.
Last updated