Local VCN Peering using Local Peering Gateways

This topic is about local VCN peering. In this case, local means that the VCNs reside in the same region. If the VCNs are in different regions, see Remote VCN Peering using a Legacy DRG.

Local peering gateways are still supported. This scenario assumes you are using a legacy DRG. Oracle currently recommends routing traffic from one VCN to another through an upgraded DRG as described in Peering VCNs in the same region through a DRG.

Overview of Local VCN Peering

Local VCN peering is the process of connecting two VCNs in the same region so that their resources can communicate using private IP addresses without routing the traffic over the internet or through your on-premises network. The VCNs can be in the same Oracle Cloud Infrastructure tenancy  or different ones. Without peering, a given VCN would need an internet gateway and public IP addresses for the instances that need to communicate with another VCN.

See Gateway Limits and Requesting a Service Limit Increase for limits-related information.

For more information, see Access to Other VCNs: Peering.

Summary of Networking Components for Peering using an LPG

At a high level, the Networking service components required for a local peering include:

  • Two VCNs with non-overlapping CIDRs, in the same region
  • A local peering gateway (LPG) on each VCN in the peering relationship.
  • A connection between those two LPGs.
  • Supporting route rules to enable traffic to flow over the connection, and only to and from select subnets in the respective VCNs (if wanted).
  • Supporting security rules to control the types of traffic allowed to and from the instances in the subnets that need to communicate with the other VCN.

The following diagram illustrates the components.

This image shows the basic layout of two VCNs that are locally peered, each with a local peering gateway.
Note

A given VCN can use the peered LPGs to reach these resources:

  • VNICs in the other VCN
  • An on-premises network attached to the other VCN, if an advanced routing scenario called transit routing has been set up for the VCNs

A VCN can't use its peered VCN to reach other destinations outside of the VCNs (such as the internet). For example, if VCN-1 in the preceding diagram were to have an internet gateway, the instances in VCN-2 could not use it to send traffic to endpoints on the internet. However, VCN-2 could receive traffic from the internet by way of VCN-1. For more information, see Important Implications of VCN Peering.

Explicit Agreement Required from Both Sides

Peering involves two VCNs that might be owned by the same party or two different ones. The two parties might both be in your company but in different departments. Or the two parties might be in entirely different companies (for example, in a service-provider model).

Peering between two VCNs requires explicit agreement from both parties in the form of Oracle Cloud Infrastructure Identity and Access Management policies that each party implements for their own VCN's compartment  or tenancy. If the VCNs are in different tenancies, each administrator must provide their tenancy OCID and put in place special policy statements to enable the peering.

Advanced Scenario: Transit Routing

There's an advanced routing scenario called transit routing that enables communication between an on-premises network and multiple VCNs over a single Oracle Cloud Infrastructure FastConnect or Site-to-Site VPN. The VCNs must be in the same region and locally peered in a hub-and-spoke layout. As part of the scenario, the VCN that is acting as the hub has a route table associated with each LPG (typically route tables are associated with a VCN's subnets).

When you create an LPG, you can optionally associate a route table with it. Or if you already have an existing LPG without a route table, you can associate a route table with it. The route table must belong to the LPG's VCN. A route table associated with an LPG can contain only rules that use the VCN's attached DRG as the target.

An LPG can exist without a route table associated with it. However, after you associate a route table with an LPG, there must always be a route table associated with it. But, you can associate a different route table. You can also edit the table's rules, or delete some or all rules.

Important Local Peering Concepts

The following concepts help you understand the basics of VCN peering and how to establish a local peering.

PEERING
A peering is a single peering relationship between two VCNs. Example: If VCN-1 peers with three other VCNs, then there are three peerings. The local part of local peering indicates that the VCNs are in the same region. A given VCN can have a maximum of 10 local peerings at a time.
Caution

The two VCNs in the peering relationship must not have overlapping CIDRs. However, if VCN-1 is peered with three other VCNs, those three VCNs can have overlapping CIDRs with each other. You would set up the subnets in VCN-1 to have route rules that direct traffic to the targeted peered VCN.
VCN ADMINISTRATORS
In general, VCN peering can occur only if both of the VCN administrators agree to it. In practice, this means that the two administrators must:
  • Share some basic information with each other.
  • Coordinate to set up the required Oracle Cloud Infrastructure Identity and Access Management policies to enable the peering.
  • Configure their VCNs for the peering.
Depending on the situation, a single administrator might be responsible for both VCNs and the related policies.
For more information about the required policies and VCN configuration, see Setting Up a Local Peering.
ACCEPTOR AND REQUESTOR
To implement the IAM policies required for peering, the two VCN administrators must designate one administrator as the requestor and the other as the acceptor. The requestor must be the one to initiate the request to connect the two LPGs. In turn, the acceptor must create a particular IAM policy that gives the requestor permission to connect to LPGs in the acceptor's compartment . Without that policy, the requestor's request to connect fails.
LOCAL PEERING GATEWAY (LPG)
A local peering gateway (LPG) is a component on a VCN for routing traffic to a locally peered VCN. As part of configuring the VCNs, each administrator must create an LPG for their VCN. A given VCN must have a separate LPG for each local peering it establishes (maximum 10 LPGs per VCN). To continue with the previous example: VCN-1 would have three LPGs to peer with three other VCNs. In the API, a LocalPeeringGateway is an object that contains information about the peering. You can't reuse an LPG to later establish another peering.
PEERING CONNECTION
When the requestor initiates the request to peer (in the Console or API), they're effectively asking to connect the two LPGs. The requestor must have information to identify each LPG (such as the LPG's compartment and name, or the LPG's OCID). Each administrator must put the required IAM policies in place for their compartment or tenancy.
Either VCN administrator can terminate a peering by deleting their LPG. In that case, the other LPG's status switches to REVOKED. The administrator could instead render the connection non-functional by removing the route rules or security rules that enable traffic to flow across the connection (see the next sections).
ROUTING TO THE LPG
As part of configuring the VCNs, each administrator must update the VCN's routing to enable traffic to flow between the VCNs. In practice, this looks just like routing you set up for any gateway (such as an internet gateway or dynamic routing gateway). For each subnet that needs to communicate with the other VCN, you update the subnet's route table. The route rule specifies the destination traffic's CIDR and your LPG as the target. Your LPG routes traffic that matches that rule to the other LPG, which in turn routes the traffic to the next hop in the other VCN.
In the following diagram, VCN-1 and VCN-2 are peered. Traffic from an instance in Subnet A (10.0.0.15) that is destined for an instance in VCN-2 (192.168.0.15) is routed to LPG-1 based on the rule in Subnet A's route table (see Callout 1: Subnet A Route Table). From there the traffic is routed to LPG-2, and then from there, on to its destination in Subnet X.
This image shows the path of traffic routed from one local peering gateway to the other.
Callout 1: Subnet A Route Table
Destination CIDR Route Target
0.0.0.0/0 Internet Gateway
172.16.0.0/12 DRG
192.168.0.0/16 LPG-1
Callout 2: Subnet X Route Table
Destination CIDR Route Target
10.0.0.0/16 LPG-2
Note

As mentioned earlier, a given VCN can use the peered LPGs to reach VNICs in the other VCN, or the on-premises network if transit routing is set up for the VCNs. But a VCN can't use the peered VCN to reach other destinations outside of the VCNs (such as the internet). For example, in the preceding diagram, VCN-2 cannot use the internet gateway attached to VCN-1.

SECURITY RULES
Each subnet in a VCN has one or more security lists that control traffic in and out of the subnet's VNICs at the packet level. You can use security lists to control the type of traffic allowed with the other VCN. As part of configuring the VCNs, each administrator must determine which subnets in their own VCN need to communicate with VNICs in the other VCN and update their subnet's security lists accordingly.
If you use network security groups (NSGs) to implement security rules, notice that you have the option to write security rules for an NSG that specify another NSG as the source or destination of traffic. However, the two NSGs must belong to the same VCN.

Important Implications of VCN Peering

If you haven't yet, read Important Implications of Peering to understand important access control, security, and performance implications for peered VCNs.

Setting Up a Local Peering

Here's the general process for setting up a peering between two VCNs in the same region:

  1. Create the LPGs: Each VCN administrator creates an LPG for their own VCN.
  2. Share information: The administrators share the basic required information.
  3. Set up the required IAM policies for the connection: The administrators set up IAM policies to enable the connection to be established.
  4. Establish the connection: The requestor connects the two LPGs.
  5. Update route tables: Each administrator updates their VCN's route tables to enable traffic between the peered VCNs as wanted.
  6. Update security rules: Each administrator updates their VCN's security rules to enable traffic between the peered VCNs as wanted.

If wanted, the administrators can perform tasks E and F before establishing the connection. In that case, each administrator must know the CIDR block or specific subnets from the other's VCN and share that in task B. After the connection is established, you can also get the CIDR block of the other VCN by viewing your own LPG's details in the Console. Look for Peer Advertised CIDR. Or if you're using the API, see the peerAdvertisedCidr parameter.

You will also need to pre-configure some IAM settings like groups prior to going through the step-by-step process.

Task B: Share information

If you're the requestor, give this information to the acceptor (for example, by email or other out-of-band method):

  • If the VCNs are in the same tenancy: Name of the IAM group that should be granted permission to create a connection in the acceptor's compartment. In the example in the next task, the group is RequestorGrp.
  • If the VCNs are in different tenancies: OCID for your tenancy, and OCID for the IAM group that should be granted permission to create a connection in the acceptor's compartment. In the example in the next task, it's the OCID for the RequestorGrp.
  • Optional: Your VCN's CIDR, or specific subnets for peering with the other VCN.

If you're the acceptor, give this information to the requestor:

  • If the VCNs are in the same tenancy: OCID for your LPG. Optionally, also the names of your VCN, LPG, and the compartment each is in.
  • If the VCNs are in different tenancies: OCID for your LPG, and OCID for your tenancy.
  • Optional: Your VCN's CIDR, or specific subnets for peering with the other VCN.
Task C: Set up the IAM policies (VCNs in same tenancy)

In this version of task C, both VCNs are in the same tenancy. If they're in different tenancies, instead see Task C: Set up the IAM policies (VCNs in different tenancies).

Both the requestor and acceptor must ensure that the right policies are in place:

  • Policy R (implemented by the requestor):

    Allow group <requestor-group-name> to manage local-peering-from in compartment RequestorComp

    The requestor is in an IAM group called <requestor-group-name>. This policy lets anyone in the group initiate a connection from any LPG in the requestor's compartment (<requestor-compartment>). Policy R can be attached to either the tenancy (root compartment) or to <requestor-compartment>. For information about why you would attach it to one versus the other, see Policy Basics.

  • Policy A (implemented by the acceptor):

    Allow group <requestor-group-name> to manage local-peering-to in compartment <acceptor-compartment>
    
    Allow group <requestor-group-name> to inspect vcns in compartment <acceptor-compartment>
    
    Allow group <requestor-group-name> to inspect local-peering-gateways in compartment <acceptor-compartment>
    

    The first statement in the policy lets the requestor connect to any LPG in the acceptor's compartment (<acceptor-compartment>). This statement reflects the required agreement from the acceptor for the peering to be established. Policy A can be attached to either the tenancy (root compartment) or to <acceptor-compartment>.

    Tip

    The second and third statements in Policy A let the requestor list the VCNs and LPGs in <acceptor-compartment>. The statements are required for the requestor to use the Console UI to select from a list of VCNs and LPGs in <acceptor-compartment> and establish the connection. The following diagram focuses only on the first statement, which is the critical one that permits the connection.
This image shows the two policies for VCNs in the same tenancy.

Both Policy R and Policy A give RequestorGrp access. However, Policy R has a resource-type called local-peering-from, and Policy A has a resource-type called local-peering-to. Together, these policies let someone in RequestorGrp establish the connection from an LPG in the requestor's compartment to an LPG in the acceptor's compartment. The API call to create the connection specifies which two LPGs.

Tip

The permission granted by Policy R might already be in place if the requestor has permission in another policy to manage all Networking components in RequesterComp. For example, there might be a general Network Admin policy like this:

 Allow group NetworkAdmin to manage virtual-network-family in compartment RequestorComp

If the requestor is in the NetworkAdmin group, then they already have the required permissions covered in Policy R (the virtual-network-family includes LPGs). And further, if the policy is instead written to cover the entire tenancy instead of only compartment RequestorComp, then the requestor already has all the required permissions in both compartments to establish the connection. In that case, policy A is not required.

Task C: Set up the IAM policies (VCNs in different tenancies)

In this version of task C, the VCNs are in different tenancies (in other words, it's a cross-tenancy peering). If the VCNs are in the same tenancy, instead see Task C: Set up the IAM policies (VCNs in same tenancy).

Both the requestor and acceptor must ensure that the right policies are in place:

  • Policy R (implemented by the requestor):

    Define tenancy Acceptor as <acceptor_tenancy_OCID>
    
    Allow group <requestor-group-name> to manage local-peering-from in compartment <requestor-compartment>
    Endorse group <requestor-group-name> to manage local-peering-to in tenancy Acceptor
    
    Endorse group <requestor-group-name> to associate local-peering-gateways in compartment <requestor-compartment>
       with local-peering-gateways in tenancy Acceptor

    The requestor is in an IAM group called <requestor-group-name>. This policy lets anyone in that group initiate a connection from any LPG in the requestor's compartment (<requestor-compartment>).

    The first statement is a "define" statement that assigns a friendly label to the acceptor's tenancy OCID. The statement happens to use "Acceptor" as the label, but it could be a value of the requestor's choice. All "define" statements in a policy must be the first ones (at the top).

    The second statement lets the <requestor-group-name> establish a connection from an LPG in the requestor's compartment.

    The third and fourth statements are special ones required because the LPGs are in different tenancies. They let the <requestor-group-name> connect an LPG in the requestor's tenancy to an LPG in the acceptor's tenancy.

    If the intent is to give the <requestor-group-name> permission to connect to an LPG in any tenancy, the policy would instead look like this:

    
    Allow group <requestor-group-name> to manage local-peering-from in compartment <requestor-compartment>
    
    Endorse group <requestor-group-name> to manage local-peering-to in 
    
    any-tenancy
    
    
    
    Endorse group <requestor-group-name> to associate local-peering-gateways in compartment <requestor-compartment> 
    with local-peering-gateways in 
    any-tenancy
    
    

    Regardless, Policy R must be attached to the requestor's tenancy (root compartment), and not the requestor's compartment. Policies that enable cross-tenancy access must be attached to the tenancy. For more information about attachment of policies, see Policy Basics.

  • Policy A (implemented by the acceptor):

    Define tenancy Requestor as <requestor_tenancy_OCID>
    
    Define group <requestor-group-name> as <RequestorGrp_OCID>
    
    Admit group <requestor-group-name> of tenancy Requestor to manage local-peering-to in compartment <acceptor-compartment>
    
    Admit group <requestor-group-name> of tenancy Requestor to associate local-peering-gateways in tenancy Requestor
        with local-peering-gateways in compartment <acceptor-compartment>
    

    Similar to the requestor's policy, this policy first uses "define" statements to assign friendly labels to the requestor's tenancy OCID and the <requestor-group-name> OCID. As mentioned earlier, the acceptor could use other values for those labels if wanted.

    The third and fourth statements let the <requestor-group-name> connect to an LPG in the acceptor's compartment (<acceptor-compartment>). These statements reflect the critical agreement required for the peering to be established. The word Admit indicates that the access applies to a group outside the tenancy where the policy resides.

    Policy A must be attached to the acceptor's tenancy (root compartment), and not the acceptor's compartment.

This image shows the location of the two policies for VCNs in different tenancies.
Task E: Configure the route tables

Configure the route tables to use the information for the other VCN that you were given in Task B: Share information, using the instructions in Configuring VCN Route Tables to Use an LPG.

Task F: Configure the security rules

Configure the security rules to use the information for the other VCN that you were given in Task B: Share information, using the instructions in Configuring Security Rules to Use an LPG.