1.8.0

Configure Access Gateway for High-Availability

This document outlines the necessary steps to deploy and configure a Magma access gateway on AWS. This document also outlines configuring the AWS gateway to serve as a secondary to a primary gateway running at an edge site.

Deployment

Build AGW AMI

Steps:

  1. Download packer onto your host machine at https://www.packer.io/downloads.html
  2. Run the following
[~] cd magma/orc8r/tools/packer
[~/magma/orc8r/tools/packer] packer build -force \
    -var "aws_access_key=YOUR_ACCESS_KEY" \
    -var "aws_secret_key=YOUR_SECRET_KEY" \
    -var "subnet=YOUR_SUBNET" \
    -var "vpc=YOUR_VPC" \
    debian-stretch-aws.json

YOUR_SUBNET and YOUR_VPC should specify an existing subnet and vpc on your AWS region. The choice of subnet and vpc won't affect the final box. These are the subnet/vpc which the box is launched into while building.

The result should show

==> Builds finished. The artifacts of successful builds are:
--> amazon-ebs: AMIs were created:
us-west-1: ami-0f1c9db5a767a0296

Deploy AGW AMI

On AWS:

  1. Navigate to the EC2 Service
  2. Select Launch Instance
  3. Select the AMI that was built in the previous step. This AMI will exist under My AMIs section.
  4. On page Choose an Instance Type, select a c4.xlarge instance type. Proceed to Configure Instance Details.
  5. On page Configure Instance Details, use the default settings. Proceed to Add Storage.
  6. On page Add Storage, use default of 8gb. Proceed to Add Tags.
  7. On page Add Tags, optionally add tags (e.g. Magma Secondary Gateway) to identify this as a secondary. Magma AGW. Proceed to Configure Security Group.
  8. On page “Configure Security Group”, create a new security group with the rules listed below. It is advised to limit the source IPs to the subnet that i the primary gateway resides in for all rules other than SSH. Proceed to Review and Launch.
TypeProtocolPort RangeSourceDescription
SSHTCP220.0.0.0/0-
SCTP (132)SCTP (132)All0.0.0.0/0-
Custom TCPTCP33860.0.0.0/0-
All UDPUDP0 - 655350.0.0.0/0
All ICMP - IPv4ICMPAll0.0.0.0/0-
  1. Review that the selected settings are as described here. Then proceed to Launch.
  2. Select Create a new key pair, then save the key pair created to your host machine. This pair will be used to access the gateway, so ensure the pair is saved in a safe and durable location.
  3. Finish by selecting Launch Instances.

ENI Configuration

Before installing Magma, we will add a second interface to gateway by creating an ENI and attaching it to the EC2 instance.

  1. In the EC2 service on AWS, navigate to the Network Interfaces section under the Network and Security tab on the side panel.
  2. Select Create network interface in the upper right corner.
  3. On the Create network interface configuration page, select the subnet for the ENI. To work properly, this subnet cannot be the same subnet that the EC2 instance was deployed with. These subnets must be in the same availability zone though.
  4. Select the same subnet that was used to deploy the EC2 instance.
  5. Once configured, select Create network interface.
  6. Navigate to the EC2 instances page.
  7. Find the recently deployed EC2 instance on the left hand side. Then select Actions → Networking → Attach network interface.
  8. On page Attach network interface, select the recently created ENI and then click Attach.

Install Magma

  1. Find the public IP for the gateway instance by navigating to Instances on the AWS EC2 service. Select the instance and copy the Public IPv4 Address in the instance summary.
  2. Add the AWS gateway key that was created when the instance was launched: ssh-add ~/.ssh/aws_key.pem
  3. SSH to EC2 instance using the public IP from step 1: ssh admin@<instance_public_ip>
  4. Now install Magma
[admin@~/] sudo su
[root@:/home/admin] wget https://raw.githubusercontent.com/facebookincubator/magma/v1.4/lte/gateway/deploy/agw_install.sh
[root@:/home/admin] bash agw_install cloud
`

When you see "AGW installation is done." It means that your AGW installation is done, you can make sure magma is running by executing:

service magma@* status

Access Gateway Configuration

  1. Follow the configuration steps to register the new gateway.
  2. To configure the gateway to serve as a secondary use the Orc8r API (NMS does not currently support this functionality).
    1. Use the POST request endpoint /lte/{network_id}/gateway_pools to create a new gateway pool.
    2. Add the primary gateway(s) to the pool via endpoint /lte/{network_id}/gateways/{gateway_id}/cellular/pooling.
      1. MME code should differ for each gateway in the pool.
      2. MME relative capacity should be set to 255 for each primary
    3. Add the secondary (AWS) gateway to the pool via endpoint /lte/{network_id}/gateways/{gateway_id}/cellular/pooling.
      1. MME code should differ for each gateway in the pool.
      2. MME relative capacity should be set to 1 for the secondary
  3. To enable secondary AGW to retrieve the connection state of the primary instances, the default value of use_ha: false should be changed to use_ha: true in /etc/magma/mme.yml. This configuration is mainly for Active-Standby configuration and should not be used if an Active-Active configuration is desired. When set as true, secondary AGW starts offloading UEs camped on it back to the primary instances when the primary instances come back up and start syncing up the states of connected eNBs to the orc8r.
  4. If the secondary AGW is in a different network with its eth1 interface configured with a private IP address, S1-U IP address needs to be configured with the public IP address of the interface separately as by default it will be configured with the eth1 IP address that is private.
    1. add "ipv4_sgw_s1u_addr": **** "IP_ADDRESS_STRING" via the endpoint /lte/{network_id}/gateways/{gateway_id}/cellular/epc, where IP_ADDRESS_STRING is a CIDR formatted IPv4 address, e.g., 203.0.113.25/32.
  5. If eNB is behind a different NAT than the AGW instance, its S1-U IP address communicated (with AGW instance) over the S1-MME interface is a private IP address. Then, eNB will not be reachable in the user plane (i.e., GTP-U traffic will not be routable back to eNB). To remedy this situation, assuming that the eNB uses the same routable IP address for S1-MME connection and S1-U connection, it is possible to force MME overwrite the S1-U private IP address with the public one during bearer context set up by changing the enable_gtpu_private_ip_correction: false to enable_gtpu_private_ip_correction: true in /etc/magma/mme.yml after ssh-ing into the AGW instance.

Note: The current functionality supports multiple primaries using the same secondary gateway. However the ENBs configured for the primaries must not overlap.

Enodeb Configuration

Any enodebs that will be used in the HA pool should be added to both the primary and secondary gateway via the NMS.

Make sure that your eNB supports MME pooling also known as S1-Flex as Magma HA feature relies on this capability. eNBs must be configured with MME pool using the management interface for the eNB vendor. The primary and secondary AGW’s routable ip addresses assigned for eth1 must be used in this configuration. Make sure that eNB simultaneously connects to each MME ip address in its pool and there are sctp heartbeat requests and responses on each AGW.

← Installing Carrier Wifi GatewaySentry Integration →