Configuring HAProxy (optional)¶

Making HAProxy highly-available¶

If multiple hosts are found in the inventory, deploy HAProxy in a highly-available manner by installing keepalived.

Edit the /etc/openstack_deploy/user_variables.yml to skip the deployment of keepalived along HAProxy when installing HAProxy on multiple hosts. To do this, set the following:

haproxy_use_keepalived: False

To make keepalived work, edit at least the following variables in user_variables.yml:

haproxy_keepalived_external_vip_cidr: 192.168.0.4/25
haproxy_keepalived_internal_vip_cidr: 172.29.236.54/16
haproxy_keepalived_external_interface: br-flat
haproxy_keepalived_internal_interface: br-mgmt

• haproxy_keepalived_internal_interface and haproxy_keepalived_external_interface represent the interfaces on the deployed node where the keepalived nodes bind the internal and external vip. By default, use br-mgmt.
• On the interface listed above, haproxy_keepalived_internal_vip_cidr and haproxy_keepalived_external_vip_cidr represent the internal and external (respectively) vips (with their prefix length).
• Set additional variables to adapt keepalived in your deployment. Refer to the user_variables.yml for more descriptions.

To always deploy (or upgrade to) the latest stable version of keepalived. Edit the /etc/openstack_deploy/user_variables.yml:

keepalived_use_latest_stable: True

The HAProxy playbook reads the vars/configs/keepalived_haproxy.yml variable file and provides content to the keepalived role for keepalived master and backup nodes.

Keepalived pings a public IP address to check its status. The default address is 193.0.14.129. To change this default, set the keepalived_ping_address variable in the user_variables.yml file.

You can define additional variables to adapt keepalived to your deployment. Refer to the user_variables.yml file for more information. Optionally, you can use your own variable file. For example:

haproxy_keepalived_vars_file: /path/to/myvariablefile.yml

Configuring keepalived ping checks¶

OpenStack-Ansible configures keepalived with a check script that pings an external resource and uses that ping to determine if a node has lost network connectivity. If the pings fail, keepalived fails over to another node and HAProxy serves requests there.

The destination address, ping count and ping interval are configurable via Ansible variables in /etc/openstack_deploy/user_variables.yml:

keepalived_ping_address:         # IP address to ping
keepalived_ping_count:           # ICMP packets to send (per interval)
keepalived_ping_interval:        # How often ICMP packets are sent

By default, OpenStack-Ansible configures keepalived to ping one of the root DNS servers operated by RIPE. You can change this IP address to a different external address or another address on your internal network.

Securing HAProxy communication with SSL certificates¶

The OpenStack-Ansible project provides the ability to secure HAProxy communications with self-signed or user-provided SSL certificates. By default, self-signed certificates are used with HAProxy. However, you can provide your own certificates by using the following Ansible variables:

haproxy_user_ssl_cert:          # Path to certificate
haproxy_user_ssl_key:           # Path to private key
haproxy_user_ssl_ca_cert:       # Path to CA certificate

Refer to Securing services with SSL certificates for more information on these configuration options and how you can provide your own certificates and keys to use with HAProxy. User provided certificates should be folded and formatted at 64 characters long. Single line certificates will not be accepted by HAProxy and will result in SSL validation failures. Please have a look here for information on converting your certificate to various formats.

Configuring additional services¶

Additional haproxy service entries can be configured by setting haproxy_extra_services in /etc/openstack_deploy/user_variables.yml

For more information on the service dict syntax, please reference playbooks/vars/configs/haproxy_config.yml

An example HTTP service could look like:

haproxy_extra_services:
  - service:
      haproxy_service_name: extra-web-service
      haproxy_backend_nodes: "{{ groups['service_group'] | default([]) }}"
      haproxy_ssl: "{{ haproxy_ssl }}"
      haproxy_port: 10000
      haproxy_balance_type: http
      # If backend connections should be secured with SSL (default False)
      haproxy_backend_ssl: True
      haproxy_backend_ca: /path/to/ca/cert.pem
      # Or if certificate validation should be disabled
      # haproxy_backend_ca: False

Additionally, you can specify haproxy services that are not managed in the Ansible inventory by manually specifying their hostnames/IP Addresses:

haproxy_extra_services:
  - service:
      haproxy_service_name: extra-non-inventory-service
      haproxy_backend_nodes:
        - name: nonInvHost01
          ip_addr: 172.0.1.1
        - name: nonInvHost02
          ip_addr: 172.0.1.2
        - name: nonInvHost03
          ip_addr: 172.0.1.3
      haproxy_ssl: "{{ haproxy_ssl }}"
      haproxy_port: 10001
      haproxy_balance_type: http

Adding additional global VIP addresses¶

In some cases, you might need to add additional internal VIP addresses to the load balancer front end. You can use the HAProxy role to add additional VIPs to all front ends by setting them in the extra_lb_vip_addresses variable.

The following example shows extra VIP addresses defined in the user_variables.yml file:

extra_lb_vip_addresses:
  - 10.0.0.10
  - 192.168.0.10

Adding Access Control Lists to HAProxy front end¶

Adding ACL rules in HAProxy is easy. You just need to define haproxy_acls and add the rules in the variable

Here is an example that shows how to achieve the goal

- service:
       haproxy_service_name: influxdb-relay
       haproxy_acls:
           write_queries:
              rule: "path_sub -i write"
           read_queries:
              rule: "path_sub -i query"
              backend_name: "influxdb"

This will add two acl rules path_sub -i write and path_sub -i query to the front end and use the backend specified in the rule. If no backend is specified it will use a default haproxy_service_name backend.