OpenStack4j > Documentation / LBaaS

LBaaS (Load Balancer as a Service)

drive the actual load balancing of requests. Thus, an Openstack operator can choose which back-end technology to use. The long-term vision for the project, however, is to provide a single API that allows a user to seamlessly move between different load balancing technologies should the operator choose to do so.

LBaaS System Architecture

*Source: https://wiki.openstack.org/wiki/Neutron_LBaaS_Arch

Pools

A pool is a container of a set of members to which the traffic is load-balanced. The pool construct is a way to share several configuration attributes between members that belong to the same pool. For example health monitors are configured on the pool, and all members of the same pool, will be monitored using these health monitors.

Querying

List all Pools for the current Tenant

List<? extends LbPool> list = os().networking().loadbalancers().lbPool().list();

Listing all Pools for the current Tenant with filters

Map<String, String> filter = Maps.newHashMap();
filter.put("name", "web_pool");

List<? extends LbPool> pools = os.networking().loadbalancers().lbPool().list(filter);

Finding a Pool by ID

LbPool pool = os().networking().loadbalancers().lbPool().get(id);

Creating / Updating / Deleting

Creating a Pool

LbPool pool = os.networking().loadbalancers().lbPool()
                .create(Builders.lbPool()
                   .name("web_pool")
                   .protocol(Protocol.HTTP)
                   .lbMethod(LbMethod.ROUND_ROBIN)
                   .subnetId(subnetId)
                   .build());

Updating a Pool

LbPool updated = os.networking().loadbalancers().lbPool()
                   .update(poolId, Builders.lbPoolUpdate()
                       .name("new_web_pool_name")
                       .description("Pool for Webservers")
                       .lbMethod(LbMethod.LEAST_CONNECTIONS)
                       .build());

Deleting a Pool

ActionResponse resp = os.networking().loadbalancers().lbPool().delete(poolId);

Statistics

Provides statistics such as BytesIn, BytesOut, ActiveConnections and TotalConnections for a specified Pool

LbPoolStats stats = os.networking().loadbalancers().lbPool().stats(poolId);

Members

The members of a pool are responsible for servicing the requests received through the vip’s virtual IP. The load-balancing method is used to distribute requests or connections between the pool members.

The weight of a member determines the portion of requests or connections it services compared to the other members of the pool. For example, if member A has a weight of 2 and member B has a weight of 1, then member A will service twice as many requests as member B. If the weight attribute is not specified, then the member’s weight is implicitly set to “1”.

Querying

List all Members for the current Tenant

List<? extends Member> list = os().networking().loadbalancers().member().list();

Listing all Members for the current Tenant with filters

Map<String, String> filter = Maps.newHashMap();
filter.put("status", "ACTIVE");

List<? extends Member> members = os.networking().loadbalancers().member().list(filter);

Finding a Member by ID

Member member = os().networking().loadbalancers().member().get(id);

Creating / Updating / Deleting

Creating a Member

Below is an example of creating a new member and adding it to the specified pool via poolId

Member member = os.networking().loadbalancers().member()
                  .create(Builders.member()
                     .address("192.168.0.22")
                     .poolId(poolId)
                     .protocolPort(80)
                     .weight(1)
                     .build());

Updating a Member

Member updated = os.networking().loadbalancers().member()
                   .update(memberId, Builders.memberUpdate()
                       .weight(2)
                       .adminStateUp(true)
                       .build());

Deleting a Member

ActionResponse resp = os.networking().loadbalancers().member().delete(memberId);

VIP

A VIP is the primary load balancing configuration object that specifies the virtual IP address and port on which client traffic is received, as well as other details such as the load balancing method to be use, protocol, etc. This entity is sometimes known in LB products under the name of a “virtual server”, a “vserver” or a “listener”.

Querying

List all VIPs for the current Tenant

List<? extends Vip> list = os().networking().loadbalancers().vip().list();

Listing all VIPs for the current Tenant with filters

Map<String, String> filter = Maps.newHashMap();
filter.put("name", "some VIP");

List<? extends Vip> vips = os.networking().loadbalancers().vip().list(filter);

Finding a VIP by ID

Vip vip = os().networking().loadbalancers().vip().get(id);

Creating / Updating / Deleting

Creating a VIP

Vip vip = os.networking().loadbalancers().vip()
            .create(Builders.vip()
               .name("web_vip")
               .tenantId(tenantId)
               .subnetId(subnetId)
               .poolId(poolId)
               .protocol(Protocol.HTTP)
               .protocolPort(80)
               .build());

Updating a VIP

Vip updated = os.networking().loadbalancers().vip()
                .update(vipId, Builders.vipUpdate()
                    .sessionPersistence(Builders.sessionPersistence()
                            .cookieName("webapp-cookie")
                            .type("HTTP_COOKIE")
                            .build())
                    .connectionLimit(10)
                    .build());

Deleting a VIP

ActionResponse resp = os.networking().loadbalancers().vip().delete(vipId);

Health Monitors

A health monitor is used to determine whether or not back-end members of the VIP’s pool are usable for processing a request. A pool can have several health monitors associated with it. There are different types of health monitors supported by the OpenStack LBaaS service:

  • PING: used to ping the members using ICMP.
  • TCP: used to connect to the members using TCP.
  • HTTP: used to send an HTTP request to the member.
  • HTTPS: used to send a secure HTTP request to the member.

Querying

List all Health Monitors for the current Tenant

List<? extends HealthMonitor> list = os().networking().loadbalancers().healthMonitor().list();

Listing all Health Monitors for the current Tenant with filters

Map<String, String> filter = Maps.newHashMap();
filter.put("name", "some monitor");

List<? extends HealthMonitor> hm = os.networking().loadbalancers().healthMonitor().list(filter);

Finding a Health Monitor by ID

HealthMonitor vip = os().networking().loadbalancers().healthMonitor().get(id);

Creating Health Monitors

The examples below are from the LBaaS WIKI and have been applied to OpenStack4j format.

Creating a PING Monitor

The monitor pings the member to ensure that the member is alive.

The PING monitor is the most basic type of health check. The following configurable properties can be specified

  • delay: This is the minimum time in seconds between regular pings of the member.
  • timeout: Maximum number of seconds for a monitor to wait for a ping reply before it times out. The value must be less than the delay value.
  • max retries: Number of permissible ping failures before changing the member’s status to INACTIVE. Must be a number between 1 and 10.
HealthMonitor mon = os.networking().loadbalancers().healthMonitor()
                      .create(Builders.healthMonitor()
                         .type(HealthMonitorType.PING)
                         .delay(20)
                         .timeout(10)
                         .maxRetries(3)
                         .build());

Creating a TCP Monitor

A TCP monitor tries to establish a TCP connection to the member to ensure that the member is healthy.

The following configurable properties can be specified for a TCP monitor:

  • delay: This is the minimum time in seconds between regular connections of the member.
  • timeout: Maximum number of seconds for a monitor to wait for a connection to be established before it times out. The value must be less than the delay value.
  • max retries: Number of permissible connection failures before changing the member’s status to INACTIVE. Must be a number between 1 and 10.
HealthMonitor mon = os.networking().loadbalancers().healthMonitor()
                      .create(Builders.healthMonitor()
                         .type(HealthMonitorType.TCP)
                         .delay(20)
                         .timeout(10)
                         .maxRetries(3)
                         .build());

Creating an HTTP or HTTPS Monitor

The HTTP and HTTPS monitor is more intelligent than the TCP monitor. It is capable of processing an HTTP or HTTPS response to determine the condition of a node. It supports the same basic properties as the TCP monitor and includes the additional attribute of path that is used to evaluate the HTTP response to a monitor probe.

  • urlPath: The HTTP path used in the HTTP request used by the monitor to test a member’s health. This must be a string beginning with a / (forward slash). The monitor expects a response from the node with an HTTP status code of 200.
HealthMonitor mon = os.networking().loadbalancers().healthMonitor()
                      .create(Builders.healthMonitor()
                         .type(HealthMonitorType.HTTP)
                         .delay(20)
                         .timeout(10)
                         .maxRetries(3)
                         .urlPath("/check")
                         .build());

Deleting

ActionResponse resp = os.networking().loadbalancers().healthMonitor().delete(healthMinotirId);

Associating Health Monitors to a Pool

The following examples show how to associate and disassociate health monitors against a pool.

// Associating a health monitor
HealthMonitor monitor = os.networking().loadbalancers().lbPool().associateHealthMonitor(lbPoolId, healthMonitorId);

// Disassociate a health monitor
ActionResponse resp = os.networking().loadbalancers().lbPool().disAssociateHealthMonitor(lbPoolId, healthMonitorId);