Red Hat Enterprise Linux 7

Load Balancer Administration

Configuring Keepalived and HAProxy

Last Updated: 2019-08-23

Red Hat Enterprise Linux 7 Load Balancer Administration
Configuring Keepalived and HAProxy

Steven Levine
Red Hat Customer Content Services
[email protected]

Stephen Wadeley
Red Hat Customer Content Services
[email protected]
Legal Notice
Copyright © 2018 Red Hat, Inc.

This document is licensed by Red Hat under the Creative Commons Attribution-ShareAlike 3.0
Unported License. If you distribute this document, or a modified version of it, you must provide
attribution to Red Hat, Inc. and provide a link to the original. If the document is modified, all Red Hat
trademarks must be removed.

Red Hat, as the licensor of this document, waives the right to enforce, and agrees not to assert,
Section 4d of CC-BY-SA to the fullest extent permitted by applicable law.

Red Hat, Red Hat Enterprise Linux, the Shadowman logo, the Red Hat logo, JBoss, OpenShift,
Fedora, the Infinity logo, and RHCE are trademarks of Red Hat, Inc., registered in the United States
and other countries.

Linux ® is the registered trademark of Linus Torvalds in the United States and other countries.

Java ® is a registered trademark of Oracle and/or its affiliates.

XFS ® is a trademark of Silicon Graphics International Corp. or its subsidiaries in the United States
and/or other countries.

MySQL ® is a registered trademark of MySQL AB in the United States, the European Union and
other countries.

Node.js ® is an official trademark of Joyent. Red Hat is not formally related to or endorsed by the
official Joyent Node.js open source or commercial project.

The OpenStack ® Word Mark and OpenStack logo are either registered trademarks/service marks
or trademarks/service marks of the OpenStack Foundation, in the United States and other
countries and are used with the OpenStack Foundation's permission. We are not affiliated with,
endorsed or sponsored by the OpenStack Foundation, or the OpenStack community.

All other trademarks are the property of their respective owners.

Abstract
Building a Load Balancer system offers a highly available and scalable solution for production
services using specialized Linux Virtual Servers (LVS) for routing and load-balancing techniques
configured through Keepalived and HAProxy. This book discusses the configuration of high-
performance systems and services using the Load Balancer technologies in Red Hat Enterprise
Linux 7.
 Table of Contents

Table of Contents
.CHAPTER
. . . . . . . . . . 1.. .LOAD
. . . . . . .BALANCER
. . . . . . . . . . . .OVERVIEW
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . .
1.1. KEEPALIVED 3
1.2. HAPROXY 3
1.3. KEEPALIVED AND HAPROXY 3

.CHAPTER
. . . . . . . . . . 2.
. . KEEPALIVED
. . . . . . . . . . . . . . OVERVIEW
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . .
2.1. A BASIC KEEPALIVED LOAD BALANCER CONFIGURATION 5
2.2. A THREE-TIER KEEPALIVED LOAD BALANCER CONFIGURATION 7
2.3. KEEPALIVED SCHEDULING OVERVIEW 8
2.4. ROUTING METHODS 10
2.5. PERSISTENCE AND FIREWALL MARKS WITH KEEPALIVED 13

.CHAPTER
. . . . . . . . . . 3.
. . SETTING
. . . . . . . . . . UP
. . . .LOAD
. . . . . . BALANCER
. . . . . . . . . . . . .PREREQUISITES
. . . . . . . . . . . . . . . . . FOR
. . . . . KEEPALIVED
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
..............
3.1. THE NAT LOAD BALANCER NETWORK 15
3.2. LOAD BALANCER USING DIRECT ROUTING 17
3.3. PUTTING THE CONFIGURATION TOGETHER 21
3.4. MULTI-PORT SERVICES AND LOAD BALANCER 22
3.5. CONFIGURING FTP 25
3.6. SAVING NETWORK PACKET FILTER SETTINGS 27
3.7. TURNING ON PACKET FORWARDING AND NONLOCAL BINDING 28
3.8. CONFIGURING SERVICES ON THE REAL SERVERS 28

.CHAPTER
. . . . . . . . . . 4.
. . .INITIAL
. . . . . . . .LOAD
. . . . . . BALANCER
. . . . . . . . . . . . CONFIGURATION
. . . . . . . . . . . . . . . . . . . WITH
. . . . . . KEEPALIVED
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .29
..............
4.1. A BASIC KEEPALIVED CONFIGURATION 29
4.2. KEEPALIVED DIRECT ROUTING CONFIGURATION 33
4.3. STARTING THE SERVICE 35

. . . . . . . . . . . 5.
CHAPTER . . HAPROXY
. . . . . . . . . . . .CONFIGURATION
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .36
..............
5.1. HAPROXY SCHEDULING ALGORITHMS 36
5.2. GLOBAL SETTINGS 37
5.3. DEFAULT SETTINGS 37
5.4. FRONTEND SETTINGS 38
5.5. BACKEND SETTINGS 39
5.6. STARTING HAPROXY 39
5.7. LOGGING HAPROXY MESSAGES TO RSYSLOG 39

APPENDIX A. EXAMPLE CONFIGURATION: LOAD BALANCING CEPH OBJECT GATEWAY SERVERS WITH
. . . . . . . . . . . .AND
HAPROXY . . . . .KEEPALIVED
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
..............
A.1. PREREQUISITES 41
A.2. PREPARING HAPROXY NODES 41
A.3. INSTALL AND CONFIGURE KEEPALIVED 42
A.4. INSTALL AND CONFIGURE HAPROXY 43
A.5. TEST YOUR HAPROXY CONFIGURATION 44

. . . . . . . . . . . .B.
APPENDIX . . REVISION
. . . . . . . . . . . HISTORY
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .46
..............

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .47
INDEX ..............

1
Load Balancer Administration

2
 CHAPTER 1. LOAD BALANCER OVERVIEW

CHAPTER 1. LOAD BALANCER OVERVIEW

The Load Balancer is a set of integrated software components that provide for balancing IP traffic
across a set of real servers. It consists of two main technologies to monitor cluster members and cluster
services: Keepalived and HAProxy. Keepalived uses Linux virtual server (LVS) to perform load balancing
and failover tasks on the active and passive routers, while HAProxy performs load balancing and high-
availability services to TCP and HTTP applications.

1.1. KEEPALIVED
The keepalived daemon runs on both the active and passive LVS routers. All routers running
keepalived use the Virtual Redundancy Routing Protocol (VRRP). The active router sends VRRP
advertisements at periodic intervals; if the backup routers fail to receive these advertisements, a new
active router is elected.

On the active router, keepalived can also perform load balancing tasks for real servers.

Keepalived is the controlling process related to LVS routers. At boot time, the daemon is started by the
systemctl command, which reads the configuration file /etc/keepalived/keepalived.conf. On the active
router, the keepalived daemon starts the LVS service and monitors the health of the services based on
the configured topology. Using VRRP, the active router sends periodic advertisements to the backup
routers. On the backup routers, the VRRP instance determines the running status of the active router. If
the active router fails to advertise after a user-configurable interval, Keepalived initiates failover. During
failover, the virtual servers are cleared. The new active router takes control of the virtual IP address
(VIP), sends out an ARP message, sets up IPVS table entries (virtual servers), begins health checks, and
starts sending VRRP advertisements.

Keepalived performs failover on layer 4, or the Transport layer, upon which TCP conducts connection-
based data transmissions. When a real server fails to reply to simple timeout TCP connection,
keepalived detects that the server has failed and removes it from the server pool.

1.2. HAPROXY
HAProxy offers load balanced services to HTTP and TCP-based services, such as Internet-connected
services and web-based applications. Depending on the load balancer scheduling algorithm chosen,
haproxy is able to process several events on thousands of connections across a pool of multiple real
servers acting as one virtual server. The scheduler determines the volume of connections and either
assigns them equally in non-weighted schedules or given higher connection volume to servers that can
handle higher capacity in weighted algorithms.

HAProxy allows users to define several proxy services, and performs load balancing services of the
traffic for the proxies. Proxies are made up of a frontend system and one or more back-end systems.
The front-end system defines the IP address (the VIP) and port on which the proxy listens, as well as the
back-end systems to use for a particular proxy.

The back-end system is a pool of real servers, and defines the load balancing algorithm.

HAProxy performs load-balancing management on layer 7, or the Application layer. In most cases,
administrators deploy HAProxy for HTTP-based load balancing, such as production web applications,
for which high availability infrastructure is a necessary part of business continuity.

1.3. KEEPALIVED AND HAPROXY

Administrators can use both Keepalived and HAProxy together for a more robust and scalable high

3
Load Balancer Administration

availability environment. Using the speed and scalability of HAProxy to perform load balancing for HTTP
and other TCP-based services in conjunction with Keepalived failover services, administrators can
increase availability by distributing load across real servers as well as ensuring continuity in the event of
router unavailability by performing failover to backup routers.

4
 CHAPTER 2. KEEPALIVED OVERVIEW

CHAPTER 2. KEEPALIVED OVERVIEW

Keepalived runs on an active LVS router as well as one or more optional backup LVS routers. The active
LVS router serves two roles:

To balance the load across the real servers.

To check the integrity of the services on each real server.

The active (master) router informs the backup routers of its active status using the Virtual Router
Redundancy Protocol (VRRP), which requires the master router to send out advertisements at regular
intervals. If the active router stops sending advertisements, a new master is elected.

NOTE

Red Hat does not support rolling updates of keepalived where the configuration changes
the VRRP version to use. All routers must be running the same version of VRRP in a
keepalived load balancer configuration. A VRRP version mismatch will lead to the
following messages:

Aug 3 17:07:19 hostname Keepalived_vrrp[123]: receive an invalid ip number count

associated with VRID!
Aug 3 17:07:19 hostname Keepalived_vrrp[123]: bogus VRRP packet received on
em2 !!!
Aug 3 17:07:19 hostname Keepalived_vrrp[123]: VRRP_Instance(vrrp_ipv6) ignoring
received advertisment...

Red Hat recommend that all systems should run the same keepalived version and
keepalived configurations should be identical where possible to avoid compatibility
issues.

2.1. A BASIC KEEPALIVED LOAD BALANCER CONFIGURATION

Figure 2.1, “A Basic Load Balancer Configuration” shows a simple Keepalived Load Balancer
configuration consisting of two layers. On the first layer is one active and several backup LVS routers.
Each LVS router has two network interfaces, one interface on the Internet and one on the private
network, enabling them to regulate traffic between the two networks. For this example the active router
is using Network Address Translation or NAT to direct traffic from the Internet to a variable number of
real servers on the second layer, which in turn provide the necessary services. Therefore, the real servers
in this example are connected to a dedicated private network segment and pass all public traffic back
and forth through the active LVS router. To the outside world, the servers appear as one entity.

5
Load Balancer Administration

Figure 2.1. A Basic Load Balancer Configuration

Service requests arriving at the LVS router are addressed to a virtual IP address, or VIP. This is a
publicly-routable address the administrator of the site associates with a fully-qualified domain name,
such as www.example.com, and is assigned to one or more virtual servers . A virtual server is a service
configured to listen on a specific virtual IP. A VIP address migrates from one LVS router to the other
during a failover, thus maintaining a presence at that IP address. A VIP is also known as a floating IP
addresses.

VIP addresses may be assigned to the same device which connects the LVS router to the Internet. For
example, if eth0 is connected to the Internet, then multiple virtual servers can be assigned to eth0.
Alternatively, each virtual server can be associated with a separate device per service. For example,
HTTP traffic can be handled on eth0 at 192.168.1.111 while FTP traffic can be handled on eth0 at
192.168.1.222.

In a deployment scenario involving both one active and one passive router, the role of the active router
is to redirect service requests from virtual IP addresses to the real servers. The redirection is based on
one of eight supported load-balancing algorithms described further in Section 2.3, “keepalived
Scheduling Overview”.

The active router also dynamically monitors the overall health of the specific services on the real servers
through three built-in health checks: simple TCP connect, HTTP, and HTTPS. For TCP connect, the
active router will periodically check that it can connect to the real servers on a certain port. For HTTP
and HTTPS, the active router will periodically fetch a URL on the real servers and verify its content.

The backup routers perform the role of standby systems. Router failover is handled by VRRP. On
startup, all routers will join a multicast group. This multicast group is used to send and receive VRRP
advertisements. Since VRRP is a priority based protocol, the router with the highest priority is elected

6
 CHAPTER 2. KEEPALIVED OVERVIEW

the master. Once a router has been elected master, it is responsible for sending VRRP advertisements at
periodic intervals to the multicast group.

If the backup routers fail to receive advertisements within a certain time period (based on the
advertisement interval), a new master will be elected. The new master will take over the VIP and send an
Address Resolution Protocol (ARP) message. When a router returns to active service, it may either
become a backup or a master. The behavior is determined by the router's priority.

The simple, two-layered configuration used in Figure 2.1, “A Basic Load Balancer Configuration” is best
for serving data which does not change very frequently — such as static web pages — because the
individual real servers do not automatically sync data between each node.

2.2. A THREE-TIER KEEPALIVED LOAD BALANCER CONFIGURATION

Figure 2.2, “A Three-Tier Load Balancer Configuration” shows a typical three-tier Keepalived Load
Balancer topology. In this example, the active LVS router routes the requests from the Internet to the
pool of real servers. Each of the real servers then accesses a shared data source over the network.

7
Load Balancer Administration

Figure 2.2. A Three-Tier Load Balancer Configuration

This configuration is ideal for busy FTP servers, where accessible data is stored on a central, highly
available server and accessed by each real server by means of an exported NFS directory or Samba
share. This topology is also recommended for websites that access a central, highly available database
for transactions. Additionally, using an active-active configuration with the Load Balancer,
administrators can configure one high-availability cluster to serve both of these roles simultaneously.

The third tier in the above example does not have to use the Load Balancer, but failing to use a highly
available solution would introduce a critical single point of failure.

2.3. KEEPALIVED SCHEDULING OVERVIEW

Using Keepalived provides a great deal of flexibility in distributing traffic across real servers, in part due
to the variety of scheduling algorithms supported. Load balancing is superior to less flexible methods,
such as Round-Robin DNS where the hierarchical nature of DNS and the caching by client machines can

8
 CHAPTER 2. KEEPALIVED OVERVIEW

lead to load imbalances. Additionally, the low-level filtering employed by the LVS router has advantages
over application-level request forwarding because balancing loads at the network packet level causes
minimal computational overhead and allows for greater scalability.

Using assigned weights gives arbitrary priorities to individual machines. Using this form of scheduling, it
is possible to create a group of real servers using a variety of hardware and software combinations and
the active router can evenly load each real server.

The scheduling mechanism for Keepalived is provided by a collection of kernel patches called IP Virtual
Server or IPVS modules. These modules enable layer 4 (L4) transport layer switching, which is designed
to work well with multiple servers on a single IP address.

To track and route packets to the real servers efficiently, IPVS builds an IPVS table in the kernel. This
table is used by the active LVS router to redirect requests from a virtual server address to and returning
from real servers in the pool.

2.3.1. Keepalived Scheduling Algorithms

The structure that the IPVS table takes depends on the scheduling algorithm that the administrator
chooses for any given virtual server. To allow for maximum flexibility in the types of services you can
cluster and how these services are scheduled, Keepalived supports the following scheduling algorithms
listed below.

Round-Robin Scheduling
Distributes each request sequentially around the pool of real servers. Using this algorithm, all the real
servers are treated as equals without regard to capacity or load. This scheduling model resembles
round-robin DNS but is more granular due to the fact that it is network-connection based and not
host-based. Load Balancer round-robin scheduling also does not suffer the imbalances caused by
cached DNS queries.

Weighted Round-Robin Scheduling

Distributes each request sequentially around the pool of real servers but gives more jobs to servers
with greater capacity. Capacity is indicated by a user-assigned weight factor, which is then adjusted
upward or downward by dynamic load information.

Weighted round-robin scheduling is a preferred choice if there are significant differences in the
capacity of real servers in the pool. However, if the request load varies dramatically, the more heavily
weighted server may answer more than its share of requests.

Least-Connection
Distributes more requests to real servers with fewer active connections. Because it keeps track of
live connections to the real servers through the IPVS table, least-connection is a type of dynamic
scheduling algorithm, making it a better choice if there is a high degree of variation in the request
load. It is best suited for a real server pool where each member node has roughly the same capacity. If
a group of servers have different capabilities, weighted least-connection scheduling is a better
choice.

Weighted Least-Connections
Distributes more requests to servers with fewer active connections relative to their capacities.
Capacity is indicated by a user-assigned weight, which is then adjusted upward or downward by
dynamic load information. The addition of weighting makes this algorithm ideal when the real server
pool contains hardware of varying capacity.

Locality-Based Least-Connection Scheduling

9
Load Balancer Administration

Distributes more requests to servers with fewer active connections relative to their destination IPs.
This algorithm is designed for use in a proxy-cache server cluster. It routes the packets for an IP
address to the server for that address unless that server is above its capacity and has a server in its
half load, in which case it assigns the IP address to the least loaded real server.

Locality-Based Least-Connection Scheduling with Replication Scheduling

Distributes more requests to servers with fewer active connections relative to their destination IPs.
This algorithm is also designed for use in a proxy-cache server cluster. It differs from Locality-Based
Least-Connection Scheduling by mapping the target IP address to a subset of real server nodes.
Requests are then routed to the server in this subset with the lowest number of connections. If all the
nodes for the destination IP are above capacity, it replicates a new server for that destination IP
address by adding the real server with the least connections from the overall pool of real servers to
the subset of real servers for that destination IP. The most loaded node is then dropped from the
real server subset to prevent over-replication.

Destination Hash Scheduling

Distributes requests to the pool of real servers by looking up the destination IP in a static hash table.
This algorithm is designed for use in a proxy-cache server cluster.

Source Hash Scheduling

Distributes requests to the pool of real servers by looking up the source IP in a static hash table. This
algorithm is designed for LVS routers with multiple firewalls.

Shortest Expected Delay

Distributes connection requests to the server that has the shortest delay expected based on number
of connections on a given server divided by its assigned weight.

Never Queue
A two-pronged scheduler that first finds and sends connection requests to a server that is idling, or
has no connections. If there are no idling servers, the scheduler defaults to the server that has the
least delay in the same manner as Shortest Expected Delay.

2.3.2. Server Weight and Scheduling

The administrator of Load Balancer can assign a weight to each node in the real server pool. This weight
is an integer value which is factored into any weight-aware scheduling algorithms (such as weighted
least-connections) and helps the LVS router more evenly load hardware with different capabilities.

Weights work as a ratio relative to one another. For instance, if one real server has a weight of 1 and the
other server has a weight of 5, then the server with a weight of 5 gets 5 connections for every 1
connection the other server gets. The default value for a real server weight is 1.

Although adding weight to varying hardware configurations in a real server pool can help load-balance
the cluster more efficiently, it can cause temporary imbalances when a real server is introduced to the
real server pool and the virtual server is scheduled using weighted least-connections. For example,
suppose there are three servers in the real server pool. Servers A and B are weighted at 1 and the third,
server C, is weighted at 2. If server C goes down for any reason, servers A and B evenly distributes the
abandoned load. However, once server C comes back online, the LVS router sees it has zero
connections and floods the server with all incoming requests until it is on par with servers A and B.

2.4. ROUTING METHODS

10
 CHAPTER 2. KEEPALIVED OVERVIEW

Red Hat Enterprise Linux uses Network Address Translation (NAT routing ) or direct routing for
Keepalived. This allows the administrator tremendous flexibility when utilizing available hardware and
integrating the Load Balancer into an existing network.

2.4.1. NAT Routing

Figure 2.3, “Load Balancer Implemented with NAT Routing” , illustrates Load Balancer utilizing NAT
routing to move requests between the Internet and a private network.

Figure 2.3. Load Balancer Implemented with NAT Routing

In the example, there are two NICs in the active LVS router. The NIC for the Internet has a real IP address
and a floating IP address on eth0. The NIC for the private network interface has a real IP address and a
floating IP address on eth1. In the event of failover, the virtual interface facing the Internet and the
private facing virtual interface are taken over by the backup LVS router simultaneously. All of the real
servers located on the private network use the floating IP for the NAT router as their default route to
communicate with the active LVS router so that their abilities to respond to requests from the Internet
is not impaired.

In this example, the LVS router's public floating IP address and private NAT floating IP address are
assigned to physical NICs. While it is possible to associate each floating IP address to its own physical
device on the LVS router nodes, having more than two NICs is not a requirement.

Using this topology, the active LVS router receives the request and routes it to the appropriate server.
The real server then processes the request and returns the packets to the LVS router which uses
network address translation to replace the address of the real server in the packets with the LVS
router's public VIP address. This process is called IP masquerading because the actual IP addresses of
the real servers is hidden from the requesting clients.

11
Load Balancer Administration

Using this NAT routing, the real servers may be any kind of machine running various operating systems.
The main disadvantage is that the LVS router may become a bottleneck in large cluster deployments
because it must process outgoing as well as incoming requests.

The ipvs modules utilize their own internal NAT routines that are independent of iptables and ip6tables
NAT. This will facilitate both IPv4 and IPv6 NAT when the real server is configured for NAT as opposed
to DR in the /etc/keepalived/keepalived.conf file.

2.4.2. Direct Routing

Building a Load Balancer setup that uses direct routing provides increased performance benefits
compared to other Load Balancer networking topologies. Direct routing allows the real servers to
process and route packets directly to a requesting user rather than passing all outgoing packets through
the LVS router. Direct routing reduces the possibility of network performance issues by relegating the
job of the LVS router to processing incoming packets only.

Figure 2.4. Load Balancer Implemented with Direct Routing

In the typical direct routing Load Balancer setup, the LVS router receives incoming server requests
through the virtual IP (VIP) and uses a scheduling algorithm to route the request to the real servers. The
real server processes the request and sends the response directly to the client, bypassing the LVS

12
 CHAPTER 2. KEEPALIVED OVERVIEW

router. This method of routing allows for scalability in that real servers can be added without the added
burden on the LVS router to route outgoing packets from the real server to the client, which can
become a bottleneck under heavy network load.

2.4.2.1. Direct Routing and the ARP Limitation

While there are many advantages to using direct routing in Load Balancer, there are limitations as well.
The most common issue with Load Balancer through direct routing is with Address Resolution Protocol
(ARP).

In typical situations, a client on the Internet sends a request to an IP address. Network routers typically
send requests to their destination by relating IP addresses to a machine's MAC address with ARP. ARP
requests are broadcast to all connected machines on a network, and the machine with the correct
IP/MAC address combination receives the packet. The IP/MAC associations are stored in an ARP cache,
which is cleared periodically (usually every 15 minutes) and refilled with IP/MAC associations.

The issue with ARP requests in a direct routing Load Balancer setup is that because a client request to
an IP address must be associated with a MAC address for the request to be handled, the virtual IP
address of the Load Balancer system must also be associated to a MAC as well. However, since both the
LVS router and the real servers all have the same VIP, the ARP request will be broadcast to all the
machines associated with the VIP. This can cause several problems, such as the VIP being associated
directly to one of the real servers and processing requests directly, bypassing the LVS router completely
and defeating the purpose of the Load Balancer setup.

To solve this issue, ensure that the incoming requests are always sent to the LVS router rather than one
of the real servers. This can be done by either filtering ARP requests or filtering IP packets. ARP filtering
can be done using the arptables utility and IP packets can be filtered using iptables or firewalld. The
two approaches differ as follows:

The ARP filtering method blocks requests reaching the real servers. This prevents ARP from
associating VIPs with real servers, leaving the active virtual server to respond with a MAC
addresses.

The IP packet filtering method permits routing packets to real servers with other IP addresses.
This completely sidesteps the ARP problem by not configuring VIPs on real servers in the first
place.

2.5. PERSISTENCE AND FIREWALL MARKS WITH KEEPALIVED

In certain situations, it may be desirable for a client to reconnect repeatedly to the same real server,
rather than have a load balancing algorithm send that request to the best available server. Examples of
such situations include multi-screen web forms, cookies, SSL, and FTP connections. In these cases, a
client may not work properly unless the transactions are being handled by the same server to retain
context. Keepalived provides two different features to handle this: persistence and firewall marks.

2.5.1. Persistence
When enabled, persistence acts like a timer. When a client connects to a service, Load Balancer
remembers the last connection for a specified period of time. If that same client IP address connects
again within that period, it is sent to the same server it connected to previously — bypassing the load-
balancing mechanisms. When a connection occurs outside the time window, it is handled according to
the scheduling rules in place.

Persistence also allows the administrator to specify a subnet mask to apply to the client IP address test
13
Load Balancer Administration

Persistence also allows the administrator to specify a subnet mask to apply to the client IP address test
as a tool for controlling what addresses have a higher level of persistence, thereby grouping connections
to that subnet.

Grouping connections destined for different ports can be important for protocols which use more than
one port to communicate, such as FTP. However, persistence is not the most efficient way to deal with
the problem of grouping together connections destined for different ports. For these situations, it is
best to use firewall marks.

2.5.2. Firewall Marks

Firewall marks are an easy and efficient way to group ports used for a protocol or group of related
protocols. For instance, if Load Balancer is deployed to run an e-commerce site, firewall marks can be
used to bundle HTTP connections on port 80 and secure, HTTPS connections on port 443. By assigning
the same firewall mark to the virtual server for each protocol, state information for the transaction can
be preserved because the LVS router forwards all requests to the same real server after a connection is
opened.

Because of its efficiency and ease-of-use, administrators of Load Balancer should use firewall marks
instead of persistence whenever possible for grouping connections. However, administrators should still
add persistence to the virtual servers in conjunction with firewall marks to ensure the clients are
reconnected to the same server for an adequate period of time.

14
 CHAPTER 3. SETTING UP LOAD BALANCER PREREQUISITES FOR KEEPALIVED

CHAPTER 3. SETTING UP LOAD BALANCER PREREQUISITES

FOR KEEPALIVED
Load Balancer using keepalived consists of two basic groups: the LVS routers and the real servers. To
prevent a single point of failure, each group should have at least two members.

The LVS router group should consist of two identical or very similar systems running Red Hat Enterprise
Linux. One will act as the active LVS router while the other stays in hot standby mode, so they need to
have as close to the same capabilities as possible.

Before choosing and configuring the hardware for the real server group, determine which of the three
Load Balancer topologies to use.

3.1. THE NAT LOAD BALANCER NETWORK

The NAT topology allows for great latitude in utilizing existing hardware, but it is limited in its ability to
handle large loads because all packets going into and coming out of the pool pass through the Load
Balancer router.

Network Layout
The topology for Load Balancer using NAT routing is the easiest to configure from a network layout
perspective because only one access point to the public network is needed. The real servers are on a
private network and respond to all requests through the LVS router.

Hardware
In a NAT topology, each real server only needs one NIC since it will only be responding to the LVS
router. The LVS routers, on the other hand, need two NICs each to route traffic between the two
networks. Because this topology creates a network bottleneck at the LVS router, Gigabit Ethernet
NICs can be employed on each LVS router to increase the bandwidth the LVS routers can handle. If
Gigabit Ethernet is employed on the LVS routers, any switch connecting the real servers to the LVS
routers must have at least two Gigabit Ethernet ports to handle the load efficiently.

Software
Because the NAT topology requires the use of iptables for some configurations, there can be a large
amount of software configuration outside of Keepalived. In particular, FTP services and the use of
firewall marks requires extra manual configuration of the LVS routers to route requests properly.

3.1.1. Configuring Network Interfaces for Load Balancer with NAT

To set up Load Balancer with NAT, you must first configure the network interfaces for the public
network and the private network on the LVS routers. In this example, the LVS routers' public interfaces
(eth0) will be on the 203.0.113.0/24 network and the private interfaces which link to the real servers
(eth1) will be on the 10.11.12.0/24 network.

IMPORTANT

At the time of writing, the NetworkManager service is not compatible with Load
Balancer. In particular, IPv6 VIPs are known not to work when the IPv6 addresses are
assigned by SLAAC. For this reason, the examples shown here use configuration files and
the network service.

On the active or primary LVS router node, the public interface's network configuration file,
15
Load Balancer Administration

On the active or primary LVS router node, the public interface's network configuration file,
/etc/sysconfig/network-scripts/ifcfg-eth0, could look something like this:

DEVICE=eth0
BOOTPROTO=static
ONBOOT=yes
IPADDR=203.0.113.9
NETMASK=255.255.255.0
GATEWAY=203.0.113.254

The configuration file, /etc/sysconfig/network-scripts/ifcfg-eth1, for the private NAT interface on the
LVS router could look something like this:

DEVICE=eth1
BOOTPROTO=static
ONBOOT=yes
IPADDR=10.11.12.9
NETMASK=255.255.255.0

In this example, the VIP for the LVS router's public interface will be 203.0.113.10 and the VIP for the NAT
or private interface will be 10.11.12.10. It is essential that the real servers route requests back to the VIP
for the NAT interface.

IMPORTANT

The sample Ethernet interface configuration settings in this section are for the real IP
addresses of an LVS router and not the floating IP addresses.

After configuring the primary LVS router node's network interfaces, configure the backup LVS router's
real network interfaces (taking care that none of the IP address conflict with any other IP addresses on
the network).

IMPORTANT

Ensure that each interface on the backup node services the same network as the
interface on the primary node. For instance, if eth0 connects to the public network on
the primary node, it must also connect to the public network on the backup node.

3.1.2. Routing on the Real Servers

The most important thing to remember when configuring the real servers network interfaces in a NAT
topology is to set the gateway for the NAT floating IP address of the LVS router. In this example, that
address is 10.11.12.10.

NOTE

Once the network interfaces are up on the real servers, the machines will be unable to
ping or connect in other ways to the public network. This is normal. You will, however, be
able to ping the real IP for the LVS router's private interface, in this case 10.11.12.9.

The real server's configuration file, /etc/sysconfig/network-scripts/ifcfg-eth0, file could look similar to
this:

16
 CHAPTER 3. SETTING UP LOAD BALANCER PREREQUISITES FOR KEEPALIVED

DEVICE=eth0
ONBOOT=yes
BOOTPROTO=static
IPADDR=10.11.12.1
NETMASK=255.255.255.0
GATEWAY=10.11.12.10


WARNING

If a real server has more than one network interface configured with a GATEWAY=
line, the first one to come up will get the gateway. Therefore if both eth0 and eth1
are configured and eth1 is used for Load Balancer, the real servers may not route
requests properly.

It is best to turn off extraneous network interfaces by setting ONBOOT=no in their

network configuration files within the /etc/sysconfig/network-scripts/ directory or
by making sure the gateway is correctly set in the interface which comes up first.

3.1.3. Enabling NAT Routing on the LVS Routers

In a simple NAT Load Balancer configuration where each clustered service uses only one port, like HTTP
on port 80, the administrator need only enable packet forwarding on the LVS routers for the requests to
be properly routed between the outside world and the real servers. However, more configuration is
necessary when the clustered services require more than one port to go to the same real server during a
user session.

Once forwarding is enabled on the LVS routers and the real servers are set up and have the clustered
services running, use keepalived to configure IP information.


WARNING

Do not configure the floating IP for eth0 or eth1 by manually editing network
configuration files or using a network configuration tool. Instead, configure them by
means of the keepalived.conf file.

When finished, start the keepalived service. Once it is up and running, the active LVS router will begin
routing requests to the pool of real servers.

3.2. LOAD BALANCER USING DIRECT ROUTING

Direct routing allows real servers to process and route packets directly to a requesting user rather than
passing outgoing packets through the LVS router. Direct routing requires that the real servers be
physically connected to a network segment with the LVS router and be able to process and direct
outgoing packets as well.

17
Load Balancer Administration

Network Layout
In a direct routing Load Balancer setup, the LVS router needs to receive incoming requests and route
them to the proper real server for processing. The real servers then need to directly route the
response to the client. So, for example, if the client is on the Internet, and sends the packet through
the LVS router to a real server, the real server must be able to connect directly to the client through
the Internet. This can be done by configuring a gateway for the real server to pass packets to the
Internet. Each real server in the server pool can have its own separate gateway (and each gateway
with its own connection to the Internet), allowing for maximum throughput and scalability. For typical
Load Balancer setups, however, the real servers can communicate through one gateway (and
therefore one network connection).

Hardware
The hardware requirements of a Load Balancer system using direct routing is similar to other Load
Balancer topologies. While the LVS router needs to be running Red Hat Enterprise Linux to process
the incoming requests and perform load-balancing for the real servers, the real servers do not need
to be Linux machines to function correctly. The LVS routers need one or two NICs each (depending
on if there is a backup router). You can use two NICs for ease of configuration and to distinctly
separate traffic; incoming requests are handled by one NIC and routed packets to real servers on the
other.

Since the real servers bypass the LVS router and send outgoing packets directly to a client, a
gateway to the Internet is required. For maximum performance and availability, each real server can
be connected to its own separate gateway which has its own dedicated connection to the network to
which the client is connected (such as the Internet or an intranet).

Software
There is some configuration outside of keepalived that needs to be done, especially for
administrators facing ARP issues when using Load Balancer by means of direct routing. Refer to
Section 3.2.1, “Direct Routing Using arptables” or Section 3.2.3, “Direct Routing Using iptables” for
more information.

3.2.1. Direct Routing Using arptables

In order to configure direct routing using arptables, each real server must have their virtual IP address
configured, so they can directly route packets. ARP requests for the VIP are ignored entirely by the real
servers, and any ARP packets that might otherwise be sent containing the VIPs are mangled to contain
the real server's IP instead of the VIPs.

Using the arptables method, applications may bind to each individual VIP or port that the real server is
servicing. For example, the arptables method allows multiple instances of Apache HTTP Server to be
running and bound explicitly to different VIPs on the system.

However, using the arptables method, VIPs cannot be configured to start on boot using standard Red
Hat Enterprise Linux system configuration tools.

To configure each real server to ignore ARP requests for each virtual IP addresses, perform the
following steps:

1. Create the ARP table entries for each virtual IP address on each real server (the real_ip is the IP
the director uses to communicate with the real server; often this is the IP bound to eth0):

arptables -A IN -d <virtual_ip> -j DROP

arptables -A OUT -s <virtual_ip> -j mangle --mangle-ip-s <real_ip>

18
 CHAPTER 3. SETTING UP LOAD BALANCER PREREQUISITES FOR KEEPALIVED

This will cause the real servers to ignore all ARP requests for the virtual IP addresses, and
change any outgoing ARP responses which might otherwise contain the virtual IP so that they
contain the real IP of the server instead. The only node that should respond to ARP requests for
any of the VIPs is the current active LVS node.

2. Once this has been completed on each real server, save the ARP table entries by typing the
following commands on each real server:

arptables-save > /etc/sysconfig/arptables

systemctl enable arptables.service

The systemctl enable command will cause the system to reload the arptables configuration on
bootup before the network is started.

3. Configure the virtual IP address on all real servers using ip addr to create an IP alias. For
example:

# ip addr add 192.168.76.24 dev eth0

4. Configure Keepalived for Direct Routing. This can be done by adding lb_kind DR to the
keepalived.conf file. Refer to Chapter 4, Initial Load Balancer Configuration with Keepalived for
more information.

3.2.2. Direct Routing Using firewalld

You may also work around the ARP issue using the direct routing method by creating firewall rules using
firewalld. To configure direct routing using firewalld, you must add rules that create a transparent proxy
so that a real server will service packets sent to the VIP address, even though the VIP address does not
exist on the system.

The firewalld method is simpler to configure than the arptables method. This method also circumvents
the LVS ARP issue entirely, because the virtual IP address or addresses exist only on the active LVS
director.

However, there are performance issues using the firewalld method compared to arptables, as there is
overhead in forwarding every return packet.

You also cannot reuse ports using the firewalld method. For example, it is not possible to run two
separate Apache HTTP Server services bound to port 80, because both must bind to INADDR_ANY
instead of the virtual IP addresses.

To configure direct routing using the firewalld method, perform the following steps on every real server:

1. Ensure that firewalld is running.

# systemctl start firewalld

Ensure that firewalld is enabled to start at system start.

# systemctl enable firewalld

2. Enter the following command for every VIP, port, and protocol (TCP or UDP) combination
intended to be serviced for the real server. This command will cause the real servers to process
packets destined for the VIP and port that they are given.

19
Load Balancer Administration

# firewall-cmd --permanent --direct --add-rule ipv4 nat PREROUTING 0 -d vip -p tcp|udp -m

tcp|udp --dport port -j REDIRECT

3. Reload the firewall rules and keep the state information.

# firewall-cmd --reload

The current permanent configuration will become the new firewalld runtime configuration as
well as the configuration at the next system start.

3.2.3. Direct Routing Using iptables

You may also work around the ARP issue using the direct routing method by creating iptables firewall
rules. To configure direct routing using iptables, you must add rules that create a transparent proxy so
that a real server will service packets sent to the VIP address, even though the VIP address does not
exist on the system.

The iptables method is simpler to configure than the arptables method. This method also circumvents
the LVS ARP issue entirely, because the virtual IP address(es) only exist on the active LVS director.

However, there are performance issues using the iptables method compared to arptables, as there is
overhead in forwarding/masquerading every packet.

You also cannot reuse ports using the iptables method. For example, it is not possible to run two
separate Apache HTTP Server services bound to port 80, because both must bind to INADDR_ANY
instead of the virtual IP addresses.

To configure direct routing using the iptables method, perform the following steps:

1. On each real server, enter the following command for every VIP, port, and protocol (TCP or
UDP) combination intended to be serviced for the real server:

iptables -t nat -A PREROUTING -p <tcp|udp> -d <vip> --dport <port> -j REDIRECT

This command will cause the real servers to process packets destined for the VIP and port that
they are given.

2. Save the configuration on each real server:

# iptables-save > /etc/sysconfig/iptables

# systemctl enable iptables.service

The systemctl enable command will cause the system to reload the iptables configuration on
bootup before the network is started.

3.2.4. Direct Routing Using sysctl

Another way to deal with the ARP limitation when employing Direct Routing is using the sysctl interface.
Administrators can configure two systcl settings such that the real server will not announce the VIP in
ARP requests and will not reply to ARP requests for the VIP address. To enable this, enter the following
commands:

echo 1 > /proc/sys/net/ipv4/conf/eth0/arp_ignore

echo 2 > /proc/sys/net/ipv4/conf/eth0/arp_announce

20
 CHAPTER 3. SETTING UP LOAD BALANCER PREREQUISITES FOR KEEPALIVED

Alternatively, you may add the following lines to the /etc/sysctl.d/arp.conf file:

net.ipv4.conf.eth0.arp_ignore = 1
net.ipv4.conf.eth0.arp_announce = 2

3.3. PUTTING THE CONFIGURATION TOGETHER

After determining which of the preceding routing methods to use, the hardware should be connected
together and configured.

IMPORTANT

The network adapters on the LVS routers must be configured to access the same
networks. For instance if eth0 connects to the public network and eth1 connects to the
private network, then these same devices on the backup LVS router must connect to the
same networks.

Also the gateway listed in the first interface to come up at boot time is added to the
routing table and subsequent gateways listed in other interfaces are ignored. This is
especially important to consider when configuring the real servers.

After connecting the hardware to the network, configure the network interfaces on the primary and
backup LVS routers. This should be done by editing the network configuration files manually. For more
information about working with network configuration files, see the Red Hat Enterprise Linux 7
Networking Guide.

3.3.1. General Load Balancer Networking Tips

Configure the real IP addresses for both the public and private networks on the LVS routers before
attempting to configure Load Balancer using Keepalived. The sections on each topology give example
network addresses, but the actual network addresses are needed. Below are some useful commands for
bringing up network interfaces or checking their status.

Bringing Up Real Network Interfaces

To open a real network interface, use the following command as root, replacing N with the number
corresponding to the interface (eth0 and eth1).

ifup ethN


WARNING

Do not use the ifup scripts to open any floating IP addresses you may configure
using Keepalived (eth0:1 or eth1:1). Use the service or systemctl command to
start keepalived instead.

Bringing Down Real Network Interfaces

To bring down a real network interface, use the following command as root, replacing N with the
21
Load Balancer Administration

To bring down a real network interface, use the following command as root, replacing N with the
number corresponding to the interface (eth0 and eth1).

ifdown ethN

Checking the Status of Network Interfaces

If you need to check which network interfaces are up at any given time, enter the following command:

ip link

To view the routing table for a machine, issue the following command:

ip route

3.3.2. Firewall Requirements

If you are running a firewall (by means of firewalld or iptables), you must allow VRRP traffic to pass
between the keepalived nodes. To configure the firewall to allow the VRRP traffic with firewalld, run
the following commands:

# firewall-cmd --add-rich-rule='rule protocol value="vrrp" accept' --permanent

# firewall-cmd --reload

If the zone is omitted the default zone will be used.

If, however, you need to allow the VRRP traffic with iptables, run the following commands:

# iptables -I INPUT -p vrrp -j ACCEPT

# iptables-save > /etc/sysconfig/iptables
# systemctl restart iptables

3.4. MULTI-PORT SERVICES AND LOAD BALANCER

LVS routers under any topology require extra configuration when creating multi-port Load Balancer
services. Multi-port services can be created artificially by using firewall marks to bundle together
different, but related protocols, such as HTTP (port 80) and HTTPS (port 443), or when Load Balancer
is used with true multi-port protocols, such as FTP. In either case, the LVS router uses firewall marks to
recognize that packets destined for different ports, but bearing the same firewall mark, should be
handled identically. Also, when combined with persistence, firewall marks ensure connections from the
client machine are routed to the same host, as long as the connections occur within the length of time
specified by the persistence parameter.

Although the mechanism used to balance the loads on the real servers, IPVS, can recognize the firewall
marks assigned to a packet, it cannot itself assign firewall marks. The job of assigning firewall marks must
be performed by the network packet filter, iptables. The default firewall administration tool in Red Hat
Enterprise Linux 7 is firewalld, which can be used to configure iptables. If preferred, iptables can be
used directly. See Red Hat Enterprise Linux 7 Security Guide for information on working with iptables in
Red Hat Enterprise Linux 7.

3.4.1. Assigning Firewall Marks Using firewalld

To assign firewall marks to a packet destined for a particular port, the administrator can use firewalld's
firewall-cmd utility.

22
 CHAPTER 3. SETTING UP LOAD BALANCER PREREQUISITES FOR KEEPALIVED

If required, confirm that firewalld is running:

# systemctl status firewalld

firewalld.service - firewalld - dynamic firewall daemon
Loaded: loaded (/usr/lib/systemd/system/firewalld.service; enabled)
Active: active (running) since Tue 2016-01-26 05:23:53 EST; 7h ago

To start firewalld, enter:

# systemctl start firewalld

To ensure firewalld is enabled to start at system start:

# systemctl enable firewalld

This section illustrates how to bundle HTTP and HTTPS as an example; however, FTP is another
commonly clustered multi-port protocol.

The basic rule to remember when using firewall marks is that for every protocol using a firewall mark in
Keepalived there must be a commensurate firewall rule to assign marks to the network packets.

Before creating network packet filter rules, make sure there are no rules already in place. To do this,
open a shell prompt, login as root, and enter the following command:

# firewall-cmd --list-rich-rules

If no rich rules are present the prompt will instantly reappear.

If firewalld is active and rich rules are present, it displays a set of rules.

If the rules already in place are important, check the contents of /etc/firewalld/zones/ and copy any
rules worth keeping to a safe place before proceeding. Delete unwanted rich rules using a command in
the following format:

firewall-cmd --zone=zone --remove-rich-rule='rule' --permanent

The --permanent option makes the setting persistent, but the command will only take effect at next
system start. If required to make the setting take effect immediately, repeat the command omitting the
--permanent option.

The first load balancer related firewall rule to be configured is to allow VRRP traffic for the Keepalived
service to function. Enter the following command:

# firewall-cmd --add-rich-rule='rule protocol value="vrrp" accept' --permanent

If the zone is omitted the default zone will be used.

Below are rules which assign the same firewall mark, 80, to incoming traffic destined for the floating IP
address, n.n.n.n, on ports 80 and 443.

# firewall-cmd --add-rich-rule='rule family="ipv4" destination address="n.n.n.n/32" port port="80"

protocol="tcp" mark set="80"' --permanent
# firewall-cmd --add-rich-rule='rule family="ipv4" destination address="n.n.n.n/32" port port="443"
protocol="tcp" mark set="80"' --permanent

23
Load Balancer Administration

# firewall-cmd --reload
success
# firewall-cmd --list-rich-rules
rule protocol value="vrrp" accept
rule family="ipv4" destination address="n.n.n.n/32" port port="80" protocol="tcp" mark set=80
rule family="ipv4" destination address="n.n.n.n/32" port port="443" protocol="tcp" mark set=80

If the zone is omitted the default zone will be used.

See the Red Hat Enterprise Linux 7 Security Guide for more information on the use of firewalld's rich
language commands.

3.4.2. Assigning Firewall Marks Using iptables

To assign firewall marks to a packet destined for a particular port, the administrator can use iptables.

This section illustrates how to bundle HTTP and HTTPS as an example; however, FTP is another
commonly clustered multi-port protocol.

The basic rule to remember when using firewall marks is that for every protocol using a firewall mark in
Keepalived there must be a commensurate firewall rule to assign marks to the network packets.

Before creating network packet filter rules, make sure there are no rules already in place. To do this,
open a shell prompt, login as root, and enter the following command:

/usr/sbin/service iptables status

If iptables is not running, the prompt will instantly reappear.

If iptables is active, it displays a set of rules. If rules are present, enter the following command:

/sbin/service iptables stop

If the rules already in place are important, check the contents of /etc/sysconfig/iptables and copy any
rules worth keeping to a safe place before proceeding.

The first load balancer related configuring firewall rules is to allow VRRP traffic for the Keepalived
service to function.

/usr/sbin/iptables -I INPUT -p vrrp -j ACCEPT

Below are rules which assign the same firewall mark, 80, to incoming traffic destined for the floating IP
address, n.n.n.n, on ports 80 and 443.

/usr/sbin/iptables -t mangle -A PREROUTING -p tcp -d n.n.n.n/32 -m multiport --dports 80,443 -j

MARK --set-mark 80

Note that you must log in as root and load the module for iptables before issuing rules for the first time.

In the above iptables commands, n.n.n.n should be replaced with the floating IP for your HTTP and
HTTPS virtual servers. These commands have the net effect of assigning any traffic addressed to the
VIP on the appropriate ports a firewall mark of 80, which in turn is recognized by IPVS and forwarded
appropriately.

24
 CHAPTER 3. SETTING UP LOAD BALANCER PREREQUISITES FOR KEEPALIVED


WARNING

The commands above will take effect immediately, but do not persist through a
reboot of the system.

3.5. CONFIGURING FTP

File Transport Protocol (FTP) is an old and complex multi-port protocol that presents a distinct set of
challenges to an Load Balancer environment. To understand the nature of these challenges, you must
first understand some key things about how FTP works.

3.5.1. How FTP Works

With most other server client relationships, the client machine opens up a connection to the server on a
particular port and the server then responds to the client on that port. When an FTP client connects to
an FTP server it opens a connection to the FTP control port 21. Then the client tells the FTP server
whether to establish an active or passive connection. The type of connection chosen by the client
determines how the server responds and on what ports transactions will occur.

The two types of data connections are:

Active Connections
When an active connection is established, the server opens a data connection to the client from port
20 to a high range port on the client machine. All data from the server is then passed over this
connection.

Passive Connections
When a passive connection is established, the client asks the FTP server to establish a passive
connection port, which can be on any port higher than 10,000. The server then binds to this high-
numbered port for this particular session and relays that port number back to the client. The client
then opens the newly bound port for the data connection. Each data request the client makes results
in a separate data connection. Most modern FTP clients attempt to establish a passive connection
when requesting data from servers.

NOTE

The client determines the type of connection, not the server. This means to effectively
cluster FTP, you must configure the LVS routers to handle both active and passive
connections.

The FTP client-server relationship can potentially open a large number of ports that
Keepalived does not know about.

3.5.2. How This Affects Load Balancer Routing

IPVS packet forwarding only allows connections in and out of the cluster based on it recognizing its port
number or its firewall mark. If a client from outside the cluster attempts to open a port IPVS is not
configured to handle, it drops the connection. Similarly, if the real server attempts to open a connection

25
Load Balancer Administration

back out to the Internet on a port IPVS does not know about, it drops the connection. This means all
connections from FTP clients on the Internet must have the same firewall mark assigned to them and all
connections from the FTP server must be properly forwarded to the Internet using network packet
filtering rules.

NOTE

In order to enable passive FTP connections, you must have the ip_vs_ftp kernel module
loaded. Run the following commands as an administrative user at a shell prompt to load
this module and and ensure that the module loads on a reboot:

echo "ip_vs_ftp" >> /etc/modules-load.d/ip_vs_ftp.conf

systemctl enable systemd-modules-load
systemctl start systemd-modules-load

3.5.3. Creating Network Packet Filter Rules

Before assigning any iptables rules for the FTP service, review the information in Section 3.4, “Multi-
port Services and Load Balancer ” concerning multi-port services and techniques for checking the
existing network packet filtering rules.

Below are rules which assign the same firewall mark, 21, to FTP traffic.

3.5.3.1. Rules for Active Connections

The rules for active connections tell the kernel to accept and forward connections coming to the internal
floating IP address on port 20 (the FTP data port).

The following iptables command allows the LVS router to accept outgoing connections from the real
servers that IPVS does not know about:

/usr/sbin/iptables -t nat -A POSTROUTING -p tcp -s n.n.n.0/24 --sport 20 -j MASQUERADE

In the iptables command, n.n.n should be replaced with the first three values for the floating IP for the
NAT interface's internal network interface defined virtual_server section of the keepalived.conf file.

3.5.3.2. Rules for Passive Connections

The rules for passive connections assign the appropriate firewall mark to connections coming in from the
Internet to the floating IP address for the service on a wide range of ports: 10,000 to 20,000.

26
 CHAPTER 3. SETTING UP LOAD BALANCER PREREQUISITES FOR KEEPALIVED


WARNING

If you are limiting the port range for passive connections, you must also configure
the FTP server, vsftpd, to use a matching port range. This can be accomplished by
adding the following lines to /etc/vsftpd.conf:

pasv_min_port=10000

pasv_max_port=20000

Setting pasv_address to override the real FTP server address should not be used
since it is updated to the virtual IP address by LVS.

For configuration of other FTP servers, consult the respective documentation.

This range should be a wide enough for most situations; however, you can increase this number to
include all available non-secured ports by changing 10000:20000 in the commands below to
1024:65535.

The following iptables commands have the net effect of assigning any traffic addressed to the floating
IP on the appropriate ports a firewall mark of 21, which is in turn recognized by IPVS and forwarded
appropriately:

/usr/sbin/iptables -t mangle -A PREROUTING -p tcp -d n.n.n.n/32 --dport 21 -j MARK --set-mark 21

/usr/sbin/iptables -t mangle -A PREROUTING -p tcp -d n.n.n.n/32 --dport 10000:20000 -j MARK --

set-mark 21

In the iptables commands, n.n.n.n should be replaced with the floating IP for the FTP virtual server
defined in the virtual_server subsection of the keepalived.conf file.

The commands above take effect immediately, but do not persist through a reboot of the system unless
they are saved. To save the changes, enter the following command:

# iptables-save > /etc/sysconfig/iptables

To ensure the iptables service is started at system start, enter the following command:

# systemctl enable iptables

You can verify whether the changes persist on a reboot by running the following command and checking
whether the changes remain:

# systemctl restart iptables

3.6. SAVING NETWORK PACKET FILTER SETTINGS

After configuring the appropriate network packet filters for your situation, save the settings so they can
be restored after a reboot. For iptables, enter the following command:

27
Load Balancer Administration

# iptables-save > /etc/sysconfig/iptables

To ensure the iptables service is started at system start, enter the following command:

# systemctl enable iptables

You can verify whether the changes persist on a reboot by running the following command and checking
whether the changes remain:

# systemctl restart iptables

See the Red Hat Enterprise Linux 7 Security Guide for more information on working with iptables in
Red Hat Enterprise Linux 7

3.7. TURNING ON PACKET FORWARDING AND NONLOCAL BINDING

In order for the Keepalived service to forward network packets properly to the real servers, each router
node must have IP forwarding turned on in the kernel. Log in as root and change the line which reads
net.ipv4.ip_forward = 0 in /etc/sysctl.conf to the following:

net.ipv4.ip_forward = 1

The changes take effect when you reboot the system.

Load balancing in HAProxy and Keepalived at the same time also requires the ability to bind to an IP
address that are nonlocal, meaning that it is not assigned to a device on the local system. This allows a
running load balancer instance to bind to an IP that is not local for failover.

To enable, edit the line in /etc/sysctl.conf that reads net.ipv4.ip_nonlocal_bind to the following:

net.ipv4.ip_nonlocal_bind = 1

The changes take effect when you reboot the system.

To check if IP forwarding is turned on, issue the following command as root:

/usr/sbin/sysctl net.ipv4.ip_forward

To check if nonlocal binding is turned on, issue the following command as root:

/usr/sbin/sysctl net.ipv4.ip_nonlocal_bind

If both the above commands return a 1, then the respective settings are enabled.

3.8. CONFIGURING SERVICES ON THE REAL SERVERS

If the real servers are Red Hat Enterprise Linux systems, set the appropriate server daemons to activate
at boot time. These daemons can include httpd for Web services or xinetd for FTP or Telnet services.

It may also be useful to access the real servers remotely, so the sshd daemon should also be installed
and running.

28
 CHAPTER 4. INITIAL LOAD BALANCER CONFIGURATION WITH KEEPALIVED

CHAPTER 4. INITIAL LOAD BALANCER CONFIGURATION

WITH KEEPALIVED
After installing Load Balancer packages, you must take some basic steps to set up the LVS router and
the real servers for use with Keepalived. This chapter covers these initial steps in detail.

4.1. A BASIC KEEPALIVED CONFIGURATION

In this basic example, two systems are configured as load balancers. LB1 (Active) and LB2 (Backup) will
be routing requests for a pool of four Web servers running httpd with real IP addresses numbered
192.168.1.20 to 192.168.1.24, sharing a virtual IP address of 10.0.0.1. Each load balancer has two interfaces
(eth0 and eth1), one for handling external Internet traffic, and the other for routing requests to the real
servers. The load balancing algorithm used is Round Robin and the routing method will be Network
Address Translation.

4.1.1. Creating the keapalived.conf file

Keepalived is configured by means of the keepalived.conf file in each system configured as a load
balancer. To create a load balancer topology like the example shown in Section 4.1, “A Basic Keepalived
configuration”, use a text editor to open keepalived.conf in both the active and backup load balancers,
LB1 and LB2. For example:

vi /etc/keepalived/keepalived.conf

A basic load balanced system with the configuration as detailed in Section 4.1, “A Basic Keepalived
configuration” has a keepalived.conf file as explained in the following code sections. In this example, the
keepalived.conf file is the same on both the active and backup routers with the exception of the VRRP
instance, as noted in Section 4.1.1.2, “VRRP Instance”

4.1.1.1. Global Definitions

The Global Definitions section of the keepalived.conf file allows administrators to specify notification
details when changes to the load balancer occurs. Note that the Global Definitions are optional and are
not required for Keepalived configuration. This section of the keepalived.conf file is the same on both
LB1 and LB2.

global_defs {

notification_email {
[email protected]
}
notification_email_from [email protected]
smtp_server 127.0.0.1
smtp_connect_timeout 60
}

The notification_email is the administrator of the load balancer, while the notification_email_from is
an address that sends the load balancer state changes. The SMTP specific configuration specifies the
mail server from which the notifications are mailed.

4.1.1.2. VRRP Instance

The following examples show the vrrp_sync_group stanza of the keeplalived.conf file in the master

29
Load Balancer Administration

The following examples show the vrrp_sync_group stanza of the keeplalived.conf file in the master
router and the backup router. Note that the state and priority values differ between the two systems.

The following example shows the vrrp_sync_group stanza for the keepalived.conf file in LB1, the
master router.

vrrp_sync_group VG1 {
group {
RH_EXT
RH_INT
}
}

vrrp_instance RH_EXT {
state MASTER
interface eth0
virtual_router_id 50
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass passw123
}
virtual_ipaddress {
10.0.0.1
}
}

vrrp_instance RH_INT {
state MASTER
interface eth1
virtual_router_id 2
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass passw123
}
virtual_ipaddress {
192.168.1.1
}
}

The following example shows the vrrp_sync_group stanza of the keepalived.conf file for LB2, the
backup router.

vrrp_sync_group VG1 {
group {
RH_EXT
RH_INT
}
}

vrrp_instance RH_EXT {
state BACKUP
interface eth0

30
 CHAPTER 4. INITIAL LOAD BALANCER CONFIGURATION WITH KEEPALIVED

virtual_router_id 50
priority 99
advert_int 1
authentication {
auth_type PASS
auth_pass passw123
}
virtual_ipaddress {
10.0.0.1
}
}

vrrp_instance RH_INT {
state BACKUP
interface eth1
virtual_router_id 2
priority 99
advert_int 1
authentication {
auth_type PASS
auth_pass passw123
}
virtual_ipaddress {
192.168.1.1
}
}

In these example, the vrrp_sync_group stanza defines the VRRP group that stays together through
any state changes (such as failover). There is an instance defined for the external interface that
communicates with the Internet (RH_EXT), as well as one for the internal interface (RH_INT).

The vrrp_instance line details the virtual interface configuration for the VRRP service daemon, which
creates virtual IP instances. The state MASTER designates the active server, the state BACKUP
designates the backup server.

The interface parameter assigns the physical interface name to this particular virtual IP instance.

virtual_router_id is a numerical identifier for the Virtual Router instance. It must be the same on all LVS
Router systems participating in this Virtual Router. It is used to differentiate multiple instances of
keepalived running on the same network interface.

The priority specifies the order in which the assigned interface takes over in a failover; the higher the
number, the higher the priority. This priority value must be within the range of 0 to 255, and the Load
Balancing server configured as state MASTER should have a priority value set to a higher number than
the priority value of the server configured as state BACKUP.

The authentication block specifies the authentication type ( auth_type) and password (auth_pass)
used to authenticate servers for failover synchronization. PASS specifies password authentication;
Keepalived also supports AH, or Authentication Headers for connection integrity.

Finally, the virtual_ipaddress option specifies the interface virtual IP address.

4.1.1.3. Virtual Server Definitions

The Virtual Server definitions section of the keepalived.conf file is the same on both LB1 and LB2.

31
Load Balancer Administration

virtual_server 10.0.0.1 80 {
delay_loop 6
lb_algo rr
lb_kind NAT
protocol TCP

real_server 192.168.1.20 80 {
TCP_CHECK {
connect_timeout 10
}
}
real_server 192.168.1.21 80 {
TCP_CHECK {
connect_timeout 10
}
}
real_server 192.168.1.22 80 {
TCP_CHECK {
connect_timeout 10
}
}
real_server 192.168.1.23 80 {
TCP_CHECK {
connect_timeout 10
}
}

In this block, the virtual_server is configured first with the IP address. Then a delay_loop configures
the amount of time (in seconds) between health checks. The lb_algo option specifies the kind of
algorithm used for availability (in this case, rr for Round-Robin; for a list of possible lb_algo values see
Table 4.1, “lv_algo Values for Virtual Server” ). The lb_kind option determines routing method, which in
this case Network Address Translation (or nat) is used.

After configuring the Virtual Server details, the real_server options are configured, again by specifying
the IP Address first. The TCP_CHECK stanza checks for availability of the real server using TCP. The
connect_timeout configures the time in seconds before a timeout occurs.

NOTE

Accessing the virtual IP from the load balancers or one of the real servers is not
supported. Likewise, configuring a load balancer on the same machines as a real server is
not supported.

Table 4.1. lv_algo Values for Virtual Server

Algorithm Name lv_algo value

Round-Robin rr

Weighted Round-Robin wrr

32
 CHAPTER 4. INITIAL LOAD BALANCER CONFIGURATION WITH KEEPALIVED

Algorithm Name lv_algo value

Least-Connection lc

Weighted Least-Connection wlc

Locality-Based Least-Connection lblc

Locality-Based Least-Connection Scheduling with Replication lblcr

Destination Hash dh

Source Hash sh

Source Expected Delay sed

Never Queue nq

4.2. KEEPALIVED DIRECT ROUTING CONFIGURATION

Direct Routing configuration of Keepalived is similar in configuration with NAT. In the following example,
Keepalived is configured to provide load balancing for a group of real servers running HTTP on port 80.
To configure Direct Routing, change the lb_kind parameter to DR. Other configuration options are
discussed in Section 4.1, “A Basic Keepalived configuration” .

The following example shows the keepalived.conf file for the active server in a Keepalived
configuration that uses direct routing.

global_defs {
notification_email {
[email protected]
}
notification_email_from [email protected]
smtp_server 127.0.0.1
smtp_connect_timeout 60
}

vrrp_instance RH_1 {
state MASTER
interface eth0
virtual_router_id 50
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass passw123
}
virtual_ipaddress {
172.31.0.1
}
}

33
Load Balancer Administration

virtual_server 172.31.0.1 80
delay_loop 10
lb_algo rr
lb_kind DR
persistence_timeout 9600
protocol TCP

real_server 192.168.0.1 80 {
weight 1
TCP_CHECK {
connect_timeout 10
connect_port 80
}
}
real_server 192.168.0.2 80 {
weight 1
TCP_CHECK {
connect_timeout 10
connect_port 80
}
}
real_server 192.168.0.3 80 {
weight 1
TCP_CHECK {
connect_timeout 10
connect_port 80
}
}
}

The following example shows the keepalived.conf file for the backup server in a Keepalived
configuration that uses direct routing. Note that the state and priority values differ from the
keepalived.conf file in the active server.

global_defs {
notification_email {
[email protected]
}
notification_email_from [email protected]
smtp_server 127.0.0.1
smtp_connect_timeout 60
}

vrrp_instance RH_1 {
state BACKUP
interface eth0
virtual_router_id 50
priority 99
advert_int 1
authentication {
auth_type PASS
auth_pass passw123
}
virtual_ipaddress {
172.31.0.1
}

34
 CHAPTER 4. INITIAL LOAD BALANCER CONFIGURATION WITH KEEPALIVED

virtual_server 172.31.0.1 80
delay_loop 10
lb_algo rr
lb_kind DR
persistence_timeout 9600
protocol TCP

real_server 192.168.0.1 80 {
weight 1
TCP_CHECK {
connect_timeout 10
connect_port 80
}
}
real_server 192.168.0.2 80 {
weight 1
TCP_CHECK {
connect_timeout 10
connect_port 80
}
}
real_server 192.168.0.3 80 {
weight 1
TCP_CHECK {
connect_timeout 10
connect_port 80
}
}
}

4.3. STARTING THE SERVICE

Start the service by entering the following command on the servers in the load balancer configuration:

# systemctl start keepalived.service

To make Keepalived service persist through reboots, enter the following command on the servers in the
load balancer configuration:

# systemctl enable keepalived.service

35
Load Balancer Administration

CHAPTER 5. HAPROXY CONFIGURATION

This chapter explains the configuration of a basic setup that highlights the common configuration
options an administrator could encounter when deploying HAProxy services for high availability
environments.

HAProxy has its own set of scheduling algorithms for load balancing. These algorithms are described in
Section 5.1, “HAProxy Scheduling Algorithms” .

HAProxy is configured by editing the /etc/haproxy/haproxy.cfg file.

Load Balancer configuration using HAProxy consists of five sections for configuration:

Section 5.2, “Global Settings”

The proxies section, which consists of 4 subsections:

The Section 5.3, “Default Settings” settings

The Section 5.4, “Frontend Settings” settings

The Section 5.5, “Backend Settings” settings

5.1. HAPROXY SCHEDULING ALGORITHMS

The HAProxy scheduling algorithms for load balancing can be edited in the balance parameter in the
backend section of the /etc/haproxy/haproxy.cfg configuration file. Note that HAProxy supports
configuration with multiple back ends, and each back end can be configured with a scheduling algorithm.

Round-Robin (roundrobin)
Distributes each request sequentially around the pool of real servers. Using this algorithm, all the real
servers are treated as equals without regard to capacity or load. This scheduling model resembles
round-robin DNS but is more granular due to the fact that it is network-connection based and not
host-based. Load Balancer round-robin scheduling also does not suffer the imbalances caused by
cached DNS queries. However, in HAProxy, since configuration of server weights can be done on the
fly using this scheduler, the number of active servers are limited to 4095 per back end.

Static Round-Robin (static-rr)

Distributes each request sequentially around a pool of real servers as does Round-Robin, but does
not allow configuration of server weight dynamically. However, because of the static nature of server
weight, there is no limitation on the number of active servers in the back end.

Least-Connection (leastconn)
Distributes more requests to real servers with fewer active connections. Administrators with a
dynamic environment with varying session or connection lengths may find this scheduler a better fit
for their environments. It is also ideal for an environment where a group of servers have different
capacities, as administrators can adjust weight on the fly using this scheduler.

Source (source)
Distributes requests to servers by hashing requesting source IP address and dividing by the weight of
all the running servers to determine which server will get the request. In a scenario where all servers
are running, the source IP request will be consistently served by the same real server. If there is a
change in the number or weight of the running servers, the session may be moved to another server
because the hash/weight result has changed.

36
 CHAPTER 5. HAPROXY CONFIGURATION

URI (uri)
Distributes requests to servers by hashing the entire URI (or a configurable portion of a URI) and
divides by the weight of all the running servers to determine which server will the request. In a
scenario where all active servers are running, the destination IP request will be consistently served by
the same real server. This scheduler can be further configured by the length of characters at the start
of a directory part of a URI to compute the hash result and the depth of directories in a URI
(designated by forward slashes in the URI) to compute the hash result.

URL Parameter (url_param)

Distributes requests to servers by looking up a particular parameter string in a source URL request
and performing a hash calculation divided by the weight of all running servers. If the parameter is
missing from the URL, the scheduler defaults to Round-robin scheduling. Modifiers may be used
based on POST parameters as well as wait limits based on the number of maximum octets an
administrator assigns to the weight for a certain parameter before computing the hash result.

Header Name (hdr)

Distributes requests to servers by checking a particular header name in each source HTTP request
and performing a hash calculation divided by the weight of all running servers. If the header is absent,
the scheduler defaults to Round-robin scheduling.

RDP Cookie (rdp-cookie)

Distributes requests to servers by looking up the RDP cookie for every TCP request and performing
a hash calculation divided by the weight of all running servers. If the header is absent, the scheduler
defaults to Round-robin scheduling. This method is ideal for persistence as it maintains session
integrity.

5.2. GLOBAL SETTINGS

The global settings configure parameters that apply to all servers running HAProxy. A typical global
section may look like the following:

global
log 127.0.0.1 local2
maxconn 4000
user haproxy
group haproxy
daemon

In the above configuration, the administrator has configured the service to log all entries to the local
syslog server. By default, this could be /var/log/syslog or some user-designated location.

The maxconn parameter specifies the maximum number of concurrent connections for the service. By
default, the maximum is 2000.

The user and group parameters specifies the user name and group name for which the haproxy
process belongs.

Finally, the daemon parameter specifies that haproxy run as a background process.

5.3. DEFAULT SETTINGS

37
Load Balancer Administration

The default settings configure parameters that apply to all proxy subsections in a configuration
(frontend, backend, and listen). A typical default section may look like the following:

NOTE

Any parameter configured in proxy subsection (frontend, backend, or listen) takes

precedence over the parameter value in default.

defaults
mode http
log global
option httplog
option dontlognull
retries 3
timeout http-request 10s
timeout queue 1m
timeout connect 10s
timeout client 1m
timeout server 1m

mode specifies the protocol for the HAProxy instance. Using the http mode connects source requests
to real servers based on HTTP, ideal for load balancing web servers. For other applications, use the tcp
mode.

log specifies log address and syslog facilities to which log entries are written. The global value refers
the HAProxy instance to whatever is specified in the log parameter in the global section.

option httplog enables logging of various values of an HTTP session, including HTTP requests, session
status, connection numbers, source address, and connection timers among other values.

option dontlognull disables logging of null connections, meaning that HAProxy will not log connections
wherein no data has been transferred. This is not recommended for environments such as web
applications over the Internet where null connections could indicate malicious activities such as open
port-scanning for vulnerabilities.

retries specifies the number of times a real server will retry a connection request after failing to connect
on the first try.

The various timeout values specify the length of time of inactivity for a given request, connection, or
response. These values are generally expressed in milliseconds (unless explicitly stated otherwise) but
may be expressed in any other unit by suffixing the unit to the numeric value. Supported units are us
(microseconds), ms (milliseconds), s (seconds), m (minutes), h (hours) and d (days). http-request 10s
gives 10 seconds to wait for a complete HTTP request from a client. queue 1m sets one minute as the
amount of time to wait before a connection is dropped and a client receives a 503 or "Service
Unavailable" error. connect 10s specifies the number of seconds to wait for a successful connection to a
server. client 1m specifies the amount of time (in minutes) a client can remain inactive (it neither
accepts nor sends data). server 1m specifies the amount of time (in minutes) a server is given to accept
or send data before timeout occurs.

5.4. FRONTEND SETTINGS

The frontend settings configure the servers' listening sockets for client connection requests. A typical
HAProxy configuration of the frontend may look like the following:

38
 CHAPTER 5. HAPROXY CONFIGURATION

frontend main
bind 192.168.0.10:80
default_backend app

The frontend called main is configured to the 192.168.0.10 IP address and listening on port 80 using the
bind parameter. Once connected, the use backend specifies that all sessions connect to the app back
end.

5.5. BACKEND SETTINGS

The backend settings specify the real server IP addresses as well as the load balancer scheduling
algorithm. The following example shows a typical backend section:

backend app
balance roundrobin
server app1 192.168.1.1:80 check
server app2 192.168.1.2:80 check
server app3 192.168.1.3:80 check inter 2s rise 4 fall 3
server app4 192.168.1.4:80 backup

The back-end server is named app. The balance specifies the load balancer scheduling algorithm to be
used, which in this case is Round Robin (roundrobin), but can be any scheduler supported by HAProxy.
For more information configuring schedulers in HAProxy, see Section 5.1, “HAProxy Scheduling
Algorithms”.

The server lines specify the servers available in the back end. app1 to app4 are the names assigned
internally to each real server. Log files will specify server messages by name. The address is the assigned
IP address. The value after the colon in the IP address is the port number to which the connection occurs
on the particular server. The check option flags a server for periodic healthchecks to ensure that it is
available and able receive and send data and take session requests. Server app3 also configures the
healthcheck interval to two seconds (inter 2s), the amount of checks app3 has to pass to determine if
the server is considered healthy (rise 4), and the number of times a server consecutively fails a check
before it is considered failed (fall 3).

5.6. STARTING HAPROXY

To start the HAProxy service, enter the following command:

# systemctl start haproxy.service

To make the HAProxy service persist through reboots, enter the following command:

# systemctl enable haproxy.service

5.7. LOGGING HAPROXY MESSAGES TO RSYSLOG

You can configure your system to log HAProxy messages to rsyslog by writing to the /dev/log socket.
Alternately you can target the TCP loopback address, however this results in slower performance.

The following procedure configures HAProxy to log messages to rsyslog.

39
Load Balancer Administration

1. In the global section of the HAProxy configuration file, use the log directive to target the
/dev/log socket.

log /dev/log local0

2. Update the frontend, backend, and listen proxies to send messages to the rsyslog service you
configured in the global section of the HAProxy configuration file. To do this, add a log global
directive to the defaults section of the configuration file, as shown.

defaults
log global
option httplog

3. If you are running HAProxy within a chrooted environment, or you let HAProxy create a chroot
directory for you by using the chroot configuration directive, then the socket must be made
available within that chroot directory. You can do this by modifying the rsyslog configuration to
create a new listening socket within the chroot filesystem. To do this, add the following lines to
your rsyslog configuration file.

$ModLoad imuxsock
$AddUnixListenSocket PATH_TO_CHROOT/dev/log

4. To customize what and where HAProxy log messages will appear, you can use rsyslog filters as
described in Basic Configuration of Rsyslog in the System Administrator's Guide.

40
EXAMPLE CONFIGURATION: LOAD BALANCING CEPH OBJECT GATEWAY SERVERS WITH HAPROXY AND KEEPALIVED

APPENDIX A. EXAMPLE CONFIGURATION: LOAD BALANCING

CEPH OBJECT GATEWAY SERVERS WITH HAPROXY AND
KEEPALIVED
This appendix provides an example showing the configuration of HAProxy and Keepalived with a Ceph
cluster. The Ceph Object Gateway allows you to assign many instances of the object gateway to a single
zone so that you can scale out as load increases. Since each object gateway instance has its own IP
address, you can use HAProxy and keepalived to balance the load across Ceph Object Gateway servers.

In this configuration, HAproxy performs the load balancing across Ceph Object Gateway servers while
Keepalived is used to manage the Virtual IP addresses of the Ceph Object Gateway servers and to
monitor HAProxy.

Another use case for HAProxy and keepalived is to terminate HTTPS at the HAProxy server. Red Hat
Ceph Storage (RHCS) 1.3.x uses Civetweb, and the implementation in RHCS 1.3.x does not support
HTTPS. You can use an HAProxy server to terminate HTTPS at the HAProxy server and use HTTP
between the HAProxy server and the Civetweb gateway instances. This example includes this
configuration as part of the procedure.

A.1. PREREQUISITES
To set up HAProxy with the Ceph Object Gateway, you must have:

A running Ceph cluster;

At least two Ceph Object Gateway servers within the same zone configured to run on port 80;

At least two servers for HAProxy and keepalived.

NOTE

This procedure assumes that you have at least two Ceph Object Gateway servers
running, and that you get a valid response when running test scripts over port 80.

A.2. PREPARING HAPROXY NODES

The following setup assumes two HAProxy nodes named haproxy and haproxy2, and two Ceph Object
Gateway servers named rgw1 and rgw2. You may use any naming convention you prefer. Perform the
following procedure on your two HAProxy nodes:

1. Install Red Hat Enterprise Linux 7.

2. Register the nodes.

# subscription-manager register

3. Enable the Red Hat Enterprise Linux 7 server repository.

# subscription-manager repos --enable=rhel-7-server-rpms

4. Update the server.

41
Load Balancer Administration

# yum update -y

5. Install admin tools (for example, wget, vim, and so on) as needed.

6. Open port 80.

# firewall-cmd --zone=public --add-port 80/tcp --permanent

# firewall-cmd --reload

7. For HTTPS, open port 443.

# firewall-cmd --zone=public --add-port 443/tcp --permanent

# firewall-cmd --reload

A.3. INSTALL AND CONFIGURE KEEPALIVED

Perform the following procedure on your two HAProxy nodes:

1. Install keepalived.

# yum install -y keepalived

2. Configure keepalived.

# vim /etc/keepalived/keepalived.conf

In the following configuration, there is a script to check the HAProxy processes. The instance
uses eth0 as the network interface and configures haproxy as the master server and haproxy2
as the backup server. It also assigns a virtual IP address of 192.168.0.100.

vrrp_script chk_haproxy {
script "killall -0 haproxy" # check the haproxy process
interval 2 # every 2 seconds
weight 2 # add 2 points if OK
}

vrrp_instance VI_1 {
interface eth0 # interface to monitor
state MASTER # MASTER on haproxy, BACKUP on haproxy2
virtual_router_id 51
priority 101 # 101 on haproxy, 100 on haproxy2
virtual_ipaddress {
192.168.0.100 # virtual ip address
}
track_script {
chk_haproxy
}
}

3. Enable and start keepalived.

42
EXAMPLE CONFIGURATION: LOAD BALANCING CEPH OBJECT GATEWAY SERVERS WITH HAPROXY AND KEEPALIVED

# systemctl enable keepalived

# systemctl start keepalived

A.4. INSTALL AND CONFIGURE HAPROXY

Perform the following procedure on your two HAProxy nodes:

1. Install haproxy.

# yum install haproxy

2. Configure haproxy for SELinux and HTTP.

# vim /etc/firewalld/services/haproxy-http.xml

Add the following lines:

<?xml version="1.0" encoding="utf-8"?>

<service>
<short>HAProxy-HTTP</short>
<description>HAProxy load-balancer</description>
<port protocol="tcp" port="80"/>
</service>

As root, assign the correct SELinux context and file permissions to the haproxy-http.xml file.

# cd /etc/firewalld/services
# restorecon haproxy-http.xml
# chmod 640 haproxy-http.xml

3. If you intend to use HTTPS, configure haproxy for SELinux and HTTPS.

# vim /etc/firewalld/services/haproxy-https.xml

Add the following lines:

<?xml version="1.0" encoding="utf-8"?>

<service>
<short>HAProxy-HTTPS</short>
<description>HAProxy load-balancer</description>
<port protocol="tcp" port="443"/>
</service>

As root, assign the correct SELinux context and file permissions to the haproxy-https.xml file.

# cd /etc/firewalld/services
# restorecon haproxy-https.xml
# chmod 640 haproxy-https.xml

4. If you intend to use HTTPS, generate keys for SSL. If you do not have a certificate, you may use
a self-signed certificate. For information on generating keys and on self-signed certificates, see
the Red Hat Enterprise Linux System Administrator's Guide.

43
Load Balancer Administration

Finally, put the certificate and key into a PEM file.

# cat example.com.crt example.com.key > example.com.pem

# cp example.com.pem /etc/ssl/private/

5. Configure HAProxy.

# vim /etc/haproxy/haproxy.cfg

The global and defaults sections of haproxy.cfg may remain unchanged. After the defaults
sections, you will need to configure frontend and backend sections, as in the following
example:

frontend http_web *:80

mode http
default_backend rgw

frontend rgw​-https
bind <insert vip ipv4>:443 ssl crt /etc/ssl/private/example.com.pem
default_backend rgw

backend rgw
balance roundrobin
mode http
server rgw1 10.0.0.71:80 check
server rgw2 10.0.0.80:80 check

6. Enable/start haproxy

# systemctl enable haproxy

# systemctl start haproxy

A.5. TEST YOUR HAPROXY CONFIGURATION

On your HAProxy nodes, check to ensure the virtual IP address from your keepalived configuration
appears.

$ ip addr show

On your calamari node, see if you can reach the gateway nodes by means of the load balancer
configuration. For example:

$ wget haproxy

This should return the same result as:

$ wget rgw1

If it returns an index.html file with the following contents then your configuration is working properly.

<?xml version="1.0" encoding="UTF-8"?>

<ListAllMyBucketsResult xmlns="http://s3.amazonaws.com/doc/2006-03-01/">

44
EXAMPLE CONFIGURATION: LOAD BALANCING CEPH OBJECT GATEWAY SERVERS WITH HAPROXY AND KEEPALIVED

<Owner>
<ID>anonymous</ID>
<DisplayName></DisplayName>
</Owner>
<Buckets>
</Buckets>
</ListAllMyBucketsResult>

45
Load Balancer Administration

APPENDIX B. REVISION HISTORY

Revision 4.1-1 Wed Aug 7 2019 Steven Levine
Preparing document for 7.7 GA publication.

Revision 3.1-2 Thu Oct 4 2018 Steven Levine

Preparing document for 7.6 GA publication.

Revision 2.1-1 Thu Mar 15 2018 Steven Levine

Preparing document for 7.5 GA publication.

Revision 2.1-0 Thu Dec 14 2017 Steven Levine

Preparing document for 7.5 Beta publication.

Revision 0.6-5 Wed Nov 22 2017 Steven Levine

Updated version for 7.4.

Revision 0.6-3 Thu Jul 27 2017 Steven Levine

Document version for 7.4 GA publication.

Revision 0.6-1 Wed May 10 2017 Steven Levine

Preparing document for 7.4 Beta publication.

Revision 0.5-9 Mon Dec 5 2016 Steven Levine

Updated Version for 7.3.

Revision 0.5-7 Mon Oct 17 2016 Steven Levine

Version for 7.3 GA publication.

Revision 0.5-6 Thu Aug 18 2016 Steven Levine

Preparing document for 7.3 Beta publication.

Revision 0.3-2 Mon Nov 9 2015 Steven Levine

Preparing document for 7.2 GA publication.

Revision 0.3-0 Wed Aug 19 2015 Steven Levine

Preparing document for 7.2 Beta publication.

Revision 0.2-6 Mon Feb 16 2015 Steven Levine

Version for 7.1 GA release

Revision 0.2-5 Thu Dec 11 2014 Steven Levine

Version for 7.1 Beta release

Revision 0.2-4 Thu Dec 04 2014 Steven Levine

Version for 7.1 Beta release

Revision 0.1-12 Tue Jun 03 2014 John Ha

Version for 7.0 GA Release

Revision 0.1-6 Mon Jun 13 2013 John Ha

Build for beta of Red Hat Enterprise Linux 7

Revision 0.1-1 Wed Jan 16 2013 John Ha

Branched from Red Hat Enterprise Linux 6 version of this Document

46
 INDEX

INDEX
A
arptables, Direct Routing Using arptables

D
direct routing
and arptables, Direct Routing Using arptables
and firewalld, Direct Routing Using firewalld

F
firewalld, Direct Routing Using firewalld
FTP, Configuring FTP
(see also Load Balancer )

H
HAProxy, haproxy
HAProxy and Keepalived, keepalived and haproxy

J
job scheduling, Keepalived , keepalived Scheduling Overview

K
Keepalived
configuration, A Basic Keepalived configuration
configuration file, Creating the keapalived.conf file
initial configuration, Initial Load Balancer Configuration with Keepalived
job scheduling, keepalived Scheduling Overview
scheduling, job, keepalived Scheduling Overview

Keepalived configuration
Direct Routing, Keepalived Direct Routing Configuration

keepalived daemon, keepalived

keepalived.conf, Creating the keapalived.conf file
Keepalivedd
LVS routers
primary node, Initial Load Balancer Configuration with Keepalived

L
least connections (see job scheduling, Keepalived )

47
Load Balancer Administration

Load Balancer
direct routing
and arptables, Direct Routing Using arptables
and firewalld, Direct Routing Using firewalld
requirements, hardware, Direct Routing, Load Balancer Using Direct Routing
requirements, network, Direct Routing, Load Balancer Using Direct Routing
requirements, software, Direct Routing, Load Balancer Using Direct Routing

HAProxy, haproxy
HAProxy and Keepalived, keepalived and haproxy
Keepalived, A Basic Keepalived configuration, Keepalived Direct Routing Configuration
keepalived daemon, keepalived
multi-port services, Multi-port Services and Load Balancer
FTP, Configuring FTP

NAT routing
requirements, hardware, The NAT Load Balancer Network
requirements, network, The NAT Load Balancer Network
requirements, software, The NAT Load Balancer Network

packet forwarding, Turning on Packet Forwarding and Nonlocal Binding

routing methods
NAT, Routing Methods

routing prerequisites, Configuring Network Interfaces for Load Balancer with NAT
three-tier, A Three-Tier keepalived Load Balancer Configuration

LVS
NAT routing
enabling, Enabling NAT Routing on the LVS Routers

overview of, Load Balancer Overview

real servers, Load Balancer Overview

M
multi-port services, Multi-port Services and Load Balancer
(see also Load Balancer )

N
NAT
enabling, Enabling NAT Routing on the LVS Routers
routing methods, Load Balancer ,Routing Methods

network address translation (see NAT)

48
 INDEX

P
packet forwarding, Turning on Packet Forwarding and Nonlocal Binding
(see also Load Balancer)

R
real servers
configuring services, Configuring Services on the Real Servers

round robin (see job scheduling, Keepalived )

routing
prerequisites for Load Balancer ,Configuring Network Interfaces for Load Balancer with NAT

S
scheduling, job (Keepalived ), keepalived Scheduling Overview

W
weighted least connections (see job scheduling, Keepalived )
weighted round robin (see job scheduling, Keepalived )

49