Heartbeat

The heartbeat program is one of the core components of the Linux-HA (High-Availability Linux) project.

At this point in time (May 2002), we estimate there are a few thousand HA clusters in production using heartbeat. Below are a few sample success stories.

Tony Willoughby of ADC Telecommunications
We use Heartbeat (and DRBD) in two products. The first is the Cuda 12000 IP Access Switch.

http://www6.adc.com/ecom/hier?NODE=OND48067

This is a Cable Modem Termination System (CMTS) which uses Linux for service provisioning and network management. From the web page:

Contains integrated service provisioning and element/network management functions eliminating the need for separate systems while capitalizing on the high availability Linux-based management module.

The second product is the FastFlow Broadband Provisioning Manager.

http://www6.adc.com/ecom/hier?NODE=OND48857

The FBPM provides the same provisioning services that are available on the Cuda 12000 on a stand alone Linux platform.

Brian Tinsley of Emageon
Emageon - Birmingham, AL
We use heartbeat to provide resource failover in our Linux cluster solutions that power our medical image archive/distribution systems. We are very proud of what we have been able to accomplish with Linux, but without software like heartbeat and Linux Virtual Server our solutions would not be half as good as what they have become. Perhaps one day your X-Ray or CAT scans will wind up passing through one of these Linux systems!

Alan Richter from Lujan Security Systems

Hello Alan.

I am using Heartbeat 0.4.9 in a commercial situation.

I work for a company (Lujsansoft) which sells Linux based badgereader systems. I am using VNC so that a. the customers can access it from whatever kind of computer they are using and b. Comtrol rocket port serial hubs so we don't have to deal with the headaches associated with long streches of RS232 or RS485 cabling.

I am using heartbeat to run the badge reader program, the driver for the serial hubs and VNC and it all works beautifully. The installation where I installed my two node cluster is a linear accelerator in Los Alamos NM USA and the badge reader system MUST stay up while the "beam" is running, it is a matter of health and safety.

Setting up heartbeat was cake and pie (piece of cake, easy as pie) I can't believe how easy it is to set up and use. Oh yeah it works perfectly too.

That's my story.

Thanks for maintaining the site. I appreciate it.

Joe Henggeler from The Weather Channel
We currently utilize heartbeat on a two-node cluster which handles transient storage for delivery/receipt of (1) critical weather data for our customers, and (2) video footage from cameramen. The hardware is : 2 Intel 1U's each connected to a 2-channel 350 GB RAID array. The services this cluster offers are ftp, nfs, http, and other custom services. The OS is Mandrake (to take advantage of ReiserFS). A future use for heartbeat within our organization is to port this to other operating systems (eg., Solaris, FreeBSD) for even more mission critical services.

From my experience, I was extremely impressed with the ease of installation/compilation on linux, and the stability of the cluster. This cluster has been running for approximately eight months (with forced manual failovers for updates and maintenance), and heartbeat has been running solid with virtually no interruptions in service.

Thanks, Alan, for maintaining this site!

Jerker Nyberg from Levonline AB
Hello Alan,

We use Heartbeat together with LVS for redundant load balancing of the webservers on Levonline AB, a webhosting company in Stockholm Sweden.

We also use Heartbeat for redundant MySQL (data is written to two servers and read from the active one), DNS and routing on the same machines.

Here is a simple illustration of our setup.

http://www.levonline.com/support/#lvs

The machines running Heartbeat are two Dell PowerEdge 2300.

Thank you for great software.

Rubin Bennett at completeconnection.com
I've done 2 that I can tell you about:

One of the leading Bagel quick service restaurant chains in the country has a Mandrake 7.2 cluster running hearbeat and DRBD/ ReiserFS. It serves as the Intranet/ SMTP gateway system for the entire organization (4000+ employees). It also runs an internal website that takes critical data from a legacy DOS system in the stores and imports it via a series of PHP pages and SQL*Plus scripts into the company's Oracle database. That site is also served by the HA cluster, and the data that gets imported is replicated between the systems.

A small independent Cable Television company in Northfield, Vermont (http://www.trans-video.net - name used with permission) HA cluster that runs all of the client/ user services for their High Speed Cable Internet Service: DHCP (failover configuration) for provisioning both Cable modems and Customer systems
TFTP (boot file for cable modems)
WWW site / home pages for users (Apache)
FTP from internal addresses for users to update their sites
SMTP for users mail

All of the above services are managed by Heartbeat for high availability. A slightly outdated description of the installation is documented at http://www.completeconnection.com/?target=/projects/tvcable.php

Renzo Alejandro Granados at conectiva.com.co
Alpopular SA: www.alpopular.com.co: Highly Available Internet/Intranet solution (web, mail, webmail, proxy, dns).

2 IBM Netfinity 5500 (M10 and M20, 256 Mb RAM, ~18 Gb Hot Swap, etc...). Using Conectiva Linux 6.0 (up to date).

Gives service to ~200 Internal Users, and access to their website (internal webmail and external content) to the Internet.

Gerfor SA: High Available Application servers (xdmcp, vnc) Giving access to applications such as StarOffice, Browsers, and clients of their mission critical system.

2 Compaq ML 350 servers (933 Mhz, 1Gb, 18Gb available to the OS after hw raid, 3 NICs) using Conectiva Linux 6.0 (up to date).

It gives service to 90 "thin clients" (old PCs) normally each server has the half of the clients ("load balancing").

All these implementations were made by Conectiva Colombia (www.conectiva.com.co)

Joel Fowler at Web-Com.com
Web-Com Development is a web systems development firm located in Santa Clara, California. We also offer hosting for sites we develop. Some examples include our own www.web-com.com and www.democracynow.net .

We are currently in the process of developing webSentry, a web site monitoring and service management application, which will be offered as a service. The application is not yet publically available.

High availability was a requirement of webSentry. It includes both batch and online elements that are managed by heartbeat . HA was also beneficial to the other virtual web sites supported by these clustered servers.

Our application infrastructure consists of the following components:
RH Linux
MySQL (with two-way replication between the clustered servers)
Apache (including perl cgi, mod_perl, mod_jk, and mod_ssl)
Tomcat

Resources under heartbeat control:
production VIP
serviceNanny.pl
webSentry Collectors (batch)

serviceNanny.pl is a perl script which (1) starts Tomcat, (2) monitors the health of important online resources (Apache, Tomcat, MySQL, DNS, and gateway access), and (3) takes appropriate remedial action (e.g. restarts Tomcat, initiates heartbeat failover, and reboots the system) to ensure online availability.

Peter Mueller at sidestep.com
sidestep.com

we utilize heartbeat in an LVS cluster with mon & ldirectord. future plans for linux clusters include file servers & various other projects, non definite yet. great product :)

Damian Ohara at motorola
Hi Alan,

We have 4 HA pairs running here at Motorola in Swindon, UK.

Pair1:
HP Netvectras (PII 233MHz), 3GB HD, 96MB RAM, 2x 3com 905, SuSE Serves SMB front for NFS mounted data on EMC Celerras

Pair2:
HP Netvectras (PII 233MHz), 3GB HD, 64MB RAM, 2x 3com 905, SuSE Serves site Anonymous FTP front for NFS mounted data on EMC Celerras

Pair3:
HP Netvectras (PII 233MHz), 3GB HD, 96MB RAM, 2x 3com 905, SuSE Serves SMB front for NFS mounted data on EMC Celerras

Pair4:
HP Netvectras (PII 233MHz), 3GB HD, 64MB RAM, 2x 3com 905, SuSE Serves SMB front for NFS mounted data on EMC Celerras

All machines are reclaimed - obsoleted from the desktop when we moved to Win2K.

We moved to Linux for this service becuase SMB on Solaris is flaky as hell (our experience) andon HP-UX is slow slow slow.

Another benefit we've found is that we can take the servers down during the day for maintenance without impact to the customer - just failover and get started!

Thanks for a great app.

Damian

Greg Louis at dynamicro.on.ca
Consultronics Limited (a small multinational manufacturer of telecommunications test equipment) has at its head office a perimeter network (DMZ), of which the outer and inner bastions are pairs of Linux machines. At the time of writing, these are running kernel 2.4.13-ac4 with heartbeat-0.4.9.0k to provide service IP addresses. Our marketing folks want 24x365 availability of http://www.consultronics.com (we sell in 80 countries and the sun never sets on our customer base), and the financial and manufacturing staff want to be sure of access (through the tri-homed inner bastion) to the inner sanctum where the company's ERP system is kept. Running heartbeat on the bastions was a first step: the intent is to make the webserver itself an HA pair and to add a secondary route to the Internet to back up the existing T1.

Patrick Scannel at fws.gov
I just used heartbeat and rsync to set up a 2 unit HA cluster that is a Samba PDC for a 100 user network. Samba does not offer BDC capabilities yet, so without your HA software, I would not have made the switch to open source. Thanks. The machines run RH 7.2 and Samba 2.2.2. They are old desktop reject machines, one of which had no available serial ports, so I stuck 3 nics in each of them. One for the network, and 2 for heartbeat/rsync. An easy heartbeat over USB would be nice.

Greg Freemyer at norcrossgroup.com
I've put in 2 HA installs for 2 different customers. Both are on Alpha's running Tru64 5.x I ported heartbeat 4.7 about 18 months ago, and that was what I used. With both I use externally shared storage, which I move from node to node. One customer is an ISP and is running the typical suite of ISP apps: radius, ftp, web-server, smtp, pop. The other is a small e-commerce company, and they are running Oracle DB and Oracle Application Server (OAS). Getting OAS to work with heartbeat was the most difficult issue I faced due to the security key being associated with the hostname.

Alex Kramarov at incredimail.com
w4.incredimail.com and w5.incredimail.com running as a cluster on www2 incredimail.com with 0.4.9.0g (next week will upgrade to the latest). They run web cluster with apache.

Kirk Lawson at heapy.com
Heapy Engineering (www.heapy.com) uses Heartbeat 0.4.7-1 with inclusion of AudibleAlarm on a two node fail-over cluster (no load balancing) Intranet Web Server. Even though it's "old" code, it's worked perfectly for us.

home
about this site
contact
legal
help