HAproxy + failover on same nodes as apache

failoverhaproxy

We're currently using HAproxy to load balance our image and application servers, but we haven't setup failover. We currently just have haproxy on one of the image boxes which then routes the requests where they need to go. Now, next week we are moving to our new colo (14 servers, it's time to stop leasing hardware).

We have 5 front end application servers. is it possible for me to put haproxy on two of those boxes, have a shared ip for failover, and simply move apache to port 9000 or some other arbitrary port? We're already at our budget for the colo for the next couple months, so I would like to avoid having two dedicated boxes for JUST HAproxy if we can avoid it.

The 5 application servers house apache and sphinx. The application servers them communicate with our mysql servers and our growing cassandra cluster.

Best Answer

HAProxy doesn't care if the web servers are on the local machine. Adding Apache to the same machine and using different ports is perfectly valid. Some day you may want to dedicate HAProxy but you probably don't need to yet (HAProxy has very low resource usage for Stack Overflow Currently).

Another option besides the multiple ports would be to have multiple IPs on each machine and bind to those. You can bind to a:

wildcard ip and a specific port
wildcard port and a specific ip
wildcard port and wildcard ip
specific ip and specific port

Also to bind to a specific VIP (Virtual failover IP) you will want the following in /etc/sysctl.conf:

net.ipv4.ip_nonlocal_bind=1

Related Solutions

Iptables – Linux equivalent of pfsync+carp for failover firewall/routing

I've read carp+pfsync is a good solution. Is that currently what most of you are using?

Yes and yes :)

As Instye notes, there are two public projects for CARP under Linux. But as you'll notice neither of them are particularly active and don't believe they include pfsync. Which is pretty important to the whole shebang.

Additionally there have been some huge advances in the PF and CARP code this year alone. Any port, including FreeBSD, often lags naturally behind in feature and bug fixes.

If the current machine isn't performing any other tasks then I'd recommend just biting the bullet and implementing OpenBSD. The learning curve won't be any steeper than getting one of the ports up and running. I don't think you'll regret it.

HAproxy to web host sub directory

You can use a regex search and replace it with a string, like so --

reqrep  <search> <string>
reqirep <search> <string>   (ignore case)
  Replace a regular expression with a string in an HTTP request line
  May be used in sections :   defaults | frontend | listen | backend
                                 no    |    yes   |   yes  |   yes
  Arguments :
    <search>  is the regular expression applied to HTTP headers and to the
              request line. This is an extended regular expression. Parenthesis
              grouping is supported and no preliminary backslash is required.
              Any space or known delimiter must be escaped using a backslash
              ('\'). The pattern applies to a full line at a time. The "reqrep"
              keyword strictly matches case while "reqirep" ignores case.

    <string>  is the complete line to be added. Any space or known delimiter
              must be escaped using a backslash ('\'). References to matched
              pattern groups are possible using the common \N form, with N
              being a single digit between 0 and 9. Please refer to section
              6 about HTTP header manipulation for more information.

Best Answer

Related Solutions

Iptables – Linux equivalent of pfsync+carp for failover firewall/routing

HAproxy to web host sub directory

Related Topic