Skip to main content

High Availability

About

High availability FreeSWITCH™ configurations including failover and central databases.

Click here to expand Table of Contents

Error rendering macro 'toc'

null

Introduction

At ClueCon 2010 the FreeSWITCH team announced the call recovery feature that's now available in FreeSWITCH.

This feature means you can restart FreeSWITCH after a failure, and keep existing calls up with only a few seconds of audio dropped.

In the case of HA clustered FreeSWITCH instances, you can automatically fail calls from box A to box B within a few seconds of a failure being detected. (typically 3-5 seconds)

Configuration Requirements

Track Calls

In the SIP profiles you must have track-calls enabled.

So if your call has 2 legs, you must enable track-calls on both profiles that the call uses.

This is done simply by using this config snip in your SIP profiles.

<param name="track-calls" value="true"/>

Be aware there is a minor performance hit using "track-calls" as call state (all session vars, similar to the mod_xml_cdr data) is stored in the core DB every time the channel state changes.

Failover to a Second Server

If you want to recover calls a second server, then you must have some way of sharing the call recovery data.

ODBC or PostgreSQL are possible methods, and to do so, you must specify a DSN for all involved sip profiles (like you needed track-calls on all involved profiles).

<param name="odbc-dsn" value="odbc://dsn:username:password"/>

And also in switch.conf.xml like:

<param name="core-db-dsn" value="odbc://dsn:username:password"/>

If you want to clone everything, you will also need to add DSN information to:

voicemail.conf.xml (also you need to rsync/drbd/NAS share the actual voicemail files)

The hostname on both machines needs to be the same, as the sql query to recover calls selects by hostname. To do this, simply set the following parameter in switch.conf.xml on both FS instances to the same value:

<param name="switchname" value="freeswitch01"/>

Instruct FreeSWITCH to recover calls

To instruct FreeSWITCH to recover from a failure all you need to do is execute:

fsctl recover

To crash test the feature, you may execute:

fsctl crash

To have FreeSWITCH to automatically recover calls on startup:

<X-PRE-PROCESS cmd="set" data="api_on_startup=fsctl recover"/>

HA configuration

Floating IP method on Debian Squeeze

I've tested this on debian squeeze, it works like a charm. :)

Please note that 10.10.10.11 should be a public facing IP, it is only used as an example here.

Install arping

apt-get install arping

Allow FreeSWITCH to bind to nonlocal IP

Add the following line to /etc/sysctl.conf

echo 'net.ipv4.ip_nonlocal_bind=1' >> /etc/sysctl.conf

Restart networking:

/etc/init.d/networking restart

Run:

sysctl -p

You should see:

net.ipv4.ip_nonlocal_bind = 1

If you don't, you did something wrong, if you do, continue...

Add an ethernet alias for the floating IP

Add the following to /etc/network/interfaces on both primary and secondary FreeSWITCH servers:

(substituting 10.10.10.11 and 255.255.255.0 with your floating IP config)

iface eth0:0 inet static address 10.10.10.11 netmask 255.255.255.0

Start primary FreeSWITCH server

Bring up the floating IP

ifup eth0:0

Make sure that the SIP profiles are set to listen on your floating IP

<param name="rtp-ip" value="10.10.10.11"/> <param name="sip-ip" value="10.10.10.11"/> <param name="presence-hosts" value="10.10.10.11"/> <param name="ext-rtp-ip" value="10.10.10.11"/> <param name="ext-sip-ip" value="10.10.10.11"/>

Start FreeSWITCH

/etc/init.d/freeswitch start

Start secondary FreeSWITCH server

Make sure the floating IP is not loaded.

ifdown eth0:0

Make sure that the SIP profiles are set to listen on your floating IP

<param name="rtp-ip" value="10.10.10.11"/> <param name="sip-ip" value="10.10.10.11"/> <param name="presence-hosts" value="10.10.10.11"/> <param name="ext-rtp-ip" value="10.10.10.11"/> <param name="ext-sip-ip" value="10.10.10.11"/>

Start FreeSWITCH

/etc/init.d/freeswitch start

TEST!

Make a call and don't hang up.

Crash the primary server

fs_cli -x 'fsctl crash'

Bring down the floating IP

ifdown eth0:0

Recover the calls on the secondary server

Bring up the floating IP

ifup eth0:0

Run arping command to update the router

arping -q -c 3 -A -I eth0:0 10.10.10.11

Recover the calls.

fs_cli -x 'sofia recover'

Voilà, the call should continue with only 3-5 seconds of audio dropped!!

To Do

  • TODO: Someone please fill this in as you play with this setup.
  • TODO: Heartbeat example Check the git-contrib in ledr/ha.d/ for some possibly good stuff.
  • TODO: UCarp config examples
  • TODO: Pacemaker config example

It is possible to bind FreeSWITCH to a floating IP that is not yet assigned to the box. These means FreeSWITCH can be in a ready & waiting state when the IP moves over, making the switchover much faster.

Normally, the OS will refuse to allow FreeSWITCH to listen on an IP that is not assigned. Sofia will get an error and the profile will not start.

On Linux you work around this using:

sysctl to set net.ipv4.ip_nonlocal_bind = 1 or echo 1 > /proc/sys/net/ipv4/ip_nonlocal_bind ubuntu: sysctl net.ipv4.ip_nonlocal_bind=1

These can be set on boot, either from a rc.d script or a /etc/sysctl.d/* file.

This means the OS will allow FreeSWITCH to use that IP even though it isn't yet assigned to the machine. It won't receive traffic or respond to ARP requests until the machine takes control of that IP, but requires FS to be already started so that it'll start handling calls as soon as it fails over.

See Also

Freeswitch HA Keepalived

Enterprise deployment IP Failover

PDF - http://opensips.org/pub/events/2016-05-10%5FOpenSIPS-Summit%5FAmsterdam/Giovanni%5FMaruzzelli-OpenSIPS%5FSummit2016-FreeSWITCH%5FHA.pdf

SmartVox Discussion of Pacemaker with OpenSIPS