Wanpipe Linux Case Studies

From david - 3/27/07 6:14 AM

Hello,

I wanted to thank Sangoma for all the great support and their quality products.  I wanted to explain instructions for a technique that I use to increase Asterisk system stability for my clients.  My technique is essential to use a two tiered Asterisk system.

THIS IS WORK IN PROGRESS, AND NOT ALL OF THE INSTRUCTIONS ARE TESTED. PLEASE CHECK FOR UPDATES!
Thanks,
David


This technique involves using two asterisk processes. One of the asterisk processes serves as the Zaptel gateway and forwards all requests to the second asterisk server via SIP. This increases system stability because we can safely reset one of the switches without affecting the call leg on the other system.

If the asterisk server becomes unstable for any reason (due to a deadlock in queue, voicemail, chan_local, etc), the switch can reset without interrupting the first asterisk process, which only job is to forward calls from Zaptel to SIP. This can solve problems associated with high call volume on a box. I call this technique 'Call virtualization'.

Followed are the detailed instructions

1.      Install asterisk, Zaptel, and the Sangoma drivers normally.  Make sure that you rebuild Zaptel after installing the Sangoma drivers because there are important patches made to Zaptel code.
2.      Install a second asterisk server in the /opt/ directory.  For verbosity we will call this /opt/asterisk-tier2. Open up Makefile and find the following line, modify it to this directory:
INSTALL_PREFIX?=/opt/astiersk-tier2
3.      Build asterisk. I always build asterisk with don’t-optimize so I can get debug information incase of a crash

# make don’t-optimize
# make install
4.      Add a user called asterisk2
5.      Change the owner of /opt/asterisk-tier2 and the source build directory to ‘asterisk2’
6.      su (as in super user) to username asterisk2. In the build directory, run ‘samples’.  It is important to do this as a different user so that you do not accidentally overwrite your config in /etc/asterisk
7.      You should now have a file called /op/asterisk-tier2/etc/asterisk/asterisk.conf
8.      Modify the file /opt/asterisk-tier2/etc/asterisk/sip.conf and change the port number to port 25060.  This is an example port number and you can use your own.
9.      Run the following command to run asterisk. Run the command as user asterisk2. Note that you must use full path names because of a bug in Asterisk.

/opt/asterisk-tier2/usr/sbin/asterisk -C /opt/asterisk-tier2/etc/asterisk/asterisk.conf  -vdgc

10.     Verify the above runs successfully.
11.     Now, modify your main asterisk process to forward all of your calls
Essentially, you want this type of dial plan:

[default]
Exten => _X.,1,Dial(SIP/Localhost:25060/${EXTEN})

soon I will update this comment with a PART II, that uses this principle to increase system stability even further. Thank you, David Elbel

Part II, Creating using Call Virtualization with a failover server
Using Call Virtualization as a make-shift load balancer.

From afaiz - 12/7/05 11:23 PM

We just completed the installation of an Asterisk-based system in the Malaysian Prime Minister's Department. The system consists of dual servers (on hot standby)  using the AFT104 cards. The 104 handles two incoming E1 lines from the telco, and another two E1 lines going out to an Avaya PABX.

 The server sits in front of the Avaya and provides call routing between the existing Avaya extensions and a mix of VoIP phones (deskphones, WiFi phones and several software-based operator consoles). The system also provides a range of services such as voicemail, faxmail and click-to-callback, and is designed as a Unified Messaging System that ties in with the client's existing resources and infrastructure.

The system has been in production since late August 2005, and since then we have had no service interruptions caused by the Sangoma cards. They perform beautifully, and virtually no echo has been observed. Integration with Asterisk is seamless!

The Sangoma team was extremely helpful and diligent in ironing out some minor teething problems. Most of our issues were resolved by upgrading the card's firmware and Wanpipe drivers at the client site, and did not take very long to complete.

Ahmad Faiz
Chief Technology Officer
Lanvik ICU Sdn Bhd

From masonc - 12/27/05 2:34 PM

I have two Adtran Channel Banks, a 600 and a 750, connected via a t1 conneciton to a Sangoma A104 providing 48 channels of FXO to an Asterisk server. I had problems correctly configuring the timing of the channel banks beacuase there is little documentation about the correct way to do this, but with some help from Sangoma I configured them as follows:

Each port of the A104 is configured as Master Clock

 /etc/wanpipe/wanpipe1.conf

TE_CLOCK        = MASTER
TE_REF_CLOCK    = 0

 /etc/wanpipe/wanpipe2.conf

TE_CLOCK        = MASTER
TE_REF_CLOCK    = 0

 Each channel bank is set to Network or loop timing, depending on what that model calls it. This allows the channel banks to take the timing from the same source and eliminates timing errors.

The Sangoma support staff have been wonderful about supporting our customizations and driver configuration.