We were wondering recently how to achieve carrier grade service with a VoIP application - we find out, that apart from the fact of having a hot backup , we needed to be able to transparently switch over the hot backup without disrupting our end users service.
The solution came out as follow :
- 2x servers with an exact copy of data (mysql multi master replication)
- DNS settings with a TTL of 600 secondes
- a monitoring agent at the DNS level, checking Asterisk Manager (port 5038) every 3 minutes
After implementation, failover was working great, and sip user agents were able to failover to the standby Asterisk server after failure. Conclusion: disruption time is low and service continuity can be assured.
Some DNS providers charge hundred of dollars to achieve this kind of deployment, but it’s possible to find affordable ones giving you failover and TTL control.
March 28, 2008 at 4:25 pm
This is a very interesting post. I have setup this configuration and I’m in the process of testing it more thoroughly. My greatest concern is that some devices/applications may not honor the TTL, or maybe their DNS is provided by a ISP that does not strictly honor TTLs.
Have you seen any problems related to this?
March 28, 2008 at 4:34 pm
Hi Dave
As long as it doesnt fall below 300, most ISPs should keep the original TTL . SIP devices should also take into account the provided TTL. I know that Mediatrix gateways accept the provided TTL with no problem.