What you need is different than has been offered by the existing HA capability. With the existing capability, call drops weren’t prevented - calls drop in the event of server death, but new calls, via the second server, can be established quickly. Live call fail-over is a whole other matter.
You’re not going to get where you want to go without spending a sh*t-ton of money. In order for this to work, you’ll have to have duplicate data paths through disparate servers running on different hardware platforms through disassociated network infrastructures.
You will need the following:
- A duplicate server running your call up to the point that it connects to the phone,
- A way to redirect the phone to a new server seamlessly.
- An infrastructure provider that can dual route all of your calls to multiple servers in different locations.
- A way, at the phone, to dedupe packets as they are processed through the network, especially as servers add and drop.
I know a system like this. It has about 200 nodes. The total cost of the system was well over $2B. The system is over 2M lines of assembly language code. The call paths are managed through dedicated communication trunks that are managed centrally with no other purpose but managing the messaging network. There is a move afoot to replace it - the price tag on the replace is about $20B so far.
Thanks for the input. Good to know!
If you are using VMware for you pbx and have multiple hosts, the the VMware HA will do very much what the FreePBX 13 HA did, albeit much easier to setup and far less hassle.
Vmware Fault Tolerant (FT) would in theory allow another machine to take over without dropping calls etc, but HA is good enough for us so not tried FT.
You cannot get around calls dropping. They wont exist on the secondary machine. Asterisk has norhing to clone those sessions between two boxes. Things are going to drop.
You would get around calls dropping With VMware Fault Tolerance, this will run a clone/shadow copy of your pbx on a second host. Should the primary host/pbx fail the shadow/clone immediately takes over without any applications etc effected eg calls in progress.
An expensive solution just for a pbx system but if you have the infrastructure in place already then cheap and easy.
We only need HA, which is much easier with VMware than the HA on FreePBX 13 which we found far more trouble than it was worth. It does work and works well on occasions, but can be a real headache at times too.
VMware also saves the cost having to have two of every commercial License for FreePBX especially the HA one which works out $3000 + $500 a year.
It will be interesting to see what they come up with for HA on SNG7. I’m sure a lot of people will find it very useful.
Please explain how the clone machine will run Asterisk during all this, which the same IP details, as the primary VM. Also as to how it will know what channels and transactional information is being cloned.
Finally, you have tested this and proven it because in your original comment about VMFT you expressed “in theory” this should work. So does it or doesn’t it?
Best to read some of the VMware documents about FT. But the whole point of FT is that the clone/shadow machine is kept fully in sync with the live machine, that is disk, memory, transactions etc all in real time. To quote one document “there is zero downtime, zero data loss, zero connection loss, and continuous service”.
Whilst we don’t run our FreePBX under FT only HA, we have run other critical services under FT and it works great, the user is unaware the service has failed over to another server, and in our case another machine room some distance away. I see no reason that Asterisk would be any different.
However its not something you would want to deploy just for a phone system I suspect, it’s not a cheap solution, you need very good network connectivity between the servers ie 10Gb.
So then the answer is no. This will not work as you are expecting or theorizing it will work. This will not replicate the needed things in Asterisk to automagically move calls over.