New HA setup advice from standalone AWS hosted install

[s2svoip]

Perhaps utilizing AWS Load Balancer along with AWS RDS (PostgreSQL) or Aurora could work which would take care of all the failover and replications needed. Haven't done it, but certainly looks like a path to explore.

View attachment 1308

Yeah for sure! that is very interesting. I like how this thread has evolved with these suggestions, making use of these parts of AWS, I am sure there is similar stuff in GCP and Azure, could be an interesting route

I only have a basic handle on AWS, trying to figure out peering across regions but with the suggestion of EFS and this, haha thats a lot of research to be done

 

[MTR]

@Adrian Fretwell Which endpoints are these? They need naming and shaming. I also want to know to avoid them or work around it.

i can tell you that Polycom and yealink works great with the TTL

Grandstream, cisco , Fanvil, aquavox (does 2 we have many doorbells) will ignore the TTL they cache it forever or until they cant reach the server... so if server1 i take it offline for an upgrade or any other reason it will jump once he notices that the the server is offline to server2 then when i switch back to server1 it will not go back it can stay like this for months will either need to reboot the device or block the ip address in iptables (then i will need to remmber to unblock it)

 

[MTR]

i can tell you that Polycom and yealink works great with the TTL

Grandstream, cisco , Fanvil, aquavox (does 2 we have many doorbells) will ignore the TTL they cache it forever or until they cant reach the server... so if server1 i take it offline for an upgrade or any other reason it will jump once he notices that the the server is offline to server2 then when i switch back to server1 it will not go back it can stay like this for months will either need to reboot the device or block the ip address in iptables (then i will need to remmber to unblock it)

i think that DNS failover is not something that is the best option, even with plycom and yealink there is still a 2 3 minutes down time.

in my experience opensips and kamaillio need much less maintenance or has downtime , i am thinking of moving over the registrations and the presence to opensips or kamaillio and the dispatcher module can do the fail-over

i would gain

1. we can handle on each fusionpbx much more end-users (registrations)
2. if we failover the server as long as the opensips and kamaillio are up the end user might not even notice that we are doing something

would love to hear your feedback on this ...

right now i am using opensips only for the DID routing and failover like this all my carrier send the the DID to my opensips and kamaillio and i have only 1 point of management,

Mark stop teaching this in the advanced class he doesn't like this anymore if anyone still has and video or documentation for the old classes please PM me

 

[DigitalDaz]

I'm happy with the 2 to 3 minutes simply because in reality this is something that strikes me no more frequently than once every couple of years. 95% of my customers do not even notice it.

Moreover, with yealinks now you can register to two servers simultaneously which makes outbound calling seamless, even with a failover.

 

[MTR]

I'm happy with the 2 to 3 minutes simply because in reality this is something that strikes me no more frequently than once every couple of years. 95% of my customers do not even notice it.

Moreover, with yealinks now you can register to two servers simultaneously which makes outbound calling seamless, even with a failover.

1. What's the max registration you do per server? Mark told me to stick to 500 and this means 500 per 2 servers?