Redundant / avoiding downtime


I am looking for some insight into avoiding downtime while patching a linux server (with asteriks and stun running on it). The distribution will be Redhat and the server will be hosted virtual.

What isn’t fully clear to me is how you can avoid downtime while patching one system. Afaik with windows servers you can cluster them up but i am not sure how this works with a linux server.

I can imagine that you need something in the range of the following:
-Second linux server
-Second dns record to point to the other server?
-Something to check if the application and server are running
-Personal details need to be stored central / in a database? Right now it happens the “regular” file (sip.conf afaik).

What i want to avoid is an environment which is redundant but does fails to detect any patch errors / program not working scenario (for example checking on ip connectivity only).

As you might have guessed my experience with this subject is basically nothing, but i hope someone can poin me in the right direction.

I looked around a bit and something like … proach/en/ might be viable options. Anyone got some input on this?

Best regards,

Having run windows server clusters in the past, they are remarkably similar both in concept and implementation but linux patching is a lot more service friendly that windows in that patches usually do not need to bring machines down in the first place. That said, I have dozens of systems running asterisk on linux high availability clusters that user dbd+heartbeat.

you can start here: … +Solutions

Well this point has come up again and i think it is better to ask for some more help so thats why i am making another post. Thanks for the link mudslide, i have looked at some of the options and am looking towards 2 things. First i’ll describe my situation and then the (hopefully) solutions i am looking at.

My situation:
-Have a linux (RHEL servers) with stun/asterisk/some home-made software running on it
-1 Dynamic file (sip.conf) > can be copied with a script though has to be checked for changed both ways
-1 incoming ip to the linux server (public on the firewall, internally a private ip)

It seems to be possible with Asterisk+Hearbeat+SIP+Multi-homed on Debian/Ubuntu (except that i use RHEL) or would i be better of using Flip1405 ( ?

The last one (Flip) seems very easy to setup.


Firstly how often do you expect to be downing the server ?

As to redundancy we supply systems based on centos / Heartbeat / DRBD and mysql replication.

Its fairly straight forward to setup and works well.

For our public systems we use DNS addresses and not ips to register to and have a short TTL on them, Then we have a UK server and a US server as Backup. We also have a different DNS address for teh US server as well so that phones that support backup servers such as the AAstra range have both addresses.



I never thought of the DNS option so that might be even easier. I guess you want to know the downtime because during it’s downtime the connectivity will be slower (because of TTL’s on the DNS part)?

How long / much delay would that give?