How to make failover work

[meneerB]

Hi, We use CUCM6.1.2,
1 publiser, 1 subscriber, 150+ phones (7941 and 7961) + extension mobility on all phones.
publisher 10.0.0.11
subscriber 10.0.0.21

I have checked what happens if one callmanager failes, and how it's setup now, it is ready for improvement.
How would you make this setup (more)failsave?

1. TFTP for booting a phone
in our w2k3 server, scope option 150 was set to the ip of the publisher. If that one failes after a powerfailure, non off the phone are able to boot...
I put a second ip in, the ip of the subscriber.
After testing: phones can boot from both IPs
So that fine now?

2. Extension mobility
Device -> Device settings -> Phone Services -> Extention Mobility.
This url connects to

http://10.0.0.21/emapp/EMAppServlet?device=#DEVICENAME#

So if the subscriber fails, no phone is able to login or logout.
I red an artical of using DNS:
Creating the same hostname for both IP's, roundrobin 0 secs and putting the hostname in the url instead of the IP.
But, roundrobin is good for this: if an ip is not pingable: the phone doesn't know this and still tries that ip address. So failing to get a url back, it cannot login.

(stupid if I may so, is this:
DNS syncs itself to other DNS servers within the same domain
DHCP is not able to do that. It's like to try if an ip is avalailable, if not 'okay, you may use that address'.
But DNS is at the other hand not able to do such a check.
grrr, always something stupid;-(
)

3. URL's
System -> Enterprise Parameters -> Phone URL Parameters
All entries are directed to the ipaddress of the publisher.
So if the publisher fails? (a bit the same as 2)

so, what are you using in your envirnment to keep it working?

 

[MauricioDB]

1.the pub and sub should have one ip each...there are in as a whole undr cisco unidifed cm group configuration...thats the failover..phones wil work if one goes down....the last thing u should be worried about if one server goes down is adding new phones to a broekn cluster! keep one server as tftp
my opnioin
2.just create a alias to the ip address of the sub or pub your using... users logged in will stay logged in. roundrobin is auto in dns and will alwyas try one then the other every time..so it may take the second try to get in..hoepfully ...but again you would be owrkign on a solution to fix the server not so much trying to get peolple logged on or off..hoepfully it doenst go down during shifts
3.u can use dns name to ip address for this...

all my stuff points to ip addresses have hundreds and have phones ..i have had one server go down but phones failed over like they should according to the cm group and people kept working ..with a slight glich then rebooted the locked up server it came back up and phones moved back everything went a-ok..i had a bug in version 7.0 im on 7.1 now so its better

 

[meneerB]

1. if you use dhcp with scope option 150, then i cannot see this working. if the server is down, how does the phone know where to get its config from...?
please give the exact menu.
2. i have to try that. the second time is when?
and no, there is no way i get the server working again on time. every minute with 150 complaining users. what do you think;-)
it just has to work, no matter what happens.
3. please give the exact menu.