Extension Mobility SRV Record Failover

[rchaseling]

Hi,

We are currently testing Extension Mobility failover on our cluster. We have DNS A records setup correctly & when we shut down the Publisher it takes a good bit of time before DNS realises it is down and thus there is a large time where users cannot login to EM.

On reading this doc - http://www.cisco.com/en/US/products/sw/voicesw/ps556/products_tech_note09186a008097cd93.shtml it says:

DNS A Record

An alternative to Cisco IOS SLB, DNS can be used to provide redundancy       for these components. In order to use DNS as a redundancy mechanism, a DNS A       record should be configured for each Cisco Unified Communications Manager       subscriber node with the same fully qualified domain name (FQDN) or host name.       This is necessary so that a DNS query for the FQDN or host name returns IP       addresses for multiple Cisco Unified Communications Manager subscriber nodes.       Thus, a phone that queries the DNS server for the FQDN or host name receives       multiple IP addresses. Given a node failure, these IP addresses can be       contacted one at a time in turn until a responding node is found. The drawback       to the use of DNS for EM redundancy is that if a node fails, a phone must wait       for a timeout before it attempts to contact the next IP address or node. This       timeout can take anywhere from 30 to 45 seconds and is repeated for each       attempt to contact a failed node. Therefore, significant delays can occur       during node failures before a response is received from a Cisco Unified       Communications Manager node.

DNS Service (SRV) records can also be used to provide       redundancy.

Its this last line we are looking at as we believe this will failover faster but we cannot get it working. Has anyone tried this or got this working??

Our SRV record is correct(?)  _http._tcp.domain.com. 86400 IN SRV 0 0 8080 callmanager.domain.com but we are unsure what the EM URL should now look like?

Has anyone got this wokrking before??

REgards

Russell

[gogasca]

Hi Rusell,

http://www.cisco.com/en/US/docs/voice_ip_comm/cucm/srnd/7x/cmapps.html#wp1043795

"Cisco  does not recommend a redundancy design using  DNS records with multiple  IP listings. With multiple IP addresses  returned to a DNS request, the  phones must wait for a timeout period  before trying the next IP  address in the list, and in most cases this  results in unacceptable  delays to the end user. In addition, this can  result in more than two  subscriber nodes with the Cisco EM Application  service enabled to  handle login/logout requests, which is not supported"

Currently IP Phones will take 60+ Seconds to fail over to the secondary server if the DNS response included one in the event the
first one is not responsive. The failover delay is due to the TCP connection retry attempts and unfortunately the parameters that
controls this behavior on the phones can not be adjusted at this time.
During the 60+ second timeout if the enduser hits Cancel SoftKey the request will immediately failover to the secondary server in the DNS
response.

Unfortunately this behavior reassures the recommendation of using Cisco Application Control Engine (ACE) or the Cisco IOS SLB feature.

HTH

PS the documented you referenced was last updated Apr 16, 2008 when we havent documented this behavior.

Thanks

[rchaseling]

Hi,

Thanks for that info and appreciate that Cisco does not recommend DNS as a failover & I'm well aware of its limitations but the client is not looking to spend any more money & does not have a 6500 to use IOS SLB. They are just after the best option & DNS will at least provide some sort of redundancy if their Publisher goes down and I'm thinking that SRV records would be better than just A records but can't seem to get it to work & believe it is the URL we are trying.

If we can't get SRV records working we'll just stick with straight A record DNS - just thought I'd ask anyway.

Thanks

Russell