We have a Primary unity server 5.0 in our NJ Datacenter. The Exchange Mailstore is OffBox in the NJ Data center (local) the secondary Unity Server is in the PA Datacenter connected via WAN 45MB/DS3
We have set up Failover
Failover advanced Options/ DISABLED automatic failover and Failback
There is a new check box "Enable File replication"
the idea is if the NJ Datacenter Fails, the User can acccess the PA Unity Server (failsafe mode/no excahnge message store )
Is this a supported config?
Failover for VM requires 3 servers. Refer the following document for configuring failover
Unity Failover only requires 2 Unity servers, 3 servers if you count the customers Exchange server. You only need a license for failover. I set this up and if the Primary server failed, it would automatically failover to the Secondary. When the Primary comes back on line, I had it reconnect to it, you can leave it unchecked and do it manually though. Both Unity servers need to be in the same AD Domain.
Couple of things here:
1. DS3 is not supported speed, take a look at these requirements:
Requirements for Separating Cisco Unity Servers by a WAN
â¢Both Cisco Unity servers must be connected with customer-managed, dedicated leased-line facilities. Carrier-managed bandwidth solutions such as an IP cloud hand off are not supported.
â¢There must be two paths of connectivity between the Cisco Unity servers.
â¢Each path of connectivity between the Cisco Unity servers must be Optical Carrier level OC-24 (1.244Gbps) or greater.
â¢The Cisco Unity servers must not be separated by a firewall.
â¢The distance between the Cisco Unity servers must be no more than 125 miles (200 kilometers).
â¢The maximum round-trip latency must be no more than 10 ms.
â¢Each location must have a redundant domain controller and global catalog servers.
â¢The Exchange environment must be customer owned.
â¢Call control must be available at each location.
2. With Unity 5.0 you do not need seperate exchnage message store, it can be installed on the failover server now.
I don;t understand why DS3 is not supported, since this is a Hot standby situation. the secondary server will not be live unless the primary Server is unavailable.
So, while the 2 servers are connected by DS3,
What Data is constantly being replicated? I actually have this scenario working currently in a LAB. I wonder what the caveats are
Data is being replicated in SQL. If the SQL replication has a hiccup, then the two servers come out of sync and major problems could occur. The Specs mention about Optical and the mileage must be new, I have not seen those specs before.
If you can sometimes talk to your local Cisco SE to see if they will buy off (meaning support) the DS3 configuration, etc. Sometimes if it's a direct Point to Point with no other traffic on that link, then they will support it.
I have had Unity Pairs out of sync before. (i made changes to the secondary while it was active *accident)
But is there a danger of the primary database getting corrupted? if we only make changes to the Primary, is it true that the only corruption that would occur would be at the secondary?
(thanks for your insight)
Well, that is essentially the problem. If the failover server starts up the services (when set to automatic) and then there are network issues with latency, or dropped packets, etc the SQL databases will come out of sync. When they do finally come back online and try to sync, issues will occur with either the failover or the primary.
It's a finicky setup with the primary/failover setup. That is why the rules are so strict about splitting up the unity servers.
I have split unitys before through DSC3 and it works fine. But you always run the risk if they database get corrupt, you could be in for some long restores. I have even had to reload the failover server completely (on a local network) because it was too far gone.
I just did a design where I have about 5-10k of Unity subscribers. Instead of loading them up all on one server with failover and splitting the failover, I actually created (2) nodes with failover in separate datacenters. If I do have a datacenter down, at least only half of the subscribers will be down, instead of all. Im not sure I can I rely on the failover to do it's job across the WAN (even though it is a direct link, and mileage, etc)
If it is a requirement, I would talk to your local Cisco SE about the failover scenarios you are required to support at your company. Sometimes the business objects are mis-understood and it turns out the failover really does not have to be split. (or something like that)
In regards of the Primary getting corrupt, it is possible that if it's failing back and forth in short time spans, the database could have slight corruption on MWI triggers, services being hung in Windows (from starting/stopping, etc) TSPs bouncing back and forth on CUCM, etc)
hope this helps.
The documentation on this is confusing. I thought Unity 5.0 allowed failover across the WAN using a 100Mbps link. It's called Unity warm standby and FO must be manually initiated.