[mgarber]

Currently our firm has one BES server in our NY office. We want to run a concurrent server in our Chicago office.

We need help figuring out how best to deploy the new server. We would like to have the ability to fail over if one server goes down. If we use one SQL database, what if the database server goes down? We would like to ensure our environment has complete failover for 100% availability.

Any help would be awesome.

Thanks.

[mahoward]

IIRC, SQL database replication is not yet fully supported. DR for SQL is a restore from backup operation. You can wait for 5.0 to support HA or spend big bucks for a solid solution like Neverfail for both the BES & SQL servers.

[mgarber]

Thanks for the reply. Let's put aside the chance of an SQL failure for now. If we went to 1 server in each office (two total servers) to load balance, if one failed, how hard is it to move the users from the failed server to the functional server? Would it be better just to cluster the servers for failover in one location?

[hdawg]

You can actually do SQL failover ... Search the BTSC Docs for disaster recovery.

What mail platform, version & service pack are you running?
what bes version & service pack are you running?

[mgarber]

Exchange 2003, ver 6.5.7638.1
BES 4.1.4.15 which I believe is SP4

[mahoward]

If the 2 BES servers are connected to the same SQL db then you can simply move the accounts from the downed server to the active server and the BB's will start working automagically.

[mgarber]

Can you clarify how you would move users off of a server that is down? I just want to make sure that its possible if the actual server blows up to move users from that server to a different server.

Thanks for all of your help.

[mahoward]

The BES Manager console connects to the SQL db primarily. So you are telling the SQL server to move the accounts to the available server. The source server does not need to be online for this to happen.

[mgarber]

Awesome.

Where do the licenses sit? If I have 50 users on each server, do I need 100 licenses on each server to accomodate failover?

[hdawg]

The licenses sit in the database. All BES servers connected to the same database (BlackBerry Domain), share the same pool of CALs.

If you have 50 users on each server (and you've got 2 servers), you need 100 licenses in the database to cover them all.

[raffip]

We have implemented a 100% available BES infrastructure. We have two BES servers connected to a clustered SQL backend. We have been able to perform upgrades to all systems without ever bringing down BES.

[hdawg]

Wirelessly posted (banana)

Originally Posted by raffip

We have implemented a 100% available BES infrastructure. We have two BES servers connected to a clustered SQL backend. We have been able to perform upgrades to all systems without ever bringing down BES.

100 uptime is not possible unless the primary server and all of its dependencies are available at all times. You have a highly available solution, but what happens if a BES fails? Do you automatically move users over? Do you have a spare server configured identically that automatically spins up? Is the data in the databased replicated or are you using shared storage? Lots of questions / details you left out of this 100% available solution.

[x14]

Originally Posted by hdawg

Wirelessly posted (banana)
100 uptime is not possible unless the primary server and all of its dependencies are available at all times.

I agree.

Eventually you have to take down the system to software upgrades. This requires downtime unless you move the users to another server which doesn't count towards system uptime.

[hdawg]

Originally Posted by x14

I agree.

Eventually you have to take down the system to software upgrades. This requires downtime unless you move the users to another server which doesn't count towards system uptime.

... and even if you move them, there is still a period when they are without full functionality when they receive new service books.

[raffip]

I apologize if i used the term 100% availablity loosely. I just meant that we can usually bring down any component of the infrastructure and make it seamless to the user. Yes, you will still have small delays of a few minutes at the most but we worked with RIM extensively when planning our Exchange 2007 environment and this was the best we could put together and it does the trick.

When necessary we move users back and forth between the two BES servers and perform monthly system maintenance, better known as Windows Security updates. Same with the SQL cluster, we fail over and back as needed. The back end data is on a Netapp, and even that has two heads for failover. Its an extremely expensive solution when you look at it, but when you are given a blank check and told "I never want BES services to be down", we spare no expense. And if you realy want more details, these Netapps also provide storage for our Exchange 2007 CCR clusters. All of this is located in our Co-location. With Exchange 2007 CCR clusters, we are able to fail Exchange services from either of two locations to our co-lo and keep email services up.

I guess I was a bit vague in my first post. Sorry, new to the community. Just trying to put some ideas out there.

[mahoward]

It is difficult to move users between servers for monthly maintenance when you number ~600 users per BES server and above, so not an option for us. Moves are also not 100% foolproof, I have seen flaky issues with it before in DR scenarios. We are moving to VMWare to outsource DR and HA from being the responsibility of the application layer, instead of waiting for Argon or paying $$$$ for 3rd party failover solutions.

[hdawg]

^ my favorite approach to date ... as long as you have the appropriate people to support it.

[x14]

We're buying NetApp for something else we're working on so I'm hoping to try running BES in VMware/NetApp.

[Legros]

Hi. We are presently on ESX 3.5. Pretty happy with the result.. at least it gives us a solution for the hardware failure. But for the software part, all solution seems to ask manual intervention... Anyone heard or try the neverfail solution or any other third party offering similar solution?
Neverfail - BlackBerry Availability

Thanks