I am currently doing a POC for a geographically dispersed highly available server farm. I want to run my purposed setup by this community to see if anyone has tried this or if anyone can poke holes in my solution. I have 2 independent ARR/Controller Servers
with 1 Web Farm on each. However Each of the ARR/Controller Servers has a different primary server, but all the servers in the farms are the same. I plan to have 1 ARR, it's Primary and 1 other server in 1 data Center and 1 ARR, it's Primary and another Server
in another Data Center. The ARR servers go back to an LTM(F5) and the LTM's go back to a GTM(F5). Since WFF doesn't support Backup Primary Servers I setup this so when applications are deployed they will be deployed to both Primary servers and WFF will sort
out the provisioning. This has worked in very small volume, but I wanted to see what others think of the setup.
When you say "ARR server" do you actually mean a control server? Obviously you only need ARR when you want to load balance and I assume you are doing that via the F5's.So a control server with a single primary and a single secondary?
If so, your set up seems fine, the only thing I would say is maybe go for more than one secondary server. Web farms are good but from my experience WFF is certainly not the finished article yet and when they fail they are a PITA to fix and can be out of
contention for days at a time.
At least with another secondary you have some extra resilience.
Yes when I say The "ARR servers" I mean the 2 Control Servers. I plan to have it setup like this: 2 Data Centers; in Data Center 1 I will have Control Server 1, Primary 1, Secondary 1. In Data Center 2 I will have Control 2, Primary 2, Secondary 2. I plan to
have independent farms setup on the 2 controller servers, however all 4 web front end servers from both data centers will be in both farms, the difference being that the primary will be different so Primary 2 is a Secondary server in the Controller 1 farm,
and Primary 1 is a secondary in Controller 2 farm. This gives me high availability for both the applications and provisioning of applications. In small volumes I haven't seen an contention that has caused a malfunction, I'm just wondering what could happen
over time or if a lot of updates happen between provisioning intervals. WFF is not exactly designed for what I want to do, so I know I'm pushing the limits a little. I wanted to see if anyone else has tried this or at the very least would consider it. I wouldn't
have to do this if we were allowed to delegate a backup primary, so if the primary went down it would automatically roll over to the backup. This doesn't sound like a hard feature to implement, but they haven't yet. Thanks for your assistance.
Your secondary servers have to be able to access the web on port 80 else you have to do a work around.
We dont use the load balancer for WFF as the software seems so fragile that we would probably REALLY be pushing our luck. Nor do we currently use it for Platform provisioning, but may in the future.
The warm up feature is not practical, so we dont use that either.
The fail-over is difficult to write as the log file will indicate that sync is shut off when it really isnt.
With all of that said, yes, we have setup our farm in some areas to have the primary for one farm, be the secondary for the second. It appears to work find so far.
The way I think of WFF, which might help you, is that the sync engine itself was apparently written by geniuses of the first degree, however, the wrapper around it was not written by the same folks. (censored who I suspect wrote the wrapper.)
rose_jw
6 Posts
Highly Available Setup: Need Advice
May 03, 2012 02:09 PM|LINK
nastymatt
124 Posts
Re: Highly Available Setup: Need Advice
May 09, 2012 08:34 AM|LINK
When you say "ARR server" do you actually mean a control server? Obviously you only need ARR when you want to load balance and I assume you are doing that via the F5's.So a control server with a single primary and a single secondary?
If so, your set up seems fine, the only thing I would say is maybe go for more than one secondary server. Web farms are good but from my experience WFF is certainly not the finished article yet and when they fail they are a PITA to fix and can be out of contention for days at a time.
At least with another secondary you have some extra resilience.
rose_jw
6 Posts
Re: Highly Available Setup: Need Advice
May 09, 2012 01:02 PM|LINK
talldaniel2
7 Posts
Re: Highly Available Setup: Need Advice
May 14, 2012 06:04 PM|LINK
Your secondary servers have to be able to access the web on port 80 else you have to do a work around.
We dont use the load balancer for WFF as the software seems so fragile that we would probably REALLY be pushing our luck. Nor do we currently use it for Platform provisioning, but may in the future.
The warm up feature is not practical, so we dont use that either.
The fail-over is difficult to write as the log file will indicate that sync is shut off when it really isnt.
With all of that said, yes, we have setup our farm in some areas to have the primary for one farm, be the secondary for the second. It appears to work find so far.
The way I think of WFF, which might help you, is that the sync engine itself was apparently written by geniuses of the first degree, however, the wrapper around it was not written by the same folks. (censored who I suspect wrote the wrapper.)