Starwind considerations with Exchange and SQL

jmchristy · Fri Apr 12, 2013 5:28 pm

I'm looking at moving my Exchange 2010 SP1 server into a virtual environment, I already have my SQL 2000 server virtualized but it's running on local storage currently.

What are some best practices or things to consider when moving either of these onto the Starwind SAN?

I'm currently running a 2-node setup, with HA. I have write back caching enabled on my current devices, and I believe this is a safe route to go since it one node fails the other switches to write through caching. I have had situations where both my nodes would go offline due to an extended unplanned power outage, so I'm worried about using write back caching with either Exchange or SQL since they have databases on the backend that are pretty sensitive when the 1's and 0's don't match.

Any feedback is greatly appreciated!

Mon Apr 15, 2013 8:52 am

Hello and welcome on board

If you know that you can have unplanned power outages that is the point where you should consider equipping your SAN boxes with UPSes.

I think it will be great if you could take a closer look at our Best Practices documentation:
http://www.starwindsoftware.com/starwin ... ces-manual

Please do not hesitate to contact us should you have any additional questions.

jmchristy · Mon Apr 15, 2013 1:07 pm

Thanks!

We certainly have UPSes setup with our SAN's, however it only gives us about an hour of run time. There are times where we lose power, or have planned outages that last more than a day. Appreciate the best practices guide, but I have that pretty much figured out.

My question is in regards to running Exchange or SQL on a device with write back caching enabled. Is this a recommended best practice with 2 node HA setup or not? I know SQL is sensitive to ANY kind of data disruption but what about Exchange 2010?

Worse case scenarios would be, power goes out in the middle of the night, UPSes run until they drain entirely and the software that is SUPPOSED to shut down the SAN's gracefully doesn't. Likelyhood isn't very high, but obviously I'd rather have 0% chance of lost data then 5% chance.

Wed Apr 17, 2013 2:12 pm

Well, if power outages is the thing that is happening pretty often then I think it will be better to have the script that would turned off your client machines and/or applications, and one of the StarWind nodes on the moment when the power is off but the UPSes are still up and after this script should turn off second StarWind node.

Otherwise I think it will be better to go with the WT caching mode or No Caching at all.

jmchristy · Wed Apr 17, 2013 6:31 pm

Thanks for the response!

It's certainly not that power outages happen "often". We may have one extended outage lasting almost an hour once a year, weather it be due to weather or upgrades to our electrical system. I don't know if that's considered often or not but if so then it's often!

We will proceed with no caching.

Wed Apr 17, 2013 6:37 pm

I'd suggest to use 3-way replication and stick with a huge UPS modules. There's virtually no way to have all the 3 nodes down @ the same time with such a conditions met.

jmchristy wrote:Thanks for the response!

It's certainly not that power outages happen "often". We may have one extended outage lasting almost an hour once a year, weather it be due to weather or upgrades to our electrical system. I don't know if that's considered often or not but if so then it's often!

We will proceed with no caching.