Please start any new threads on our new
site at https://forums.sqlteam.com. We've got lots of great SQL Server
experts to answer whatever question you can come up with.
| Author |
Topic |
|
bogey
Posting Yak Master
166 Posts |
 Posted - 2005-08-30 : 10:27:30
|
| Our current server sqlprod\sqlprod_01 is our production machine. Its running an instance of sqlserver called sqlprod_01. I'd like to setup a failover machine. Since all applications reference the "instance" does this mean that my failover should have the same instance name? Currently the way failover works is you unplug the production machine and rename the failover to whatever the prod was. This means that the failover must have machinename\same instance name as production.Am I correct in this assumtion? |
|
|
eyechart
Master Smack Fu Yak Hacker
3575 Posts |
Posted - 2005-08-30 : 10:50:29
|
| this type of failover gets tricky becuase SQL Server doesn't like it when you rename the host after SQL Server has been installed. This can be done (I think) but I believe it gets kinda messy.I just re-read your post, are you running with both machines hot on the same network, or is your failover machine cold until it is needed? If it is not online, then you can of course name the box the same as the production server. If you have two identical boxes though, you might want to consider using MSCS clustering instead of the manual method you described. It would basically handle the scenario you discussed automatically. There is definitely some additional costs involved with clustering though. You can also look into using Veritas clustering, which behave similarly to MSCS clustering. It is actually cheaper to implement because it does not require that you run enterprise edition SQL Server or Advanced/Enterprise edition of windows server like MSCS clustering does. -ec |
 |
|
|
bogey
Posting Yak Master
166 Posts |
Posted - 2005-08-30 : 11:02:14
|
| Cost is not an issue. When you say cold, what do you mean? Both machines are on the same network. The failover would have the same instance name just a different machine name. |
|
|
|
eyechart
Master Smack Fu Yak Hacker
3575 Posts |
Posted - 2005-08-30 : 11:08:31
|
quote:
Originally posted by bogey
Cost is not an issue. When you say cold, what do you mean? Both machines are on the same network. The failover would have the same instance name just a different machine name.
when I said cold I meant powered off. When you connect to an instance you connect to server\instance, so the machine names will have to be made the same at some point so you don't have to go and alter all the clients connection strings or DSNs.If cost is not an issue then look into setting up a cluster. The MSCS cluster is the easier, better documented and supported approach. You will need to use Enterprise edition SQL and advanced server win2k/win2k3. You will also need a shared storage subsystem, like an HP MSA500 or MSA1000. If physical hardware fails, you just failover to the other node and everything is automatic. Your approach of renaming the system and performing a manual failover gets messy and is difficult to do especially in a critical situation where everyone is waiting on you to get things back up and running. You also might run into issues with routers and ARP caches because the MAC address of your host has changed. MSCS gets around all that becuase it virtualizes the MAC address, so you don't have to flush ARP caches.EDIT:MSCS performs a gratiutous ARP request that updates the switches/routers/servers ARP tables when a failover happens. It doesn't virtualize the MAC address as I stated earlier. The end result is that clients should be able to connect to the correct node when a failover occurs.-ec |
|
|
|
bogey
Posting Yak Master
166 Posts |
Posted - 2005-08-30 : 11:15:07
|
| Thanks eyechart - The machine is HOT and the production nor standby database is NOT running on windows 2000/2003 advanced server. |
|
|
|
|
|
|
|
|
Disaster Recovery with an Instance??