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 |
|
pareshmotiwala
Constraint Violating Yak Guru
323 Posts |
 Posted - 2007-03-16 : 16:07:30
|
| I have just started this new thread.We have a Standard Edition of SQL 2005 and windows 2003.Active/Passive Cluster.The failover time is the default of 180 seconds. My boss is not ok with it. He feels the whole exercise is a waste of time.Tara suggested that they have an active/active/active/active cluster going in production.I learn Active Active clustering is not supported by SQL 2005 SE is it true? and also if at all it was true in some version, it would require serious code re-writing? or that the application will have to be "Cluster Aware"I have the SQL 2005 Administration Book, which suggests the above.Again, my apologies for hijacking an earlier thread by Sharmag. |
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2007-03-16 : 16:13:16
|
| We use SQL Server 2005 Enterprise Edition on a 4 node active/active/active/active cluster. We have 11 instances installed on that cluster.Yes active active clustering is supported in SQL Server 2005. It requires no code changes. The application does not need to be cluster aware. I don't have the book that you got this information from, so please type into this thread exactly what it says. Perhaps you are mis-reading it.I don't understand what your boss has a problem with. Our cluster groups failover in about 30 seconds. What exercise is he talking about?Tara Kizerhttp://weblogs.sqlteam.com/tarad/ |
 |
|
|
mcrowley
Aged Yak Warrior
771 Posts |
Posted - 2007-03-16 : 16:14:14
|
| I believe SQL Server 2005 Standard only supports clusters of two physical nodes. In order to go up to 4 (or 8 for that matter), you need to get SQL Server Enterprise Edition. Applications that connect to the database would not have to be rewritten, unless they would need some sort of re-connection logic for when there is a failover. When an active-active cluster fails over, the IP address of the SQL Server is assumed by the physical node that is picking up the load. In short, the clients do not care which physical server has the SQL Server running on it, the IP address is available, and they can all connect to that.Database Mirroring does not have a mechanism to assume an IP address, so that does require a code change to most connection strings. This change I think is only available with the SQLNCLI provider. |
|
|
|
vladimir.stokic
Starting Member
25 Posts |
Posted - 2007-03-23 : 06:33:40
|
| Hi there!Pareshmotiwala, I believe that the time of 180 seconds is a consequence of having a passive node in the cluster. However, if your applications uses only one database, you cannot have an Active/Active configuration, because (to my knowledge) two instances of SQL Server cannot work on the same database (please correct me if I am wrong). So, the failover time is as it is due to the fact that the passive node needs to be brought up to active mode. In Tara's case, she already has active nodes (running several instances of SQL Server each), so they do not need to be brought up, and that saves time. They just need to take over the databases that the failed node owned.That having been said, I would like to ask if there is any way to reduce the failover time in an Active/Passive cluster. Again, the assumtion is that the application uses just one database.Regarding the database mirroring, I was informed that the failover is transparent to the application in that case too, because MSSQL Native client is the one to change the IP of the Principal Server on the client side. Please, correct me if I am wrong.Also, I know that this is MSSQL forum, but I was just curious if anybody knew the usual failover times of some other DBServers (e.g. Oracle or Sybase)?Thanks. |
|
|
|
|
|
|
|
|
Active Active Clustering