August 28, 2013 at 3:13 am
dokledon (8/28/2013)
I say again
Erm, no.whether stand alone or clustered you can only have one default instance, it has nothing to do with separate groups but the registry keys which are replicated across all nodes and validated at install time.
For StandAlone, no doubt.
For Cluster, this exaclty what I wanted to check. Opinions seem to be divided, but I start to believe it is really impossible.
So for my scenario, can I use Named Instances + 1433 + Alias for all my SQL Cluster Instances ?
ClusterGroup1:
SQL1/Inst1 ; IP A ; 1433 ; owner Node 1
ClusterGroup2:
SQL2/Inst2 ; IP B ; 1433 ; owner Node 1
ClusterGroup3:
SQL3/Inst3 ; IP C ; 1433 ; owner Node 2
ClusterGroup4:
SQL4/Inst4 ; IP D ; 1433 ; owner Node 2
Thanks,
Dok
No
again:
"the port has to be unique. But you should be able to just connect via Network Name, and have the SQL Server Browser Service have the port resolved for you - and redirect your Client to it automatically."
ist all one Cluster, so nothing can be repeated: neither instance Name, nor port - no matter on which node they are acive at a given time.
unless you really want to mix up clustered instances with non-clustered instances (!)
Andreas
---------------------------------------------------
MVP SQL Server
Microsoft Certified Master SQL Server 2008
Microsoft Certified Solutions Master Data Platform, SQL Server 2012
www.insidesql.org/blogs/andreaswolter
www.andreas-wolter.com
August 28, 2013 at 3:15 am
dokledon (8/28/2013)
For StandAlone, no doubt.
For stand alone or cluster, no doubt!
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
August 28, 2013 at 3:25 am
Andreas Wolter-332291 (8/28/2013)
Noagain:
"the port has to be unique. But you should be able to just connect via Network Name, and have the SQL Server Browser Service have the port resolved for you - and redirect your Client to it automatically."
ist all one Cluster, so nothing can be repeated: neither instance Name, nor port - no matter on which node they are acive at a given time.
unless you really want to mix up clustered instances with non-clustered instances (!)
That's not correct. With clustered installations of SQL you can use the same portnumber!! Because this portnumber is used in combination with the IP-address and every cluster group has it's own unique IP-address.
I've implemented this many times during my previous employment.
August 28, 2013 at 3:36 am
HanShi (8/28/2013)
Andreas Wolter-332291 (8/28/2013)
Noagain:
"the port has to be unique. But you should be able to just connect via Network Name, and have the SQL Server Browser Service have the port resolved for you - and redirect your Client to it automatically."
ist all one Cluster, so nothing can be repeated: neither instance Name, nor port - no matter on which node they are acive at a given time.
unless you really want to mix up clustered instances with non-clustered instances (!)
That's not correct. With clustered installations of SQL you can use the same portnumber!! Because this portnumber is used in combination with the IP-address and every cluster group has it's own unique IP-address.
I've implemented this many times during my previous employment.
You are right.
I should not write quick replies before breakfast.
Yes, per Group it has to be unique.
The WFC does not matter
Andreas
---------------------------------------------------
MVP SQL Server
Microsoft Certified Master SQL Server 2008
Microsoft Certified Solutions Master Data Platform, SQL Server 2012
www.insidesql.org/blogs/andreaswolter
www.andreas-wolter.com
August 28, 2013 at 3:40 am
HanShi (8/28/2013)
With clustered installations of SQL you can use the same portnumber!! Because this portnumber is used in combination with the IP-address and every cluster group has it's own unique IP-address.
You can indeed use the same port number with a different IP, you cannot use the same instance name even in a cluster scenario.
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
August 28, 2013 at 3:56 am
Perry Whittle (8/28/2013)
...., you cannot use the same instance name even in a cluster scenario.
I've read more documentation and I stand corrected. You are absolutely right! Sorry for the misleading earlier posts.
Quote MS documentation (http://technet.microsoft.com/en-us/library/ms179410(v=sql.105).aspx):
SQL Server depends on distinct registry keys and service names within the failover cluster to ensure that SQL Server functionality continues after a failover. Therefore, the name you provide for the instance of SQL Server, including the default instance, must be unique across all nodes in the failover cluster. Using unique instance names ensures that instances of SQL Server that are configured to fail over to a single server have distinct registry keys and service names.
August 28, 2013 at 3:56 am
Andreas Wolter-332291 (8/28/2013)
HanShi (8/28/2013)
Andreas Wolter-332291 (8/28/2013)
Noagain:
"the port has to be unique. But you should be able to just connect via Network Name, and have the SQL Server Browser Service have the port resolved for you - and redirect your Client to it automatically."
ist all one Cluster, so nothing can be repeated: neither instance Name, nor port - no matter on which node they are acive at a given time.
unless you really want to mix up clustered instances with non-clustered instances (!)
That's not correct. With clustered installations of SQL you can use the same portnumber!! Because this portnumber is used in combination with the IP-address and every cluster group has it's own unique IP-address.
I've implemented this many times during my previous employment.
You are right.
I should not write quick replies before breakfast.
Yes, per Group it has to be unique.
The WFC does not matter
But:
Do you happen to have a Whitepaper/KB where Microsoft explicitely states that this configuration is supported.
I mean, we are talking about HA. And in the end it will be the same NIC/MAC that has to resolve the forwarding to the same virtual IP.
And I have not read anything about the technique behind and how stable it really is.
I'd be grateful for a link - otherwise i will certainly still stay away from such configurations..
Andreas
---------------------------------------------------
MVP SQL Server
Microsoft Certified Master SQL Server 2008
Microsoft Certified Solutions Master Data Platform, SQL Server 2012
www.insidesql.org/blogs/andreaswolter
www.andreas-wolter.com
August 28, 2013 at 11:24 am
For clarification
I have a 3 node cluster comprising Node1, Node2 and Node3.
I create a clustered default instance on Node2 and Node3. I then attempt to install an instance (clustered or standalone) on Node1 and receive the error attached below
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
August 28, 2013 at 12:31 pm
Perry Whittle (8/28/2013)
For clarificationI have a 3 node cluster comprising Node1, Node2 and Node3.
I create a clustered default instance on Node2 and Node3. I then attempt to install an instance (clustered or standalone) on Node1 and receive the error attached below
I never even tried that for obvious reasons :w00t:, but I thought I had about people hacking their Systems to get such done.. :hehe:
Andreas
---------------------------------------------------
MVP SQL Server
Microsoft Certified Master SQL Server 2008
Microsoft Certified Solutions Master Data Platform, SQL Server 2012
www.insidesql.org/blogs/andreaswolter
www.andreas-wolter.com
August 29, 2013 at 2:06 am
I have never hacked SQL server to attempt this not sure i'd even want to. Not even sure this would actually be possible due to the registry hive structure.
When following the SQL Server installation rules the screenshot below is the error you would expect to see
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
August 29, 2013 at 2:46 am
Perry Whittle (8/29/2013)
I have never hacked SQL server to attempt this not sure i'd even want to. Not even sure this would actually be possible due to the registry hive structure.When following the SQL Server installation rules the screenshot below is the error you would expect to see
You have no idea what's possible, if you just really want to.
How about this one?:
A customer managed to install the SQL Server Agent seperately - on a different box - than SQL Server. Because he thought this is how a "3 Tier System" should be designed like. SQL Server worked.. Kind of.. SQL Agent not so much...
:w00t:
error Messages? Who cares.. just Keep trying until they go away...:sick:
Andreas
---------------------------------------------------
MVP SQL Server
Microsoft Certified Master SQL Server 2008
Microsoft Certified Solutions Master Data Platform, SQL Server 2012
www.insidesql.org/blogs/andreaswolter
www.andreas-wolter.com
August 29, 2013 at 3:53 am
I think there is no more doubt about Default instances in a Cluster.
1) What about Static port number ? Can we use default 1433 for all instances ?
2) How connect to the Clustered Named Instance just using the SQL Cluster name (ie Name: SQL1\Inst1 ; connect with SQL1) ?
August 29, 2013 at 4:01 am
dokledon (8/29/2013)
I think there is no more doubt about Default instances in a Cluster.1) What about Static port number ? Can we use default 1433 for all instances ?
Yes, that is possible.
dokledon (8/29/2013)
2) How connect to the Clustered Named Instance just using the SQL Cluster name (ie Name: SQL1\Inst1 ; connect with SQL1) ?
If you want to connect without using the instancename, you need to connect using the IP-address (or servername) and portnumber. The ip-address/servername is NOT from the hostserver/cluster, but from the virtual SQL server.
August 29, 2013 at 6:43 pm
If you want to connect without using the instancename, you need to connect using the IP-address (or servername) and portnumber. The ip-address/servername is NOT from the hostserver/cluster, but from the virtual SQL server.
Thank you HanShi.
And to connect just using Virtual Server Name ?
How could we do that we Named Instances ?
Dok.
August 30, 2013 at 5:54 am
dokledon (8/29/2013)
If you want to connect without using the instancename, you need to connect using the IP-address (or servername) and portnumber. The ip-address/servername is NOT from the hostserver/cluster, but from the virtual SQL server.
Thank you HanShi.
And to connect just using Virtual Server Name ?
How could we do that we Named Instances ?
Dok.
Like I said in the quoted post: connect using the Virtual Servername and the portnumber: SQL1,1433 (depending on the type of the connector you may have to replace the comma to a colon or semicolon)
Viewing 15 posts - 16 through 30 (of 30 total)
You must be logged in to reply to this topic. Login to reply