May 13, 2010 at 3:57 pm
I have been able to successfully instal the Oracle 10.2.0.4 client on a non-produciton standanalone machine, but looking for information on the install for the production server. The production server is Windows 2008 Sql 2008 64 bit active passive cluster. The OS is on its own physical disks and shared disk is san attached. I am unsure how or what configurations to set up to complete the install. Do I install the client on C: with the OS for each node and set the Oracle folder to c:, or is a shared service required through the cluster administrator and install oracle binaries to shared disk? Any suggestions are appreciated.
August 13, 2010 at 1:07 pm
I'm doing the same thing and have not been able to find any information either. Have you found anything since you posted here?
I have and active-active setup. Active1 is a SQL2k5 node and Active2 is a SQL2k8 node. Both are 64-bit and need to have linked servers to Oracle databases.
Thanks
-Mario
August 16, 2010 at 8:25 am
It will depend on the version of the OS. For Windows 2008 64 bit I used "10204_vista_w2k8_x64_production_client". I installed this on physical Disk with the Window OS on both nodes. Since the System databases for sql server are on shared disk I only had to set up the linked server 1 time, but the oracle binaries are on each node with values in the path environment variable, so 2 installs here.
August 16, 2010 at 10:59 am
(I know this is pretty long but I figured it would be better that way to try to save someone else from lot of googling and headaches)
I played with this a little bit over the weekend and here's the info of what's happened so far:
- Active Active Cluster
- Both servers are (pnode1 and pnode2) Windows Server 2008 R2 Enterprise x64
- For pnode1 we have C: on local disk, D: (data), L: (log) and T: (tempdb) on the san
- For pnode2 we have C: on local disk, E: (data), M: (log) and U: (tempdb) on the san
- On pnode1: installed (and clustered) SQL Server 2005 x64 Enterprise (this is a default instance, but after all the installs, I now wish I had made it a named instance) on pnode1 and out of that we have the clustered virtual node vnode1.
- On pnode1: installed SQL Server 2005 SP2 on vnode1
- Moved files around, including system, so MDFs on D:, LDFs on L and all tempdb files to T:
- On pnode2: installed (and clustered) SQL Server 2008 R2 Enterprise (see note 1 bellow as to why R2) with a named instance "sql2k8" and during installation, specified MDFs on E:, LDFs on M: and tempdb's MDF + LDF files on U:
Now, to the Oracle stuff:
1) RDPed on to vnode1
2) installed Oracle client from 11107_w2k8_x64 DVD on C:\Oracle (see notes 2 and 3 bellow) with custom install to install only: sqlplus, Oracle MTS for windows, ODBC + OLEDB + .Net oracle provider and Oracle Net
3) offlined vnode1
4) restarted pnode1
5) moved vnode1 back to pnode1
6) onlined vnode1
7) with SSMS, logged on to vnode1 and set "allow inprocess" for the Oracle OLEDB Provider
8) created linked server to oracle database
9) select * from ORADB..SYS.DUAL; works fine
10) repeated 1 through 9 on pnode2/vnode2
11) verified vnode 1 resides in pnode1 and vnode2 on pnode2
12) in vnode1: select * from ORADB..SYS.DUAL; works
13) in vnode2: select * from ORADB..SYS.DUAL; works
14) move vnode1 to pnode2
15) in vnode1: select * from ORADB..SYS.DUAL; works
15) in vnode2: select * from ORADB..SYS.DUAL; works
16) move vnode2 to pnode1
17) in vnode1: select * from ORADB..SYS.DUAL; works
18) in vnode2: select * from ORADB..SYS.DUAL; works
19) move vnode1 to pnode1
20) in vnode1: select * from ORADB..SYS.DUAL; works
21) in vnode2: select * from ORADB..SYS.DUAL; works
22) move vnode2 to pnode2
23) in vnode1: select * from ORADB..SYS.DUAL: works
24) in vnode2: select * from ORADB..SYS.DUAL; works
so at this point, I can query oracle through the linked server from both sql instances from both physical nodes, regardless of which physical node a particular sql instances happens to be on.
But here's where the problems began:
25) pull the power chord from pnode2
26) vnode2 failed over to pnode1 and came on line just fine
27) ssms to vnode1 and can still query oracle
28) ssms to vnode2 and can still query oracle
29) power up pnode2, but keep vnode2 on pnode1 (node is set to not failback)
30) ssms to vnode1 and can still query oracle
31) ssms to vnode2 and can still query oracle
32) move vnode2 to pnode2
33) ssms to vnode1 and can still query oracle
34) ssms to vnode2 and CANNOT query oracle, error message: Msg 7302, Level 16, State 1, Line 1 Cannot create an instance of OLE DB provider "OraOLEDB.Oracle" for linked server "ORADB".
35) move vnode2 to pnode1
36) ssms to vnode1 and can still query oracle
37) ssms to vnode2 and can query oracle
38) move vnode2 to pnode2
39) ssms to vnode1 and can still query oracle
40) ssms to vnode2 and CANNOT query oracle
41) move vnode1 to pnode2
42) ssms to vnode1 and CANNOT query oracle
43) ssms to vnode2 and CANNOT query oracle
44) move vnode1 and vnode2 to pnode1
45) can query oracle from both vnodes
I tried all sorts of things, from tweaking the provider to recreating the linked server, etc. Here's what finally "resolved" the problem it and there have been no errors thus far, with tests on moving the virtual nodes around and also pulling the power chord as done above:
1) SSMS to vnode1
2) drop oracle linked server
3) SSMS to vnode2
4) drop oracle linked server
5) RDP to pnode1
6) offline vnode1
7) RDP to pnode2
8) offline vnode2
9) restart pnode1 and pnode2
10) RDP to pnode1
10) make sure vnode1 is on pnode1 and vnode2 on pnode2 and both still offline
12) install oracle client, custom install so that only the OLEDB and .Net providers are installed, so do not install: sqlplus, MTS, Oracle Net
13) RDP to pnode2
14) repeat 12
15) online vnode1 on pnode1
16) online vnode2 on pnode2
17) ssms to vnode1
18) set "allow inprocess" for oracle provider
19) create linked server
20) repeat 17, 18 and 19 on vnode2
21) repeat all "select * from ORADB..SYS.DUAL" tests above including moving the nodes around AND pulling the power chords and everything is still working.
Maybe it was a glitch the first time around. The only difference is that the first time I installed the Oracle client, I did it by RDPing to the virtual nodes (and uninstalled from the virtual nodes too). The Second time around, I installed by RDPing to the physical nodes AND installed only the 2 providers (OLEDB and .Net), but nothing else.
I'd be interested in knowing about others experiences with this sort of setup.
Note 1: We first tried to install 2k5 on pnode1 and 2k8 (non-R2) on pnode2, but had problems installing 2k8. After looking around, I found the slipstream thing to install 2k8 sp1 stuff and then the 2k8 instance, but that still did not work and screwed up the cluster enough to the point we couldn't install or uninstall anything else without getting all sorts for errors. So the cluster was rebuilt and we installed 2k5 on pnode1 and 2k8r2 on pnode2 without any errors during the install.
Note 2: I don't know the exact file name of the 11G client that was downloaded from Oracle because the Oracle DBA that downloaded it is no longer with us. However I do know that as of this writing *(8/16/2010) it is the most up to date Oracle 11G x64 client. We use the 11g client even though we're on Oracle 10G because the 10G x64 client (it has "Vista" on its name) downloaded at the time gave us a bunch of errors when I tried to install it on another win2k8r2EE with sql2k5x64EE but the 11Gx64 did not
Note 3: When installing the Oracle 11Gx64 client, you'll get an error about the installer expecting OS version 6.0 but finding 6.1. Just click on the checkbox that says "user verified" or "manually verified" or something like that (I did not write down the exact wording) for both that check and the next one (which does not run because the first one did not pass) to enable the Next button and install the client.
Viewing 4 posts - 1 through 3 (of 3 total)
You must be logged in to reply to this topic. Login to reply