Oracle Grid Infrastructure 12.1.0.2 on Windows caveats

Recently I needed to set up a two-node RAC cluster on Windows 2012 at a customers site. Therefore we picked the latest and greatest available version for that, which is 12.1.0.2 as of now. Since this is not my first cluster on Windows I took my checklist and went through all the prerequisites, renamed network interfaces, disabled automatic DNS registration, changed the bind order, created partitions, things like that.

So I continued to run through all those steps in the installer and finally the prerequisites check told me that the so called “Automatic Metric” value for the public and interconnect network is the same while it should not. This Automatic Metric feature was introduced with Windows 2012 and does a kind of automatic ordering of network interfaces depending on their speed. And since our public network was only 1GBit in contrast to the interconnect network having 4GBit it was put behind the interconnect network. That’s why the documentation (http://docs.oracle.com/database/121/CWWIN/networks.htm#CIHJFCGC) tells us to disable this feature and do the ordering manually which we then did.

Now that all prerequisites checks were successful the installer started to do all the work. Until the creation of the Container Database for the Oracle Grid Infrastructure Management Repository…. You know, that’s this new management database which was optional in 12.1.0.1 and became mandatory with 12.1.0.2. That means there is no way around this database. A quick research in MOS and the web gave no useful hints. That’s why I started to walk through the logfiles. Finally I came to the dbca log for the Management Database at “D:\app\oracle\cfgtoollogs\dbca\_mgmtdb\trace.log”:

[CRSNative.genericStartResource:313]  Failed to start resource: Name: ora.mgmtdb, node: myracnode01, filter: null, msg CRS-5020: Not all endpoints are registered for listener MGMTLSNR
CRS-2674: Start of 'ora.MGMTLSNR' on 'myracnode01' failed
[Thread-24] [ 2015-02-27 10:36:10.649 CET ] [HADatabaseUtils.start:1244]  Error starting mgmt database in local node, PRCR-1013 : Failed to start resource ora.mgmtdb
PRCR-1064 : Failed to start resource ora.mgmtdb on node myracnode01
CRS-5020: Not all endpoints are registered for listener MGMTLSNR
CRS-2674: Start of 'ora.MGMTLSNR' on 'myracnode01' failed

Ok, the management listener has some issues, it can’t register all it’s endpoints? Why is that?

C:\Windows\system32> crsctl stat res -t
--------------------------------------------------------------------------------
Name           Target  State        Server                   State details
--------------------------------------------------------------------------------
Local Resources
--------------------------------------------------------------------------------
ora.ASMNET1LSNR_ASM.lsnr
               ONLINE  ONLINE       myracnode01              STABLE
               ONLINE  ONLINE       myracnode02              STABLE
ora.GIMGMT.dg
               ONLINE  ONLINE       myracnode01              STABLE
               ONLINE  ONLINE       myracnode02              STABLE
ora.LISTENER.lsnr
               ONLINE  ONLINE       myracnode01              STABLE
               ONLINE  ONLINE       myracnode02              STABLE
ora.net1.network
               ONLINE  ONLINE       myracnode01              STABLE
               ONLINE  ONLINE       myracnode02              STABLE
ora.ons
               ONLINE  ONLINE       myracnode01              STABLE
               ONLINE  ONLINE       myracnode02              STABLE
--------------------------------------------------------------------------------
Cluster Resources
--------------------------------------------------------------------------------
ora.LISTENER_SCAN1.lsnr
      1        ONLINE  ONLINE       myracnode02              STABLE
ora.LISTENER_SCAN2.lsnr
      1        ONLINE  ONLINE       myracnode01              STABLE
ora.LISTENER_SCAN3.lsnr
      1        ONLINE  ONLINE       myracnode01              STABLE
ora.MGMTLSNR
      1        ONLINE  INTERMEDIATE myracnode01              Not All Endpoints Re
                                                             gistered 192.168.50.
                                                             33 192.168.50.34,STA
                                                             BLE
ora.asm
      1        ONLINE  ONLINE       myracnode01              Started,STABLE
      2        ONLINE  ONLINE       myracnode02              Started,STABLE
      3        OFFLINE OFFLINE                               STABLE
ora.cvu
      1        ONLINE  ONLINE       myracnode01              STABLE
ora.oc4j
      1        ONLINE  ONLINE       myracnode01              STABLE
ora.scan1.vip
      1        ONLINE  ONLINE       myracnode02              STABLE
ora.scan2.vip
      1        ONLINE  ONLINE       myracnode01              STABLE
ora.scan3.vip
      1        ONLINE  ONLINE       myracnode01              STABLE
ora.myracnode01.vip
      1        ONLINE  ONLINE       myracnode01              STABLE
ora.myracnode02.vip
      1        ONLINE  ONLINE       myracnode02              STABLE
--------------------------------------------------------------------------------

Let’s look into the listener configuration:

C:\Windows\system32> srvctl config mgmtlsnr
Name: MGMTLSNR
Type: Management Listener
Owner: nt authority\system
Home:
End points: TCP:1521
Management listener is enabled.
Management listener is individually enabled on nodes:
Management listener is individually disabled on nodes:

C:\Windows\system32> srvctl config listener
Name: LISTENER
Type: Database Listener
Network: 1, Owner: nt authority\system
Home:
End points: TCP:1521
Listener is enabled.
Listener is individually enabled on nodes:
Listener is individually disabled on nodes:

Look at that, both listeners use the same TCP port. That is not the case on Linux. On Linux the management listener uses port 1531. Should it be that easy? I tried it:

C:\Windows\system32> srvctl modify mgmtlsnr -endpoints TCP:1531
C:\Windows\system32> srvctl stop mgmtlsnr
C:\Windows\system32> srvctl start mgmtlsnr
C:\Windows\system32> srvctl config mgmtlsnr
Name: MGMTLSNR
Type: Management Listener
Owner: nt authority\system
Home:
End points: TCP:1531
Management listener is enabled.
Management listener is individually enabled on nodes:
Management listener is individually disabled on nodes:

Now the management listener is running on port 1531. A quick look at “crsctl stat res -t” showed no more issues:

--------------------------------------------------------------------------------
Cluster Resources
--------------------------------------------------------------------------------
ora.MGMTLSNR
      1        ONLINE  ONLINE       myracnode01              192.168.50.35 192.16
                                                             8.50.36,STABLE

Now I re-run the assistant and it went past that step without any errors. Easy fix, but not quite easy to find the cause of the issue.

Finally the installer came up with lots of green check marks:
installer-successful

I hope this helps all of you who need to cope with Oracle clusters on Windows.

 

Advertisements