Monday 20 July 2015

DELETE A NODE

NOTE:- This method not removing any physical file, Its work only unpin and update inventory.
11g R2 RAC has extremely simplified the process of deleing a node to the cluster as compared to 10g RAC or 11g R1 RAC.  To know the  procedure to delete a node in 10g RAC or 11g R1 RAC pls click here.
Now, after the introduction of SCAN and GPNP, we need to follow very simple steps.

Current scenario:
I have three nodes in the cluster presently.

Host names :
- host01.example.com
- host02.example.com
- host03.example.com

Node to be deleted : 
- host03.example.com


Example Note:
-Unpin Node

Run the following command as root to determine whether the node you want to delete is active and whether it is pinned.


[root@host01 ~]# $GRID_HOME/bin/olsnodes -s -t
host01        Active  Unpinned
host02        Active  Unpinned
host03        Active  Unpinned

If the node being removed is already unpinned then you do not need to run the crsctl unpin css command below and can proceed to the next step.

If the node being removed is pinned with a fixed node number, then run the crsctl unpin css command as root from a node that is to remain a member of the Oracle RAC in order to unpin the node and expire the CSS lease on the node you are deleting.


[root@racnode1 ~]# $GRID_HOME/bin/crsctl unpin css -n host03
CRS-4667: Node racnode3 successfully unpinned.
If Cluster Synchronization Services (CSS) is not running on the node you are deleting, then the crsctl unpin css command in this step fails.



Procedure:

– Login as root on the node(s) to be deleted(host03) navigate to  <gridhome>/crs/install 
[root@host03]#cd /u01/app/11.2.0/grid/crs/install


– Disable oracle clusterware applications and applications  on the node (host03) 
[root@host03]#./rootcrs.pl -delete -force


– From a node that will remain(host01) , delete the node
root@host01]#crsctl delete node -n host03

– on a node that will remain , login as grid user, navigate to gridhome/oui/bin and update node list

[grid@host01]$ cd /u01/app/11.2.0/grid/oui/bin

[grid@host0$] ./runInstaller -updateNodeList ORACLE_HOME=/u01/app/11.2.0/grid “CLUSTER_NODES={host01,host02}”
  


– check that only nodes 1 and 2 remain
$ crsctl stat res –t


Reference:
      Bug 16788764 : 12.1SH0506: UNABLE TO RUN ADDNODE OF AN INACTIVE/PINNED NODE

      Bug 16971369 : ON OCCASIONS NODE DELETE FOLLOWING GRID DECONFIG CAUSES NODE TO BE PINNED

No comments: