Blog

12c addnode procedure with ASM

Share this article

A quick note and work-around for a ASM problem I encountered while adding a RAC node to our 12c labo setup:

The environment in question was a new 2 linux (OL 6.6) node 12.1.0.2 RAC (on OVM 3.2.2) with the April 2015 PSU.
ASM is used to provide the shared storage.

When running the “cluvfy stage -pre nodeadd” check before actually adding the node,  a warning was given on the ASM check:

Checking Oracle Cluster Voting Disk configuration...
ERROR:
PRVF-5449 : Check of Voting Disk location "ORCL:DG_GRID_ASM01" failed on the following nodes:
demo-rac12-03,demo-rac12-01
demo-rac12-03:No such file or directory
demo-rac12-01:No such file or directory
ERROR:
PRVF-5449 : Check of Voting Disk location "ORCL:DG_GRID_ASM02" failed on the following nodes:
demo-rac12-03,demo-rac12-01
demo-rac12-03:No such file or directory
demo-rac12-01:No such file or directory
ERROR:
PRVF-5449 : Check of Voting Disk location "ORCL:DG_GRID_ASM03" failed on the following nodes:
demo-rac12-03,demo-rac12-01
demo-rac12-03:No such file or directory
demo-rac12-01:No such file or directory
PRVF-5431 : Oracle Cluster Voting Disk configuration check failed

I ignored this warning as it failed as well on the reference node and blamed it on not having the the cvuqdisk rpm installed (which is required for the cvu shared device check).
But when running the actual addnode.sh script, I got the same failure and addnode.sh would not continue.

Checking on MOS I found note 1989622.1 – addnode.sh failed due to “PRVF-5449 : Check Of Voting Disk Location Failed” for ORCL and AFD device, which describes this as a bug on 12c and offers a patch for it.
Unfortunately, the provided patch conflicts with the April 2015 PSU patch.
But I also found another note, describing the same issue (1945572.1), which states that you can work around the problem by changing the asm_diskstring parameter from “ORCL:*” to the actual location of the ASM devices (“/dev/oracleasm/disks/*”) and restart the cluster.

Note 1945572.1 is for 11.2, but after making the changes on my 12c setup the verification succeeded

[oracle@demo-rac12-01 addnode]$ cluvfy comp vdisk -n demo-rac12-01,demo-rac12-02,demo-rac12-03 -verbose
Verifying Voting Disk
Checking Oracle Cluster Voting Disk configuration...
WARNING:
PRVF-4038 : CRS is not installed on nodes:
demo-rac12-03
Verification will proceed with nodes:
demo-rac12-01,demo-rac12-02
Oracle Cluster Voting Disk configuration check passed
UDev attributes check for Voting Disk locations started...
Result: UDev attributes check passed for Voting Disk locations
Verification of Voting Disk was successful.

And I could execute the addnode.sh script, after which I changed the asm_diskstring back to the “ORCL:*” value.

In the note, they use a strange 2 step approach to set the parameter (first use asmcmd to change the parameter in the GPnP profile and then use alter system to change it in the spfile).
I have done some tests, but I can’t find a reason for this. Just changing it with alter system (scope spfile) seems to work just as well.

I don’t know if this work-around also works for ACFS, but at least there is a solution for pure ASM disks (but you will need to schedule downtime).

Tags: Blog
Exadata install partner specialization announcement!
Exitas Technical Lunch: High Availability in the LAMP stack

You May Also Like