My Oracle Support Banner

How to reconfigure the quorum disks if one or more quorum disks are missing in Exadata. (Doc ID 2901581.1)

Last updated on OCTOBER 09, 2023

Applies to:

Oracle Database - Enterprise Edition - Version 12.2.0.1 and later
Linux x86-64

Goal

On Exadata that has three cell servers, the best practice for using high redundancy diskgroups recommends that quorum disks are created on DB nodes.
The instructions here apply situations where at least one the quorum-disks is lost, and the goal is to reconfigure the quorum disks at this point.

Grid Infrastructure (GI) requires that a majority of voting files are online, and asm requires that a majority of PST are online, so for a high redundancy disk group, the best practice is to have five failgroups that can host voting files for GI and PST files for asm diskgroups, one per failgroup. This allows asm and GI to tolerate losing two voting files or PST files, but this can be a problem when there are only three failgroups (i.e. three cell servers in Exadata) because only three voting files and PST are created, one per failgroup. As a result, this allows only one cell server to be down (like Normal redundancy diskgroup) although all data are mirrored in all three failgroups.

To overcome this, Oracle recommends that quorum disks are created because each quorum disk will be its own failgroup. Only one quorum disk per diskgroup per node can be used, and on Exadata that has only three cell servers, asm will use two quorumd disks from DB nodes to have five failgroups for the high redundancy diskgroup.

If quorum disks are needed, Oracle recommends that quorum disks be created on every DB node. For example, if the Exadata has four DB nodes and three cell servers, there should be 4 quorum disks (one per node) for each high redundancy diskgroup. GI and ASM use as many quorum disks as needed. In case of Exadata that has three cell servers, two quorum disks are used to create 5 failgroups. Having unused quorum disks on DB nodes 3 and 4 prompts asm to failover the offlined voting file and PST files from DB nodes 1 and 2 to unused quorum disks on DB nodes 3 and 4.

Because quorum disks host voting files and PST files, losing quorum disks reduces the number of available voting files and PST files. For example, if both quorum disks are lost or offline, only three out of 5 voting files and PST files are available, so losing a cell server in Exadata (e.g. due to a planned maintenance) will cause the GI and asm diskgroups to be offline that will crash databases. Therefore, having quorum disks always available is very important.

Solution

To view full details, sign in with your My Oracle Support account.

Don't have a My Oracle Support account? Click to get started!


In this Document
Goal
Solution
References


My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.