Configuration and Tuning GPFS for Digital Media Environments

50 Configuration and Tuning GPFS for Digital Media Environments

Figure 2-11 Relationship and workflow between components, functions, and data access

2.2.7 Replication (disk failure groups)

As we have mentioned before, GPFS is a journaling file system (logs all

metadata activity). To maintain the file system online (mounted) and consistent,

several protection mechanisms are used:

򐂰 Data and metadata replication

򐂰 Log file replication

򐂰 Disk failure groups

The disk failure groups are used for keeping the file system (not the cluster!)

online based on maintaining a quorum file system descriptors (disks) available. If

you are familiar with AIX logical volume manager, the information about a volume

group is stored on the Volume Group Descriptor Area (VGDA), and a quorum of

VGDAs must be available for the VG to be either brought online or to remain

varied on.

GPFS reserves a storage area on every disk which is part of a file system for the

file system descriptor (FSDesc), but by default writes the FSDesc on three of the

disks which are part of that file system (if available).

FS Mgr.

Token Mgr.

lease

mmfsd

node A

MetaNode

lease

mmfsd

node B

lease

mmfsd

node C

MM M M M MM M M M

MM M M M MM M M M

MM M M M MM M M M

MM M M M MM M M M

MM M M M MM M M M

CfgMgr

mmfsd

node D

Request for

accessing a file

1

2

3

4

a.mpg

Chapter 2. Introduction to GPFS 51

This behavior poses the problem that in case of multiple disks failure, the file

system may be unmounted due to the lack of FSDesc quorum (even though the

total number of disks still available for that file system is larger that 50%+1).

To solve this problem, GPFS uses disk failure groups. The characteristics of the

failure groups are:

򐂰 A number identifying the failure group to which a disk belongs

򐂰 Is used by GPFS during metadata and data placement on the disks of a file

system

򐂰 Ensures that no two replicas of the same block will become unavailable due to

a single failure

򐂰 Are set by GPFS by default

򐂰 Can also be defined either at NSD creation time (mmcrnsd/mmcrvsd

commands) or later (mmchdisk command)

򐂰 The syntax for the disk descriptor file is the following

DiskName:PrimaryNSDServer:BackupNSDServer:DiskUsage:FailureGroup:NSDName

򐂰 It is important to set failure groups correctly to have proper/effective file

system replication (metadata and/or data)

The following paragraphs explain in more detail considerations for choosing the

hardware and the failure groups to provide maximum availability for your file

system.

Data replication

The GPFS replication feature allows you to specify how many copies of a file to

maintain. File system replication assures that the latest updates to critical data

are preserved in the event of hardware failure. During configuration, you assign a

replication factor to indicate the total number of copies you want to store,

currently only two copies are supported.

Replication allows you to set different levels of protection for each file or one level

for an entire file system. Since replication uses additional disk capacity and

stresses the available write throughput, you might want to consider replicating

only file systems that are frequently read from but seldom written to or only for

metadata. In addition, only the primary copy is used for reading, unlike for

instance in RAID 1, hence no read performance increase can be achieved.

Failure groups

GPFS failover support allows you to organize your hardware into a number of

failure groups to minimize single points of failure. A failure group is a set of disks

that share a common point of failure that could cause them all to become

simultaneously unavailable, such as the disk enclosures, RAID controllers or

storage units such as the DS4500.