Setting up AIX file services

Setting up the hardware

For essential data it is a good idea to keep them on two separate mirrored identical disks.
Additional disks are usually automatically recognized on boot. If they are added to a running system, you have to configure them manually:

   smitty
   Devices - Install/Configure Devices Added After IPL

Just keep the menu item

   INPUT device / directory for software   [none]

unless the new device needs extra drivers (for disks almost never).

Setting up a volume group

All physical disks must be assigned to a volume group (VG) before they can be used. If the desired VG, for example vgmisc, does not yet exist, create it as follows:

   smitty
   System Storage Management - Logical Volume Manager - 
   Volume Groups - Add a Volume Group

On AIX >=5.3
you have to choose the VG type: Original, Big or Scalable.
Original corresponds to the VGs known from AIX<=5.2 and usually suffices. For special requirements matched by the other two, refer to IBM documentation for details.

For all versions of AIX (at the time of this writing up to 7.1) one may accept the default settings, except:

  VOLUME GROUP name                                  [vgmisc]
  Physical partition SIZE in megabytes                                       +
* PHYSICAL VOLUME names                              []                      +

The entry Physical partition SIZE in megabytes specifies the smallest increment by which disk space is allocated to logical volumes.
Note that a partition in AIX speak has nothing to do with the partitions known from PC operating systems (their analogy would roughly be logical volumes mentioned later).
An Original VG can have up to 1016 "Physical partitions" (PPs) which have to be chosen large enough so that the whole disk can be covered. Here's a quick lookup of PP size vs disk size:

Disk PP size

36GB 32MB

146GB 256MB

300GB 512MB

450GB 512MB

600GB 1024MB

Disk	PP size
36GB	32MB
146GB	256MB
300GB	512MB
450GB	512MB
600GB	1024MB

For the PHYSICAL VOLUME names you'll have the choice among all active disks not yet assigned to a VG. If you plan disk mirroring, you may choose both or just one, and add the second one later.

Setting up a logical volume

Create a new logical volume (LV), for example lvscratch:

   smitty
   System Storage Management - Logical Volume Manager - 
   Logical Volumes - Add a Logical Volume

   VOLUME GROUP name [vgmisc]

Most default settings can be left unchanged, except:

example for AIX <5.2 :

  Logical volume NAME [lvscratch]
* Number of LOGICAL PARTITIONS [128]                    #
  Logical volume TYPE          [jfs]                   +
  MAXIMUM NUMBER of LOGICAL PARTITIONS [512]            #

This defines a logical volume of 64GB maximum and 16GB actual size, given a physical partition size of 128MB.

example for AIX >=5.2 :

  Logical volume NAME                                [lvscratch]
* Number of LOGICAL PARTITIONS                       [1]                      #
  Logical volume TYPE                                [jfs2]                  + 
  RANGE of physical volumes                           minimum                + 
  Number of COPIES of each logical                    2                      +
    partition                                                                 
  MAXIMUM NUMBER of LOGICAL PARTITIONS               [2048]                   #

Number of LOGICAL PARTITIONS is set to the minimum value, corresponding to a single physical partition (PP). It can be increased conveniently later via the file system size definition.
Number of COPIES corresponds to the actual number of disks in the VG. Set it to 1, if only one disk is used.
MAXIMUM NUMBER of LOGICAL PARTITIONS is an educated guess covering the whole disk. It can be increased later, if needed.

Setting up a file system

Create a new file system, for example fsscratch:

Example for AIX <5.2 and/or jfs :

  smitty
  
  System Storage Management - File Systems - Add File Systems - JFS - Add a JFS on a Previously Defined LV - Add a Large File Enabled Journaled File System
* LOGICAL VOLUME name  lvscratch                                  
* MOUNT POINT                  [/fsscratch]
  Mount AUTOMATICALLY at system restart?   yes                    +
  Start Disk Accounting?                              no                     +

The choice

  Number of bytes per inode                           4096                   +
  Allocation Group Size (MBytes)                      64

defines a filesystem of max 64GB size.
Specifying

  Number of bytes per inode                           16384                  +
  Allocation Group Size (MBytes)                      8

allows a size of 256GB.

Example for AIX >=5.2 and/or jfs2:

  smitty

  System Storage Management - File Systems - 
  Add / Change / Show / Delete File Systems - 
  Enhanced Journaled File Systems - 
  Add an Enhanced Journaled File System on a Previously Defined Logical Volume

Default settings can be left unchanged, except:


* LOGICAL VOLUME name                                 lvscratch               +
* MOUNT POINT                                        [/fsscratch]
  Mount AUTOMATICALLY at system restart?              yes                    +
  ENABLE Quota Management?                            all                    +

Mount the filesystem:

  smitty
  System Storage Management - File Systems - Mount a File Systems
   
    
  FILE SYSTEM name                                   [/dev/lvscratch]        +
  DIRECTORY over which to mount                      [/fsscratch]            +

# or on the command line:

/usr/sbin/mount /dev/lvscratch /fsscratch

Adjust the filesystem size, for example:

  smitty
  System Storage Management - File Systems -
  Add / Change / Show / Delete File Systems -
  [Enhanced] Journaled File Systems -
  Change / Show Characteristics of an [Enhanced] Journaled File System 
    
  File system name                                    /fsscratch
  NEW mount point                                    [/fsscratch]
  SIZE of file system
          Unit Size                                   Gigabytes              +
          Number of units                            [128]                    #

If applicable and not yet done, follow these instructions to mirror the VG and/or the LV. At least disable the quorum for mirrored VGs:
```
   chvg -Q'n' vgmisc
```
otherwise both disks would be needed to keep the VG online, which is not what you want in case just one disk fails.

Set appropriate permissions, for example a group-enabled FS:

chmod g+sw              /fsscratch
chown ourgroup:ourgroup /fsscratch

For file systems which should be cleared periodically (typically some scratch or /tmp) adapt the /etc/skulker script or use some equivalent (scavenger) and place a respective entry in the crontab file, for example:
```
crontab -e

0 3 * * * /usr/sbin/skulker
# or 
0 3 * * * /opt/bin/scavenger -t+1 /fsscratch >/dev/null
```
LAST BUT NOT LEAST: Don't forget to setup ADSM/TSM backup !

Enabling NFS access

If the filesystem should be accessed by other nodes (NFS clients), the following steps are necessary.

Export the filesystem:

smitty

System Storage Management - File Systems - Change File Systems - 
Network File System (NFS) - Add a Directory to Exports List

You may just accept the default settings, except:

* Pathname of directory to export                    [/fsscratch]             /

* Security method 1                                  
      Hosts & netgroups allowed client access        []              
      Hosts allowed root access                      []

netgroups are defined via NIS and conveniently combine groups of hosts allowed client access under a unique name. If nothing is specified, all hosts are allowed, except the root user, which has to be enabled separately.
If necessary, you may choose (carefully and restrictive!) hosts which should be enabled for root access.

Enable NFS mounts on the client machines:
- If the filesystem is subject to automount:
  1. Adapt the respective configuration file on every client machine (including fileserver itself), for example /etc/auto.s
```
bio -rw,bg,soft,intr,retry=10 fileserver:/fsscratch/bio
```
  2. Maintenance of configuration files may be facilitated using rdist:
```
rdist -f /bio/local/../autofs.distfile
```
    will distribute the configuration files over all machines defined in the distfile.
  3. If the filesystem is referenced by NFS clients, unmount it on every client:
```
   umount -f /mount_point_for_fsscratch
```
    If, as usual, this can't be accomplished because somebody is using the filesystem, use
```
fuser /mount_point_for_fsscratch
# or:
lsof
```
    to discover and kill -9 the offending processes.
    Brute force version:
```
fuser -kxuc /mount_point_for_fsscratch
```
    kills everything associated with /mount_point_for_fsscratch.
    Update the automount daemon on all clients, for example:
```
  umount -f /mount_point_for_fsscratch; stopsrc -s automountd; startsrc -s automountd
```
    or, if this doesn't work:
```
  umount -f /mount_point_for_fsscratch; stopsrc -s automountd; /usr/sbin/automount -m -v -f /etc/auto_master
```
- If fixed mounts are used ... (to come)

Last update: 5-Dec-2012, M.Kraemer