4    Creating LSM Disks, Disk Groups, and Volumes

This chapter describes how to:

Use the information from the worksheets you filled out in Chapter 2 to create disk groups and LSM volumes.

4.1    Overview of Creating LSM Disks

You create an LSM disk when you initialize or encapsulate a disk or disk partition for LSM use. Specifying a disk name, such as dsk10, initializes the entire disk as an LSM sliced disk. Specifying a partition name, such as dsk10g or dsk10c, initializes the partition as an LSM simple disk. Encapsulating a disk or disk partition that contains data you want placed under LSM control creates a nopriv disk.

Initializing an LSM disk:

Note

If the disk is new to the system and LSM is already running, enter the hwmgr scan scsi command, then the voldctl enable command, to make LSM recognize the disk.

To create LSM sliced or simple disks, you can use either the voldisksetup command or the voldiskadd script. (To encapsulate a disk or partition, see Section 4.6.1.)

4.1.1    Overview of Configuration Database Copies

By default, LSM configures each sliced or simple disk with the potential to have one copy of the configuration database for its disk group. For disk groups with fewer than four disks, configure each disk to have two copies of the configuration database to ensure multiple copies in case one or more disks fail. Use the voldisksetup command to specify the number of configuration database copies on an LSM disk.

An LSM sliced or simple disk can have 0, 1, or 2 copies of the configuration database for its disk group. LSM enables the specified number of configuration database copies on each disk only when you add the disk to a disk group and only if necessary to maintain the proper number and distribution of copies for the disk group as a whole. For most system configurations, initialize your LSM disks with the default number of copies and allow LSM to manage them.

To maintain the proper number and distribution of LSM configuration database copies in Fibre Channel environments, see the Best Practice entitled Ensuring Redundancy of LSM Configuration Databases on a Fibre Channel at the following URL:

http://www.tru64unix.compaq.com/docs/best_practices

4.1.2    Overview of Disk Offsets

When you initialize a disk for LSM use, by default LSM skips over the first 16 blocks on the disk to preserve the disk header and bootstrap information. The public region of a sliced disk and the private region of a simple disk start at the first block after this default offset. Using the voldisksetup command, you can specify a different offset, if necessary.

For example, you can set up your LSM disks to align I/O requests to the chunk size of an underlying RAID hardware device. For more information on this specific application, see the Best Practice entitled Aligning LSM Disks and Volumes to Hardware RAID Devices at the following URL:

http://www.tru64unix.compaq.com/docs/best_practices

4.1.3    Creating LSM Sliced Disks

The following examples show how to create LSM sliced disks with various attributes:

Optionally (but recommended), make a backup copy of the disk label information for all your LSM disks (Section 4.1.5).

After you initialize a disk or disk partition as an LSM disk, you can add it to a disk group. For information on creating a disk group, see Section 4.2. For information on adding an LSM disk to an existing disk group, see Section 5.2.2.

4.1.4    Creating LSM Simple Disks

The default private region offset for simple disks is 16 blocks on the a or c partitions and 0 blocks on all other partitions. The default number of configuration database copies is 1.

The following examples show how to create LSM simple disks with various attributes:

Optionally (but recommended), make a backup copy of the disk label information for all your LSM disks (Section 4.1.5).

After you initialize an LSM disk, you can add it to a disk group. For information on creating a disk group, see Section 4.2. For information on adding an LSM disk to an existing disk group, see Section 5.2.2.

4.1.5    Backing Up Disk Label Information

Back up the updated disk label information for each LSM disk. Having this information will simplify the process of replacing a failed disk, by allowing you to copy the failed disk's attributes to a new disk. After a disk fails, you cannot read its disk label and therefore cannot copy that information to a new disk.

You can back up the disk label information before or after adding a disk to a disk group; the information does not change.

To back up the disk label information: 

# disklabel dskn > file

For more information, see disklabel(8).

4.2    Creating Disk Groups

The default rootdg disk group is created when you initialize LSM and always exists on a system running LSM. You can create additional disk groups to organize your disks into logical sets. Each disk group that you create must have a unique name and must contain at least one sliced or simple LSM disk to store the disk group's configuration database. An LSM disk can belong to only one disk group. [Footnote 1]

For large LSM configurations, consider keeping rootdg fairly small (ten disks or fewer) and create other disk groups for the rest of your LSM configuration. If possible, use rootdg only for volumes relating to the system disk (on a standalone system), the clusterwide root, /usr, and /var file system domains, and members' swap devices.

Having separate disk groups gives you the ability to move LSM volumes to different systems or clusters.

You can create an LSM disk group using the following commands:

4.2.1    Creating LSM Disks and Disk Groups Using the voldiskadd Script

The voldiskadd script lets you do all of the following tasks in one interactive session:

Note

If a disk group will have fewer than four disks, see Section 4.1.1.

You can invoke the voldiskadd script with or without a disk name. If you invoke the script by itself, it prompts you for the following information:

Example 4-1 uses a disk named dsk9 to create a disk group named dg1:

Example 4-1:  Creating an LSM Disk Group with the voldiskadd script

# voldiskadd dsk9


Add or initialize disks
 
Menu: VolumeManager/Disk/AddDisks
 
  Here is the disk selected.
 
  dsk9
 
Continue operation? [y,n,q,?] (default: y) [Return]
 
  You can choose to add this disk to an existing disk group, a
  new disk group, or leave the disk available for use by future
  add or replacement operations. To create a new disk group,
  select a disk group name that does not yet exist. To leave
  the disk available for future use, specify a disk group name
  of "none".
 
Which disk group [<group>,none,list,q,?] (default: rootdg) dg1
 
  There is no active disk group named dg1.
 
Create a new group named dg1? [y,n,q,?] (default: y) [Return]
 
  The default disk name that will be assigned is:
 
  dg101
 
Use this default disk name for the disk? [y,n,q,?] (default: y) [Return]
 
Add disk as a spare disk for dg1? [y,n,q,?] (default: n) [Return]
 
  A new disk group will be created named dg1 and the selected disks
  will be added to the disk group with default disk names.
  dsk9
 
Continue with operation? [y,n,q,?] (default: y) [Return]
 
  The following disk device has a valid disk label, but does 
  not appear to have been initialized for the Logical Storage 
  Manager. If there is data on the disk that should NOT be 
  destroyed you should encapsulate the existing disk partitions 
  as volumes instead of adding the disk as a new disk.
 
  dsk9
 
Initialize this device? [y,n,q,?] (default: y) [Return]
 
  Initializing device dsk9.
 
  Creating a new disk group named dg1 containing the disk
  device dsk9 with the name dg101.
 
Goodbye.

4.2.2    Creating Disk Groups Using the voldg Command

Use the voldg command to create disk groups from disks that are initialized for LSM, including disks configured to have a nondefault number of configuration database copies (Section 4.1).

By default, LSM maintains a minimum of four active copies of the configuration database in each disk group. You can specify a different number of active copies even if you are using disks initialized with the defaults. You can specify a maximum equal to the number of sliced and simple disks in the disk group.

For example, if you create a disk group with ten sliced or simple disks, each of which is configured by default to store one copy, you can set the number of copies for that disk group to ten. If you configured each disk to store two copies, you can set the number of copies for the disk group to 20.

For any disk group, the maximum number of active configuration database copies is derived from the number of sliced or simple disks in the group and the number of copies each of those disks is configured to store.

Note

You can set the number of configuration database copies that a disk group will maintain only when you create the disk group. You cannot change the number of active copies for existing disk groups.

Maintaining more than the default number of copies can affect performance, because every change to the LSM configuration is written to all active copies of the database. If you want to use a nondefault number, choose a number of copies sufficient to meet your environment's needs but small enough to minimize the performance impact.

To create a disk group with default values using the voldg command: 

# voldg init disk_group disk [disk...]

For example: 

# voldg init dg1 dsk5 dsk6 dsk7 dsk9 dsk10 dsk11 dsk12

If a disk group will have fewer than four disks, configure each disk to have two copies of the disk group's configuration database (Section 4.1) to ensure that the disk group has multiple copies in case one or more disks fail.

To create a disk group and set the number of configuration copies to 10: 

# voldg init newdg disks nconfig=10

For example: 

# voldg init newdg dsk100 dsk101 dsk102... dsk110 nconfig=10

4.3    Creating LSM Volumes for New Data

To create an LSM volume for a new file system or application, use the volassist command. The volassist command either finds the necessary space within the disk group and creates all the objects for the volume or uses attributes you supply on the command line, such as specific disk names. You must assign a name and length (size) on the command line.

You can specify values for other LSM volume attributes on the command line or in a text file that you create. If you do not specify a value for an attribute, LSM uses a default value.

By default, LSM uses a stripe width of 64K bytes for striped plexes and 16K bytes for RAID5 plexes. However, you can use a different stripe width if you have applications that use an I/O transfer size that requires a different value, or if you have created hardware devices with a particular stripe width and you want the LSM volume created from those devices to align its writes (data stripes) to the hardware's stripe width.

For more information, see the Best Practice entitled Aligning LSM Disks and Volumes to Hardware RAID Devices, available at the following URL:

http://www.tru64unix.compaq.com/docs/best_practices/sys_bps.html

4.3.1    Overview of LSM Volume Attributes

The following lists the priority given to assignable attributes:

  1. Values on the command line

  2. Values in a file that you specify by using the volassist -d option

  3. Values in the /etc/default/volassist file

  4. Default values

To display the default values for volume attributes: 

# volassist help showattrs


#Attributes:
 layout=nomirror,nostripe,span,nocontig,raid5log,noregionlog,nofpalog,diskalign,nostorage
 mirrors=2 columns=0 nlogs=1 regionlogs=1 raid5logs=1 fpalogs=1
 min_columns=2 max_columns=8
 regionloglen=0 raid5loglen=0 logtype=region
 stripe_stripeunitsize=128 raid5_stripeunitsize=32
 usetype=fsgen diskgroup= comment="" fstype=
 user=0 group=0 mode=0600
 probe_granularity=2048
 alloc=
 wantalloc=
 mirror=

Some volume attributes have several options to define them. Some options define an attribute globally, where others define an attribute for a specific plex type. For example, you can specify the size of a stripe data unit using the stripeunit (or stwidth) option for both striped or RAID5 plexes, the stripe_stripeunit (or stripe_stwid) option specifically for striped plexes, or the raid5_stripeunit (or raid5_stwid) option specifically for RAID5 plexes.

For a complete list of attributes, see volassist(8). Table 4-1 describes some of the common attributes for which you can specify a value.

Table 4-1:  Common LSM Volume Attributes

Attribute Description Attribute Options
Plex type layout={concatenated|striped|raid5}
Usage type -U {fsgen|gen|raid5}
Number of plexes (mirrors). Default is 2. mirror={number|yes|no}

Type of log:

  • drl for mirrored volumes (only)

  • region for mirrored or RAID5 volumes

logtype={drl|region|none}
Number of FPA logs (for mirrored volumes only) nfpalog=number
Size of stripe width for a striped or RAID5 plex, in blocks, sectors, kilobytes, megabytes, or gigabytes stripeunit=data_unit_size or stwid=data_unit_size
Number of columns for a striped or RAID5 plex; typically, the number of disks in each plex nstripe=number_of_columns or ncolumn=number_of_columns

Creating a text file that specifies many of these attributes is useful if you create many LSM volumes that use the same nondefault values for attributes. Any attribute that you can specify on the command line can be specified on a separate line in the text file. By default, LSM looks for the /etc/default/volassist file when you create an LSM volume. If you created an /etc/default/volassist file, LSM creates each volume using the attributes that you specify on the command line and in the /etc/default/volassist file.

Example 4-2 shows an LSM volume attributes file called /etc/default/volassist that creates an LSM volume using a four-column striped plex with two mirrors, a stripe width of 32K bytes, and no log.

Example 4-2:  LSM Volume Attribute Defaults File

# LSM Vn.n
# volassist defaults file. Use '#' for comments
# number of stripes
nstripe=4
# layout
layout=striped
# mirroring
nmirror=2
# logging
logtype=none
# stripe size
stripeunit=32k

For example, to create an LSM volume using the attributes in the /etc/default/volassist file: 

# volassist make volume length

If you have created a custom attributes file and want LSM to use the applicable attributes in that file when creating the volume, specify the attributes file as follows: 

# volassist make volume length -d filename

With this option, LSM creates the volume using both the attributes that you specify on the command line (such as name and length) and those in the named file. If you specify an attribute that conflicts with the contents of the file, the command line takes precedence.

To specify a length (size) for the volume, enter a number and the appropriate suffix:

Suffix Unit
b Blocks
s Sectors (default)
k Kilobytes
m Megabytes
g Gigabytes
t Terabytes

4.3.2    Creating LSM Volumes with a Single Concatenated Plex

A volume with a single concatenated plex is also called a simple volume. It provides no data redundancy; if a disk fails, the data is lost. To avoid this, you can either create the volume initially with mirrored contatenated plexes (Section 4.3.3) or add another data plex to the volume later (Section 5.5.2).

You can either let LSM find space on any available disks in the disk group or specify which disks you want LSM to use in creating the volume.

For volumes that will support a file system, use the default fsgen usage type. For volumes that will contain raw data, such as a database, use the gen usage type.

4.3.3    Creating LSM Volumes with Mirrored, Concatenated Plexes

To provide data redundancy (high availability) you can create an LSM volume with two or more concatenated plexes. To further increase availability, you can specify disks on a different bus for each data plex and the DRL plex by making the volume in steps, allowing you to control which disks LSM uses to create each plex.

For volumes that will support a file system, use the default fsgen usage type. For volumes that will contain raw data, such as a database, use the gen usage type. You specify a usage type only when you create the volume, but not when you add a mirror plex or a log plex to an existing volume.

To improve availability of the logs, you can add multiple logs when you create the volume, with the nlog=count attribute. You can also add one or more FPA logs to the volume when you create it, with the nfpalog=count attribute. An FPA log is not required; it is used only when you create a secondary volume from one plex of a primary volume, (Section 5.4.2.2). Multiple FPA logs ensure that FPA logging remains in effect in case of disk failure.

4.3.3.1    Creating a Mirrored, Concatenated Volume in One Step

4.3.3.2    Creating a Mirrored, Concatenated Volume with Plexes on Different Buses

To create a mirrored, concatenated volume with each plex on a different bus:

  1. Create the volume with a single concatenated plex, specifying the disks: 

    # volassist [-g disk_group] [-U use_type] make \
volume length disks

    For example: 

    # volassist -g dg1 -U gen make vol2 3g dsk2 dsk3 dsk4

  2. Add another concatenated plex (mirror) to the volume, specifying disks on a different bus: 

    # volassist [-g disk_group] mirror volume disks

    For example: 

    # volassist -g dg1 mirror vol2 dsk5 dsk6 dsk7

    When you add a mirror to a volume manually, LSM does not add a DRL plex.

  3. Add a DRL plex to the volume, specifying a disk that is not used by one of the data plexes: 

    # volassist -g dg1 addlog volume disk

    For example: 

    # volassist -g dg1 addlog vol2 dsk8

    The volume is ready for use.

4.3.4    Creating LSM Volumes with a Single Striped Plex

An LSM volume with a striped plex can offer faster performance than a volume with a concatenated plex. You can specify the number of disks you want LSM to stripe the data over (the number of stripe columns), or you can let LSM stripe the data over as many disks as necessary based on the volume size and the stripe width you specify, if other than the default.

A volume with a single striped plex does not provide data redundancy; if a disk fails, the data is lost. To avoid this, you can either create the volume initially with mirrored striped plexes (Section 4.3.5) or add another data plex to the volume later (Section 5.5.2).

Note

In general, do not use LSM to stripe data if you also use a hardware controller to stripe data. In some specific cases such a configuration can improve performance but only if:

The number of LSM columns in each plex in the volume should be equal to the number of hardware RAID controllers. See your hardware RAID documentation for information about how to choose the best number of columns for the hardware RAID set.

This section contains a number of examples of creating volumes with one or more striped plexes. In the examples, the syntax for specifying a different stripe width is shown. To use the default stripe width, omit that option.

By default, the volassist command creates columns for a striped plex on disks in alphanumeric order, regardless of their order on the command line. To improve performance, create the columns in each plex using disks on different buses. For more information about specifying the disk order for columns in a striped plex, see Section 4.4.2.

When creating volumes with a striped plex, you must specify the number of stripe columns per plex. Each column must be on a different disk; therefore, this is the number of disks over which to stripe the plex.

For volumes that will support a file system, use the default fsgen usage type. For volumes that will contain raw data, such as a database, use the gen usage type. You specify a usage type only when you create the volume, but not when you add a mirror plex or a log plex to an existing volume.

The following examples show how to create LSM striped volumes with different properties:

4.3.5    Creating LSM Volumes with Mirrored, Striped Plexes

To provide data redundancy (high availability) and improved performance, you can create an LSM volume with two or more striped plexes.

You can either let LSM find space on any available disks in the disk group or specify which disks you want LSM to use in creating the volume. To further increase performance and availability, you can specify disks on a different bus for each plex and the DRL plex by making the volume in steps, allowing you to control which disks LSM uses to create each plex.

To improve availability of the logs, you can add multiple logs with the nlog=count attribute when you create the volume. You can also add one or more FPA logs to the volume with the nfpalog=count attribute when you create it. An FPA log is not required; it is used only when you create a secondary volume from one plex of a primary volume (Section 5.4.2.2). Multiple FPA logs ensure that FPA logging remains in effect in case of disk failure.

When creating volumes with a striped plex, you must specify the number of stripe columns per plex. Each column must be on a different disk; therefore, this is the number of disks over which to stripe the plex.

For volumes that will support a file system, use the default fsgen usage type. For volumes that will contain raw data, such as a database, use the gen usage type. You specify a usage type only when you create the volume, but not when you add a mirror plex or a log plex to an existing volume.

4.3.5.1    Creating a Mirrored, Striped Volume in One Step

4.3.5.2    Creating a Mirrored, Striped Volume with Plexes on Different Buses

For improved availability, you can create an LSM volume with each data plex and the log plex on a different bus. If your configuration does not support this, you can still use the following procedure to specify which disks LSM uses to create each data plex and the log plex.

The following procedure shows how to ensure that LSM creates each plex on the disks you specify. The ncolumn option forces LSM to stripe the plex across all the named disks. The ncolumn value must equal the number of disks you specify.

Note

Each data plex is a complete copy of the volume, and uses as much disk space as the volume size you specify.

  1. Create the volume with a single striped plex, specifying disks on one bus: 

    # volassist [-g disk_group] [-U use_type] make volume length \
layout=stripe ncolumn=number_of_columns \
[stwid=data_unit_size] disks

    For example, to create a 1GB volume named vstripe in the dg1 disk group with the default usage type of fsgen and one 3-column striped plex with the default stripe width on disks dsk10, dsk11, and dsk12: 

    # volassist -g dg1 make vstripe 1g layout=stripe \
ncolumn=3 dsk10 dsk11 dsk12

    The volume looks similar to the following: 

    Disk group: dg1
 
V  NAME         USETYPE      KSTATE   STATE    LENGTH   READPOL   PREFPLEX
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
v  vstripe      fsgen        ENABLED  ACTIVE   2097152  SELECT    vstripe-01
pl vstripe-01   vstripe      ENABLED  ACTIVE   2097408  STRIPE    3/128    RW
sd dsk10-01     vstripe-01   dsk10    0        699136   0/0       dsk10    ENA
sd dsk11-01     vstripe-01   dsk11    0        699136   1/0       dsk11    ENA
sd dsk12-01     vstripe-01   dsk12    0        699136   2/0       dsk12    ENA
 

  2. Add a mirror plex to the volume, specifying disks on a different bus: 

    # volassist [-g disk_group] mirror volume disks

    For example: 

    # volassist -g dg1 mirror vol_stripe dsk19 dsk20 dsk21

  3. Add a DRL plex to the volume, specifying a disk that is not used by one of the data plexes, and if possible, on a different bus: 

    # volassist [-g disk_group] addlog volume disk

    For example: 

    # volassist -g dg1 addlog vol_stripe dsk26

    The completed volume is ready for use and looks similar to the following: 

    Disk group: dg1
 
V  NAME         USETYPE      KSTATE   STATE    LENGTH   READPOL   PREFPLEX
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
v  vstripe      fsgen        ENABLED  ACTIVE   2097152  SELECT    -
pl vstripe-01   vstripe      ENABLED  ACTIVE   2097408  STRIPE    3/128    RW
sd dsk10-01     vstripe-01   dsk10    0        699136   0/0       dsk10    ENA
sd dsk11-01     vstripe-01   dsk11    0        699136   1/0       dsk11    ENA
sd dsk12-01     vstripe-01   dsk12    0        699136   2/0       dsk12    ENA
pl vstripe-02   vstripe      ENABLED  ACTIVE   2097408  STRIPE    3/128    RW
sd dsk19-01     vstripe-02   dsk19    0        699136   0/0       dsk19    ENA
sd dsk20-01     vstripe-02   dsk20    0        699136   1/0       dsk20    ENA
sd dsk21-01     vstripe-02   dsk21    0        699136   2/0       dsk21    ENA
pl vstripe-03   vstripe      ENABLED  ACTIVE   LOGONLY  CONCAT    -        RW
sd dsk26-01     vstripe-03   dsk26    0        65       LOG       dsk26    ENA
 

4.3.6    Creating LSM Volumes with a RAID 5 Plex

A volume with a RAID 5 data plex uses distributed parity to provide data redundancy. When you create the volume, you can use the default values for the number of columns in the plex (minimum of three, maximum of eight) and the stripe width (16K bytes), as well as let LSM use any available space in the disk group to create the volume. Or you can specify the number of columns, the stripe width, and the disks to use.

If you specify the disks to use, by default the volassist command creates the columns for a RAID5 plex on disks in alphanumeric order, regardless of their order on the command line, and automatically creates a RAID5 log plex for the volume on a separate disk. LSM will not create the log plex for a RAID 5 volume on a disk used by the data plex, as it does for mirrored volumes. You must specify enough disks to create the data plex and log plex.

To improve performance, you can create the columns on disks on different buses. For more information about specifying the disk order for columns in a RAID5 plex, see Section 4.4.3.

The usage type for all volumes with a RAID5 plex is raid5, regardless of what the volume is used for. When you specify layout=raid5, LSM automatically applies the raid5 usage type.

4.3.7    Creating LSM Volumes for Swap Space

To protect against system or cluster member crashes due to swap disk errors, you can create LSM mirrored volumes for swap space. For recommendations on the amount of swap space to configure, see the System Administration manual or the Cluster Administration manual.

HP recommends that you use multiple disks for secondary swap devices and add the devices as several individual volumes, instead of striping or concatenating them into a single large volume. This makes the swapping algorithm more efficient.

The way you use LSM volumes for swap space differs depending on the environment:

All swap volumes for both standalone systems and cluster members must belong to the rootdg disk group. If there is not enough free disk space to create the volumes, add more disks to rootdg (see Section 5.2.2).

4.3.7.1    Creating Swap Volumes

Swap volumes, if mirrored, should not use dirty region logging (DRL). After you create a swap volume, you should also modify the volume's recovery policy, so that LSM will not resynchronize the plexes after a system failure. On a cluster member, choose disks that are local to the member, if possible. The disks must belong to the rootdg disk group.

To create an LSM volume for secondary swap space on a standalone system or additional swap space on a cluster member:

  1. Create the volume with no DRL plex, if mirrored. Assign the volume one of the following usage types: 

    # volassist -U use_type make volume length [nmirror=count] \
[layout=nolog] [disks]

    For example:

  2. If the volume is mirrored, change the volume's recovery policy to prevent plex resynchronization after a system crash; for example: 

    # volume set start_opts=norecov swapvol_2

  3. Make the LSM volume available as swap space using the swapon command; for example: 

    # swapon /dev/vol/rootdg/swapvol_2

  4. Edit the sysconfigtab file to add the volume's device special file to the swapdevice kernel attribute value within the vm: section.

    For example: 

    vm:
  swapdevice = /dev/vol/rootdg/swapvol, /dev/vol/rootdg/swapvol_2

    In a cluster, be sure to modify the appropriate member's file, which is a context-dependent symbolic link (CDSL) in the form /cluster/members/member{n}/boot_partition/etc/sysconfigtab.

4.3.7.2    Mirroring Swap Volumes

The following procedure applies only to secondary swap volumes on a standalone system and to any unmirrored swap volume on a cluster member.

Note

To mirror the primary swap volume on a standalone system, use the volrootmir command. See Section 3.4.2.1.3.

In a cluster, choose an LSM disk that belongs to the rootdg disk group and is local to the member whose swap volume you want to mirror. You can run the command from any cluster member.

To mirror a secondary swap volume on a standalone system or any unmirrored swap volume on a cluster member:

  1. Mirror the volume. In a cluster, specify a disk that is local to the appropriate cluster member: 

    # volassist mirror volume [dskN]

    For example:

  2. Change the volume's recovery policy to prevent plex resynchronization after a system crash; for example: 

    # volume set start_opts=norecov swapvol_2

For more information on configuring additional swap space, see swapon(8), and sysconfig(8).

4.4    Creating LSM Volumes with Nondefault Properties

This section describes how to create LSM volumes with attributes or properties that you cannot always specify with the high-level commands. Only users who have a thorough understanding of LSM, and the need to create such volumes, are advised to use these procedures.

4.4.1    Creating a Striped Plex with Subdisks of Different Sizes

If you have LSM disks of different sizes and want to use space on particular disks to maximize your use of storage, you can use the low-level commands to manually create the subdisks and plex columns, then create the plex, and then create and start the volume.

The makeup of each column can differ, but each column should total the same number of sectors. For example, each column can contain a different number of subdisks, of different sizes.

Each subdisk should be a multiple of the stripe width (and, therefore, so will each column) so that writes align evenly to subdisk boundaries.

The following table shows how three columns that each total 256000 sectors (125 MB) can comprise different numbers and sizes of subdisks. The three columns make up one plex of 768000 sectors (375 MB). These subdisk names and sizes are used as examples in the following procedure.

Striped Plex: plex-01
Column 0 Column 1 Column 2

Subdisk dsk3-01, 128000 Sectors

Subdisk dsk4-01, 128000 Sectors

Subdisk dsk6-01, 64000 Sectors

Subdisk dsk9-01, 64000 Sectors

Subdisk dsk11-01, 64000 Sectors

Subdisk dsk12-01, 64000 Sectors

Subdisk dsk13-01, 51200 Sectors

Subdisk dsk14-01, 51200 Sectors

Subdisk dsk15-01, 51200 Sectors

Subdisk dsk16-01, 51200 Sectors

Subdisk dsk17-01, 51200 Sectors
Total: 256000 Sectors (125 MB) Total: 256000 Sectors (125 MB) Total: 256000 Sectors (125 MB)

The following procedure shows how to make subdisks of different sizes, create a plex with columns of equal size, and create a volume using that plex. The plex uses the default stripe width of 64K bytes and has three columns, with the subdisk sizes shown in the previous table.

Each subdisk is on a different disk and starts at offset 0 in the public region; therefore you need specify only a length for the subdisk. (To create a subdisk on a disk that already has other subdisks, you must specify the offset — the starting point — of the new subdisk as well as its length.)

To create a volume with different sized subdisks:

  1. Decide the size of the volume and the number of columns you want in the plex (or each plex, if mirrored); for the examples that follow, the volume is 375 MB, so each column is one-third of 375, or 125 MB.

  2. Divide the column size (125 MB or 256000 sectors) by the number of intended subdisks per column (in the following examples, 2, 4, and 5) to determine the size of each subdisk. In the first column (two subdisks), each subdisk will be 256000/2, or 128000 sectors, and so on.

    Alternatively, you can work backwards from the amount of space available on various disks, determine how large a subdisk you can create in that space that is a multiple of the stripe width, and calculate how many subdisks you need for each column. The space availability might influence the number of columns.

  3. Create the subdisks: 

    volmake sd subdisk_name disk len=length

    For example: 

    # volmake sd dsk3-01 dsk3 len=128000
# volmake sd dsk4-01 dsk4 len=128000
 
# volmake sd dsk6-01 dsk6 len=64000
# volmake sd dsk9-01 dsk9 len=64000
# volmake sd dsk11-01 dsk11 len=64000
# volmake sd dsk12-01 dsk12 len=64000
 
# volmake sd dsk13-01 dsk13 len=51200
# volmake sd dsk14-01 dsk14 len=51200
# volmake sd dsk15-01 dsk15 len=51200
# volmake sd dsk16-01 dsk16 len=51200
# volmake sd dsk17-01 dsk17 len=51200
 
 

  4. Create a plex with the desired number of columns and associate the appropriate subdisks to each column.

    If there are many subdisks involved, you can create an "empty" plex first, then associate the subdisk groups in separate steps, one step for each column.

    You do not need to specify a size for the empty plex; as you associate subdisk columns, the plex size is updated to map to the end of the longest column. When you create a plex manually, you must set a stripe width.

    For example: 

    # volmake plex plex-01 layout=stripe ncolumn=3 stwidth=64k
 
# volsd assoc plex-01 dsk3-01:0 dsk4-01:0
# volsd assoc plex-01 dsk6-01:1 dsk9-01:1 dsk11-01:1 dsk12-01:1
# volsd assoc plex-01 dsk13-01:2 dsk14-01:2 dsk15-01:2 \
dsk16-01:2 dsk17-01:2

    The plex looks similar to the following: 

    # volprint -pht plex-01


    Disk group: rootdg
 
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
pl plex-01      -            DISABLED -        768000   STRIPE    3/128    RW
sd dsk3-01      plex-01      dsk3     0        128000   0/0       dsk3     ENA
sd dsk4-01      plex-01      dsk4     0        128000   0/128000  dsk4     ENA
sd dsk6-01      plex-01      dsk6     0        64000    1/0       dsk6     ENA
sd dsk9-01      plex-01      dsk9     0        64000    1/64000   dsk9     ENA
sd dsk11-01     plex-01      dsk11    0        64000    1/128000  dsk11    ENA
sd dsk12-01     plex-01      dsk12    0        64000    1/192000  dsk12    ENA
sd dsk13-01     plex-01      dsk13    0        51200    2/0       dsk13    ENA
sd dsk14-01     plex-01      dsk14    0        51200    2/51200   dsk14    ENA
sd dsk15-01     plex-01      dsk15    0        51200    2/102400  dsk15    ENA
sd dsk16-01     plex-01      dsk16    0        51200    2/153600  dsk16    ENA
sd dsk17-01     plex-01      dsk17    0        51200    2/204800  dsk17    ENA
 

  5. Create the volume using the plex: 

    # volmake -U use_type vol volume_name plex=plex_name

    For example: 

    # volmake -U fsgen vol vol-01 plex=plex-01

  6. Start the volume: 

    # volume start vol-01

    The volume is started and ready for use, and looks similar to the following: 

    # volprint -vht vol-01


    Disk group: rootdg
 
V  NAME         USETYPE      KSTATE   STATE    LENGTH   READPOL   PREFPLEX
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
v  vol-01       fsgen        ENABLED  ACTIVE   768000   ROUND     -
pl plex-01      vol-01       ENABLED  ACTIVE   768000   STRIPE    3/128    RW
sd dsk3-01      plex-01      dsk3     0        128000   0/0       dsk3     ENA
sd dsk4-01      plex-01      dsk4     0        128000   0/128000  dsk4     ENA
sd dsk6-01      plex-01      dsk6     0        64000    1/0       dsk6     ENA
sd dsk9-01      plex-01      dsk9     0        64000    1/64000   dsk9     ENA
sd dsk11-01     plex-01      dsk11    0        64000    1/128000  dsk11    ENA
sd dsk12-01     plex-01      dsk12    0        64000    1/192000  dsk12    ENA
sd dsk13-01     plex-01      dsk13    0        51200    2/0       dsk13    ENA
sd dsk14-01     plex-01      dsk14    0        51200    2/51200   dsk14    ENA
sd dsk15-01     plex-01      dsk15    0        51200    2/102400  dsk15    ENA
sd dsk16-01     plex-01      dsk16    0        51200    2/153600  dsk16    ENA
sd dsk17-01     plex-01      dsk17    0        51200    2/204800  dsk17    ENA
 

4.4.2    Creating a Striped Plex with Disks on Different Buses

You can improve performance for a volume with striped plexes by striping each plex over disks on different buses. If you have enough buses, you can mirror the volume on different buses from those supporting the first plex. For example, if you had 12 buses, you could stripe one plex over the first six buses, and stripe a second plex over the other six buses.

Caution

For mirrored volumes, do not create all the data plexes on the same bus; this eliminates the volume's availability. If the bus fails, you lose the entire volume.

Before you begin, decide which LSM disks you want to use, identify which bus each disk is on, and plan how you want to create the volume based on how you want LSM to stripe and mirror the data over the buses.

To create a volume with plexes that are striped down buses and mirrored across buses, you must use low-level commands to create each subdisk, create each plex from those subdisks, create the volume from the plexes, add a log plex, and start the volume.

The following procedure creates a volume with two plexes and a DRL plex using these disks and buses in the rootdg disk group:

Plex plex-01 Plex plex-02
Bus 1 Bus 2 Bus 3 Bus 4
dsk0 dsk7 dsk14 dsk21
dsk1 dsk8 dsk15 dsk22
dsk2 dsk9 dsk16 dsk23
dsk3 dsk10 dsk17 dsk24
dsk4 (for the DRL plex)

The first plex will stripe alternately over buses 1 and 2. The second plex will stripe alternately over buses 3 and 4. The log plex will be on bus 1, the same as the first plex, but on a different disk. Although recommended, it is not always possible to place the log plex on a different bus from the data plexes.

To create an LSM mirrored, striped volume with each plex on different buses:

  1. Create the subdisks on the disks on bus 1 and 2; for example: 

    # volmake sd dsk0-01 dsk0 len=16m
# volmake sd dsk1-01 dsk1 len=16m
# volmake sd dsk2-01 dsk2 len=16m
# volmake sd dsk3-01 dsk3 len=16m
 
# volmake sd dsk7-01 dsk7 len=16m
# volmake sd dsk8-01 dsk8 len=16m
# volmake sd dsk9-01 dsk9 len=16m
# volmake sd dsk10-01 dsk10 len=16m

  2. Create the striped plex, specifying the subdisks in alternating bus order; for example: 

    # volmake plex plex-01 layout=stripe stwidth=64k \
sd=dsk0-01,dsk7-01,dsk1-01,dsk8-01,dsk2-01,dsk9-01, \
dsk3-01,dsk10-01

    This creates an 8-column striped plex that alternates between buses 1 and 2. It is more efficient to stripe over all the disks in a plex.

    The plex looks similar to the following: 

    # volprint -p plex-01


    Disk group: rootdg
 
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
pl plex-01      -            DISABLED -        262144   STRIPE    8/128    RW
sd dsk0-01      plex-01      dsk0     0        32768    0/0       dsk0     ENA
sd dsk7-01      plex-01      dsk7     0        32768    1/0       dsk7     ENA
sd dsk1-01      plex-01      dsk1     0        32768    2/0       dsk1     ENA
sd dsk8-01      plex-01      dsk8     0        32768    3/0       dsk8     ENA
sd dsk2-01      plex-01      dsk2     0        32768    4/0       dsk2     ENA
sd dsk9-01      plex-01      dsk9     0        32768    5/0       dsk9     ENA
sd dsk3-01      plex-01      dsk3     0        32768    6/0       dsk3     ENA
sd dsk10-01     plex-01      dsk10    0        32768    7/0       dsk10    ENA
 

  3. Create the subdisks for the second data plex on the disks on bus 3 and 4; for example: 

    # volmake sd dsk14-01 dsk14 len=16m
# volmake sd dsk15-01 dsk15 len=16m
# volmake sd dsk16-01 dsk16 len=16m
# volmake sd dsk17-01 dsk17 len=16m
 
# volmake sd dsk21-01 dsk21 len=16m
# volmake sd dsk22-01 dsk22 len=16m
# volmake sd dsk23-01 dsk23 len=16m
# volmake sd dsk24-01 dsk24 len=16m

  4. Create the second data plex, specifying the subdisks in alternating bus order; for example: 

    # volmake plex plex-02 layout=stripe stwidth=64k \
sd=dsk14-01,dsk21-01,dsk15-01,dsk22-01,dsk16-01,dsk23-01, \
dsk17-01,dsk24-01

    This creates an 8-column striped plex that alternates between buses 3 and 4.

    The plex looks similar to the following: 

    # volprint -pht plex-02


    Disk group: rootdg
 
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
pl plex-02      -            DISABLED -        262144   STRIPE    8/128    RW
sd dsk14-01     plex-02      dsk14    0        32768    0/0       dsk14    ENA
sd dsk21-01     plex-02      dsk21    0        32768    1/0       dsk21    ENA
sd dsk15-01     plex-02      dsk15    0        32768    2/0       dsk15    ENA
sd dsk22-01     plex-02      dsk22    0        32768    3/0       dsk22    ENA
sd dsk16-01     plex-02      dsk16    0        32768    4/0       dsk16    ENA
sd dsk23-01     plex-02      dsk23    0        32768    5/0       dsk23    ENA
sd dsk17-01     plex-02      dsk17    0        32768    6/0       dsk17    ENA
sd dsk24-01     plex-02      dsk24    0        32768    7/0       dsk24    ENA
 

  5. Create the LSM volume using both data plexes; for example: 

    # volmake vol vol_mirr plex=plex-01,plex-02

    The volume looks similar to the following: 

     # volprint -vht vol_mirr


    Disk group: rootdg
 
V  NAME         USETYPE      KSTATE   STATE    LENGTH   READPOL   PREFPLEX
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
v  vol_mirr     fsgen        DISABLED EMPTY    262144   ROUND     -
pl plex-01      vol_mirr     DISABLED EMPTY    262144   STRIPE    8/128    RW
sd dsk0-01      plex-01      dsk0     0        32768    0/0       dsk0     ENA
sd dsk7-01      plex-01      dsk7     0        32768    1/0       dsk7     ENA
sd dsk1-01      plex-01      dsk1     0        32768    2/0       dsk1     ENA
sd dsk8-01      plex-01      dsk8     0        32768    3/0       dsk8     ENA
sd dsk2-01      plex-01      dsk2     0        32768    4/0       dsk2     ENA
sd dsk9-01      plex-01      dsk9     0        32768    5/0       dsk9     ENA
sd dsk3-01      plex-01      dsk3     0        32768    6/0       dsk3     ENA
sd dsk10-01     plex-01      dsk10    0        32768    7/0       dsk10    ENA
pl plex-02      vol_mirr     DISABLED EMPTY    262144   STRIPE    8/128    RW
sd dsk14-01     plex-02      dsk14    0        32768    0/0       dsk14    ENA
sd dsk21-01     plex-02      dsk21    0        32768    1/0       dsk21    ENA
sd dsk15-01     plex-02      dsk15    0        32768    2/0       dsk15    ENA
sd dsk22-01     plex-02      dsk22    0        32768    3/0       dsk22    ENA
sd dsk16-01     plex-02      dsk16    0        32768    4/0       dsk16    ENA
sd dsk23-01     plex-02      dsk23    0        32768    5/0       dsk23    ENA
sd dsk17-01     plex-02      dsk17    0        32768    6/0       dsk17    ENA
sd dsk24-01     plex-02      dsk24    0        32768    7/0       dsk24    ENA
 

  6. Add a DRL plex to the volume, if possible, specifying a disk that is not used by one of the data plexes; for example: 

    # volassist addlog vol_mirr dsk4

    The volume looks similar to the following: 

    Disk group: rootdg
 
V  NAME         USETYPE      KSTATE   STATE    LENGTH   READPOL   PREFPLEX
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
v  vol_mirr     fsgen        DISABLED EMPTY    262144   ROUND     -
pl plex-01      vol_mirr     DISABLED EMPTY    262144   STRIPE    8/128    RW
sd dsk0-01      plex-01      dsk0     0        32768    0/0       dsk0     ENA
sd dsk7-01      plex-01      dsk7     0        32768    1/0       dsk7     ENA
sd dsk1-01      plex-01      dsk1     0        32768    2/0       dsk1     ENA
sd dsk8-01      plex-01      dsk8     0        32768    3/0       dsk8     ENA
sd dsk2-01      plex-01      dsk2     0        32768    4/0       dsk2     ENA
sd dsk9-01      plex-01      dsk9     0        32768    5/0       dsk9     ENA
sd dsk3-01      plex-01      dsk3     0        32768    6/0       dsk3     ENA
sd dsk10-01     plex-01      dsk10    0        32768    7/0       dsk10    ENA
pl plex-02      vol_mirr     DISABLED EMPTY    262144   STRIPE    8/128    RW
sd dsk14-01     plex-02      dsk14    0        32768    0/0       dsk14    ENA
sd dsk21-01     plex-02      dsk21    0        32768    1/0       dsk21    ENA
sd dsk15-01     plex-02      dsk15    0        32768    2/0       dsk15    ENA
sd dsk22-01     plex-02      dsk22    0        32768    3/0       dsk22    ENA
sd dsk16-01     plex-02      dsk16    0        32768    4/0       dsk16    ENA
sd dsk23-01     plex-02      dsk23    0        32768    5/0       dsk23    ENA
sd dsk17-01     plex-02      dsk17    0        32768    6/0       dsk17    ENA
sd dsk24-01     plex-02      dsk24    0        32768    7/0       dsk24    ENA
pl vol_mirr-01  vol_mirr     DISABLED EMPTY    LOGONLY  CONCAT    -        RW
sd dsk4-01      vol_mirr-01  dsk4     0        65       LOG       dsk4     ENA
 

  7. Start the LSM volume; for example: 

    # volume start vol_mirr

    The volume is ready for use.

4.4.3    Creating a RAID 5 Plex with Disks on Different Buses

You can improve performance for a volume with a RAID5 plex by striping the plex over disks on different buses.

Before you begin, decide which LSM disks you want to use, identify which bus each disk is on, and plan how you want to create the volume based on how you want LSM to stripe the RAID 5 data plex over the buses.

Each column of subdisks should be the same size and be a multiple of the data unit size. For example, a data unit size (stripe width) of 16K bytes for a RAID5 plex corresponds to 32 blocks (sectors), so the total of the subdisks in each column should be a multiple of 32.

If each column consists of one subdisk (the typical configuration), then the subdisk size should be a multiple of 32. If a column consists of two subdisks, each subdisk can be a different size as long as the total is a multiple of 32.

You can optionally identify a separate disk, ideally on a different bus, for the RAID5 log plex, or allow LSM to use any available disk.

Caution

If both the log plex and one column of the data plex are on the same bus and the bus fails, you lose the entire volume. If possible, put the log plex on a bus that does not support a disk used in the data plex.

The following procedure creates a RAID 5 volume using these disks and buses in the rootdg disk group:

Bus 1 Bus 2 Bus 3 Bus 4 Bus 5
dsk1 dsk4 dsk8 dsk11 dsk22 (for the RAID 5 log plex)
dsk2 dsk5 dsk9 dsk12

The finished volume will stripe data and parity alternately over buses 1 through 4. The log plex will be on bus 5.

To create a RAID 5 plex with disks on different buses:

  1. Create the subdisks; for example: 

    # volmake sd dsk1-01 dsk1 len=1m
# volmake sd dsk2-01 dsk2 len=1m
# volmake sd dsk4-01 dsk4 len=1m
# volmake sd dsk5-01 dsk5 len=1m
# volmake sd dsk8-01 dsk8 len=1m
# volmake sd dsk9-01 dsk9 len=1m
# volmake sd dsk11-01 dsk11 len=1m
# volmake sd dsk12-01 dsk12 len=1m

  2. Create the RAID 5 data plex, specifying the subdisks in an order that alternates between the buses in the order you want; for example: 

    # volmake plex plex_r5 layout=raid5 stwidth=16k \
sd=dsk1-01,dsk4-01,dsk8-01,dsk11-01,dsk2-01,dsk5-01, \
dsk9-01,dsk12-01

    This creates an eight-column RAID5 data plex named plex_r5, where the columns proceed from bus 1 through bus 4, then repeat.

    The plex looks similar to the following: 

    Disk group: rootdg
 
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
pl plex_r5      -            DISABLED -        14336    RAID      8/32     RW
sd dsk1-01      plex_r5      dsk1     0        2048     0/0       dsk1     ENA
sd dsk4-01      plex_r5      dsk4     0        2048     1/0       dsk4     ENA
sd dsk8-01      plex_r5      dsk8     0        2048     2/0       dsk8     ENA
sd dsk11-01     plex_r5      dsk11    0        2048     3/0       dsk11    ENA
sd dsk2-01      plex_r5      dsk2     0        2048     4/0       dsk2     ENA
sd dsk5-01      plex_r5      dsk5     0        2048     5/0       dsk5     ENA
sd dsk9-01      plex_r5      dsk9     0        2048     6/0       dsk9     ENA
sd dsk12-01     plex_r5      dsk12    0        2048     7/0       dsk12    ENA
 

  3. Create the LSM volume using the data plex; for example: 

    # volmake -U raid5 vol vol5 plex=plex_r5

    This creates an LSM volume named volr5 from plex plex_r5, with a usage type of raid5 (required for all volumes with a RAID5 plex).

  4. Add a RAID 5 log plex to the volume, optionally specifying the disk (by default, LSM uses a disk not already used in the volume, if available); for example: 

    # volassist addlog volr5 dsk22

  5. Start the LSM volume; for example: 

    # volume start volr5

    The volume is ready for use and looks similar to the following: 

    Disk group: rootdg
 
V  NAME         USETYPE      KSTATE   STATE    LENGTH   READPOL   PREFPLEX
PL NAME         VOLUME       KSTATE   STATE    LENGTH   LAYOUT    NCOL/WID MODE
SD NAME         PLEX         DISK     DISKOFFS LENGTH   [COL/]OFF DEVICE   MODE
 
v  volr5        raid5        ENABLED  ACTIVE   14336    RAID      -
pl plex_r5      volr5        ENABLED  ACTIVE   14336    RAID      8/32     RW
sd dsk1-01      plex_r5      dsk1     0        2048     0/0       dsk1     ENA
sd dsk4-01      plex_r5      dsk4     0        2048     1/0       dsk4     ENA
sd dsk8-01      plex_r5      dsk8     0        2048     2/0       dsk8     ENA
sd dsk11-01     plex_r5      dsk11    0        2048     3/0       dsk11    ENA
sd dsk2-01      plex_r5      dsk2     0        2048     4/0       dsk2     ENA
sd dsk5-01      plex_r5      dsk5     0        2048     5/0       dsk5     ENA
sd dsk9-01      plex_r5      dsk9     0        2048     6/0       dsk9     ENA
sd dsk12-01     plex_r5      dsk12    0        2048     7/0       dsk12    ENA
pl volr5-01     volr5        ENABLED  LOG      2560     CONCAT    -        RW
sd dsk22-01     volr5-01     dsk22    0        2560     0         dsk22    ENA
 

4.5    Configuring File Systems to Use LSM Volumes

After you create an LSM volume, you can use it the same way you can use a disk partition. Because LSM uses the same interfaces as disk device drivers, you can specify an LSM volume in any operation where you can specify a disk or disk partition.

The following sections describe how to configure AdvFS and UFS to use an LSM volume.

4.5.1    Configuring AdvFS Domains to Use LSM Volumes

AdvFS treats LSM volumes as it does any other storage device. (For more information on creating an AdvFS domain, see the AdvFS Administration manual.)

Note

If an AdvFS domain needs more storage, you can either create a new LSM volume and add it to the domain using the AdvFS addvol command or grow an LSM volume that is already part of the domain (Section 5.4.9).

AdvFS domains can use a combination of LSM volumes and physical storage. For more information, see the AdvFS Administration manual.

4.5.2    Configuring UFS File Systems to Use LSM Volumes

To configure a UFS file system to use an LSM volume:

  1. Create a file system using the LSM disk group and volume name: 

    # newfs [options] /dev/rvol/disk_group/volume

    The following example creates a UFS file system on an LSM volume named vol_ufs in the dg1 disk group: 

    # newfs /dev/rvol/dg1/vol_ufs

    You do not need to specify the name of the disk group for LSM volumes in the rootdg disk group.

    For more information, see newfs(8).

  2. Use the LSM block special device name to mount the file system: 

    # mount /dev/vol/disk_group/volume /mount_point

    For example: 

    # mount /dev/vol/dg1/vol_ufs /mnt2

4.6    Creating LSM Volumes for Existing Data

You can use LSM to manage existing data by encapsulating the disk or disk partition containing the data. LSM converts the disk or partition to an LSM nopriv disk and creates an LSM volume from the encapsulated disk or disk partition.

You can encapsulate:

4.6.1    Encapsulating Disks or Disk Partitions

Encapsulating existing data is a two-part process: you create the encapsulation scripts with the volencap command, then run the volreconfig command to execute those scripts.

The encapsulation procedure configures the named disks and disk partitions as LSM nopriv disks, using information in the disk label and in the /etc/fstab file, and creates a separate LSM volume from each nopriv disk. By default, the nopriv disk and volume will belong to the rootdg disk group, unless you specify a different disk group.

If you encapsulate an entire disk (by not specifying a partition letter), such as dsk3, LSM creates a nopriv disk and a volume for each in-use partition.

Encapsulation provides a way to put existing data under LSM control, which lets you use LSM to mirror the data in the volume and thereby provide redundancy and high availability. However, recovery of a failed LSM nopriv disk can be complex, and there are other cases where nopriv disks complicate matters. After you encapsulate a domain, immediately move the volumes to LSM sliced or simple disks in the disk group if possible, before mirroring or performing other operations on the volumes.

After the encapsulation, entries in the /etc/fstab file or in the /etc/sysconfigtab file are changed to use the LSM volume name instead of the block device name of the disk or disk partition.

To encapsulate a disk or disk partition:

  1. Back up the data on the disk or disk partition to be encapsulated.

  2. Unmount the disk or partition or take the data off line. If you cannot unmount the disk or partition or take the data off line, LSM must restart the system to complete the encapsulation procedure.

  3. Create the LSM encapsulation script: 

    # volencap [-g disk_group] {disk|partition}

    The following example creates an encapsulation script for a disk named dsk3: 

    # volencap dsk3

    Note

    Although you can encapsulate several disks or disk partitions at the same time, we recommend that you encapsulate each disk or disk partition separately.

  4. Complete the encapsulation process: 

    # volreconfig

    If the encapsulated disk or disk partition is in use, the volreconfig command queries you whether to restart the system now or later.

  5. Optionally (but recommended), move the volume to sliced or simple disks in the same disk group before mirroring the volume or performing other operations. See Section 5.1.5.

4.6.2    Creating LSM Volumes for AdvFS Domains

You can place the storage for an existing AdvFS domain under LSM control by either encapsulating the domain or migrating the domain to an LSM volume.

No mount point changes are necessary during encapsulation or migration, because the mounted filesets are abstractions to the domain. The domain can be activated normally after the encapsulation or migration process completes. After the domain is activated, the filesets remain unchanged and the result of encapsulation or migration is transparent to AdvFS domain users.

4.6.2.1    Encapsulating an AdvFS Domain

Encapsulation provides a way to put existing data under LSM control, which lets you use LSM to mirror the data in the volume to provide redundancy and high availability. However, encapsulation creates a nopriv disk, and recovery of a failed nopriv disk can be complex, and there are other cases where nopriv disks complicate matters. After you encapsulate a domain, immediately move the volumes to LSM sliced or simple disks in the disk group if possible, before mirroring or performing other operations on the volumes.

To encapsulate an AdvFS domain:

  1. Back up the data in the AdvFS domain with the vdump utility.

  2. Unmount all filesets.

    If the domain is in use (you cannot unmount the filesets), you can create the encapsulation script (step 3) and run volreconfig (step 4) when convenient to complete the encapsulation procedure.

  3. Create the LSM encapsulation script: 

    # volencap domain

    The following example creates an encapsulation script for an AdvFS domain named dom1: 

    # volencap dom1

  4. Complete the encapsulation procedure: 

    # volreconfig

    If the AdvFS domain is mounted, the volreconfig command prompts you to restart the system.

    The /etc/fdmns directory is updated on successful creation of LSM volumes.

  5. Optionally (but recommended), move the volume to sliced or simple disks in the same disk group before mirroring the volume or performing other operations. See Section 5.1.5.

4.6.2.2    Migrating an AdvFS Domain

The volmigrate command lets you migrate any AdvFS domain (except for the root_domain on a standalone system) to an LSM volume. This operation uses a different disk than the disk the domain uses originally and therefore does not require a restart.

The volmigrate command creates a volume with the properties you specify, including:

There must be sufficient LSM disks in the same disk group, and the disks must be large enough to contain the domain. For more information on disk requirements and the options for striping and mirroring, see volmigrate(8).

To migrate a domain to an LSM volume: 

# volmigrate [-g disk_group] [-m num_mirrors] [-s num_stripes] \
domain disk_media_name...

The volmigrate command creates a volume with the specified characteristics, moves the data from the domain to the volume, removes the original disk or disks from the domain, and leaves those disks unused. The volume is started and ready for use, and no restart is required.