![]() |
![]() |
[ | | | ]
This chapter describes how to:
HSM-created stub files on your space-managed file systems are bound to the space-managed file system itself. Because of this, you cannot perform the following:
Before you can migrate files to Tivoli Storage Manager storage, first mount your file systems either automatically or manually, and then add space management.
For AIX JFS file systems: Mount your file systems twice. Enter these commands to mount them:
mount /fs mount -v fsm /fs
The first mount is for the physical AIX JFS file system. The second mount (with type fsm) is for the space management file system.
Mount the file system migrator (FSM) over the AIX JFS file system before you add space management services. The file system migrator is a kernel extension that intercepts all file system operations and provides any space management support that you require. If you do not mount the FSM, anyone with permission to access migrated files can access and modify your stub files. If a stub file is modified, the HSM client cannot access the migrated file.
After you add space management to an AIX JFS file system, do not unmount the FSM and directly access the native file system. If you unmount the FSM from a file system, enter the following command to remount it:
mount -v fsm /fs
The /etc/filesystems file contains a stanza for each file system to which you add space management. The HSM client updates these stanzas so that the file system and the FSM both mount automatically whenever you restart your system.
Add space management to all file systems for which you need space management services. When you add space management to your file system, the HSM client:
For Solaris VxFS file systems: If the /etc/vfstab file contains an entry for your file system, a new attribute is not added.
For AIX JFS file systems: If the /etc/filesystems file contains a stanza for your file system, the HSM client performs the following:
For AIX GPFS and Linux86 GPFS file systems: The dsmwatchd daemon starts at system startup (inittab entry). Ensure that DMAPI is enabled on all GPFS file systems that the HSM client manages. Enter the following command to query this information:
For AIX GPFS: /usr/lpp/mmfs/bin/mmlsfs DevicePath -z For Linux86 GPFS: /opt/lpp/mmfs/bin/mmlsfs DevicePath -z
If it is required, change the value to:
For AIX GPFS: /usr/lpp/mmfs/bin/mmchfs DevicePath -z yes For Linux86 GPFS: /opt/lpp/mmfs/bin/mmchfs DevicePath -z yes
Attention: During the mount process and while the HSM client is adding space management to your file systems, do not attempt to access any files in your file systems or perform any tasks against your file systems.
When you add space management to your file systems, select space management settings that will control the following:
This information is stored in the dsmmigfstab file located in config/dsmmigfstab. You can modify the settings in the dsmmigfstab file using the dsmmigfs command (see dsmmigfs). You can also use the dsmmigfs command to display the settings in this file (see dsmmigfs).
At any time after you add space management to your local file systems, you can update the settings, if necessary.
Copy several files into your migrated file system after you add space management and run the dsmmigrate command. If you are running open registration, the command prompts you for your node password and contact information the first time that you run it.
The following sections provide information to help you select space management settings for your file systems.
The HSM client does not migrate a file unless doing so saves space on your local file system. For AIX JFS file systems, before a file is considered for migration, the size must be greater than both the stub file size plus one byte and the file system fragment size. If the minmigfilesize option is defined, the file must be greater than this option value. For example, if the stub file size is 2047 bytes, and the defined block size for a file system is 4096 bytes, the file size must be greater than 4096 bytes before it is eligible for migration.
For AIX GPFS, Linux86 GPFS, Solaris VxFS, and HP-UX file systems, the file size must be greater than both the stub file size and the file system block size before it is eligible for migration.
The threshold percentages that you set for your file system affect when threshold migration starts and stops. A high threshold determines when threshold migration starts. A low threshold determines when file migration stops.
Specify a value of zero through 100 percent. The default for a high threshold is 90 percent. The default for a low threshold is 80 percent. For example, if you allocate 10 GB for a file system, and you must maintain at least 1 GB of free space, set the high threshold to 90 percent. If space usage equals or exceeds 90 percent when the HSM client checks space usage on your file system, files automatically begin migrating to Tivoli Storage Manager storage. The HSM client migrates files beginning with the first file that is listed in the current migration candidates list for your file system.
The percentage that you specify for a low threshold must be the same as, or lower than, the percentage that you specify for a high threshold. For example, to stop migrating files when there are 20 MB of available free space on your file system, set the low threshold to 80 percent.
If there are no additional candidates in the migration candidates list after threshold migration starts, and if space usage drops below the high threshold that you set, threshold migration stops. The dsmscout daemon builds a new migration candidates list when candidates are available on your file system. Threshold migration starts again the next time your file system exceeds the high threshold.
The premigration percentage controls premigration of additional files after threshold or demand migration completes. The names of any migrated files are removed from the current migration candidates list. To premigrate the next files that are listed in the migration candidates list, copies of the files are sent to Tivoli Storage Manager storage, and the original files remain as premigrated files on your local file system.
The default for the premigration percentage is the difference between the percentage that you specify for the low threshold and the percentage that you specify for the high threshold. The default percentage premigrates enough files to make the next occurrence of threshold migration faster. For example, if the high threshold is 90 percent and the low threshold is 80 percent, the premigration percentage is 10 percent. When space usage drops to 80 percent, additional files premigrate until at least 10 percent of the occupied space on your file system contains premigrated files that are listed at the beginning of the current migration candidates list. The next time threshold migration is required, the HSM client replaces those files with stub files on your local file system. This quickly reduces space usage to a low threshold without requiring additional time to copy the files to Tivoli Storage Manager storage.
If demand migration is required, having your eligible files already premigrated hastens that process as well. The HSM client quickly releases at least ten percent of the space on your local file system, and migrates any additional files that are necessary to return space usage to the low threshold.
During premigration, the HSM client skips any files that are premigrated and premigrates only those files that are required to reach the premigration percentage. Increase or decrease the premigration percentage when you want to change that percentage.
If the percentage that you specify for the low threshold is the same as the percentage for the high threshold, the default premigration percentage is zero. The HSM client does not premigrate any files after threshold or demand migration completes.
If the premigration percentage equals or exceeds the percentage that you specify for the low threshold, the HSM client premigrates all remaining files in your file system that are currently eligible for automatic migration.
Set the age and size factors to determine the order in which eligible files migrate to Tivoli Storage Manager storage during automatic migration. The age factor determines how much consideration is given to the number of days since a file was last accessed. The size factor determines how much consideration is given to the size of the file. Files are prioritized for automatic migration in descending order by their priority score. Set a value from zero through 999999999 for both age and size factors. The default is one for both factors.
Table 12 displays how to prioritize your files for migration.
If a resident file and a premigrated file have the same priority score, the
HSM client assigns the premigrated file a higher priority.
Table 12. Setting the Age and Size Factors
Prioritize Your Files | Set the Factors |
---|---|
Prioritize your files for migration by age only. | Set the age factor to one and the size factor to zero. |
Prioritize your files for migration by size only. | Set the age factor to zero and the size factor to one. |
Prioritize your files for migration according to their age. | Increase the age factor. |
Prioritize your files for migration according to their size. | Increase the size factor. |
Prioritize your files for migration in order by age, and within age, by size. | Set the size factor to one, and set the age factor to the size of the largest file that you want to store, or to the size (in 1 KB blocks) of the file system itself. The HSM client assigns the highest priority to the oldest files. If two or more files are the same age, files are prioritized by size. |
Prioritize your files for migration in order by size, and within size, by age. | Set the age factor to one, and set the size factor to the number of days equal to the longest life span of a file on your file system. The HSM client assigns the highest priority to the largest files. If two or more files are the same size, files are prioritized by age. |
The quota that you set determines the maximum number of megabytes of data that you can migrate and premigrate from your file system to storage. Specify a quota value from zero through 999999999. The default is the number of megabytes that are assigned for your file system. For example, if 20 gigabytes are assigned for your file system, the HSM client migrates and premigrates your files from that file system until the total number of megabytes that migrate and premigrate equals 20 gigabytes.
When files premigrate, they use space on both your local file system and in storage. When files migrate, stub files use some of the space on your local file system.
If you set the quota to zero for your file system, files do not migrate to storage. Set the quota for your file system to a value that is large enough to accommodate projected growth. Check with your Tivoli Storage Manager administrator to determine whether there are any restrictions on the amount of data that you can migrate and premigrate to storage.
Because dsmscoutd does not perform reconcilations, this is the maximum number of migration candidates that the dsmscout daemon searches for during a scan of your file systems. When this number is reached, the dsmscout daemon stops. The range of values that you can enter are zero through 9999999. The default is 10000.
A stub file contains information that is necessary to locate and recall a migrated file. It can contain leading bytes of data called leader data from your original file. If you access only this data but you do not modify it, the migrated file is not recalled from Tivoli Storage Manager storage. Storing leader data in stub files is especially useful if you frequently run one or more programs that read only the information located at the beginning of a large number of files.
For HSM clients on AIX JFS file systems, the minimum stub file size is 511 bytes and the default size is 4095. You can select a stub file size that is smaller than your file system block size. However, this is an inefficient use of space on your file system. Although the HSM client uses only the number of bytes that you specify for stub files, your operating system allocates one block for each stub file. The excess number of bytes remain unused.
For HSM clients on Solaris VxFS and HP-UX VxFS file systems, valid stub file sizes are a multiple of the physical blocksize. For HSM clients on AIX GPFS and Linux86 GPFS file systems, valid stub file sizes are zero, or a multiple of the physical blocksize. The minimum physical block size is 512 bytes. The default logical block size is 4096 bytes.
When selecting a stub file size, consider the advantage of storing additional leader data (files are not recalled if only leader data is accessed and not modified), over the disadvantage of using additional space on your local file system for stub files (larger stub files consume more storage space on your local file system).
To add space management to one or more file systems, follow these steps:
These file systems contain files that your operating system frequently uses.
To add space management to your local file systems from the command line, use the dsmmigfs command. For example, to add space management to the /home file system and set a high threshold of 85 percent and a low threshold of 75 percent, enter the following:
dsmmigfs add -ht=85 -lt=75 /home
See dsmmigfs for more information about this command.
A nested file system is a file system mount point that is contained within another file system. For example:
/test /test/migfs1
The /test file system is a parent file system and /test/migfs1 is a nested file system within /test. They are both mount points.
To add space management to a parent file system and any nested file systems, follow these steps:
To add space management to a parent file system but not to its nested file system, perform steps one through three above.
For AIX JFS file systems only
For an AIX JFS-nested file system that mounts automatically when you restart your system, follow these steps to mount the parent file system before you mount the nested file system:
mount /test/migfs1
Where /test/migfs1 is the name of the nested file system.
To add space management to a file system that the NFS server exports, follow these steps:
/usr/sbin/showmount -a
/usr/etc/exportfs -a
After you add space management, you can manage your file systems easily. This section describes how to:
Changes to space management settings take effect in the following manner:
Update space management settings for your file system at any time. Space management can be active, inactive, or globally inactive for a file system. To update space management settings for your file system, follow these steps:
To update space management settings for your file system from the command line, use the dsmmigfs command. For example, to change the age factor to five and the size factor to one for the /home file system, enter:
dsmmigfs update -a=5 -si=1 /home
See dsmmigfs for more information about this command.
Deactivate space management from your file systems to temporarily prevent migration, recall, or reconciliation processes from occurring. Any migration, recall, or reconciliation process that currently is in progress completes first. When you reactivate space management on your file system, all space management services resume.
Deactivating space management from your file system:
To deactivate space management from your file systems, follow these steps:
To deactivate space management from your file systems from the command line, use the dsmmigfs command. For example:
dsmmigfs deactivate /home
See dsmmigfs for more information about this command.
You can temporarily deactivate space management from all file systems on your client node. When you globally deactivate space management, you cannot perform migration, recall, or reconciliation on any file system. However, any migration, recall, or reconciliation process that is in progress completes before space management is globally deactivated.
Globally deactivate space management before you perform a system maintenance task or before your Tivoli Storage Manager administrator plans to export migrated files from one server and import them to another. When your Tivoli Storage Manager administrator completes the export and import processes, globally reactivate space management on the file systems that you previously deactivated.
Globally deactivating space management:
To globally deactivate space management, follow these steps:
To globally deactivate space management on your client node from the command line, enter:
dsmmigfs globaldeactivate
See dsmmigfs for more information about this command.
Once you deactivate space management from your file system, you can reactivate it at any time. To reactivate space management, follow these steps:
Use the dsmmigfs command to reactivate space management for your file systems from the command line. For example:
dsmmigfs reactivate /home
See dsmmigfs for more information about this command.
You can globally reactivate space management for your client node at any time. All file systems to which you added space management return to their previous states. For example, if you activated space management for a file system when you globally deactivated space management for your client node, it becomes active again. Or, if you deactivated space management for a file system when you globally deactivated space management for your client node, it remains inactive.
To globally reactivate space management for your client node, follow these steps:
Use the dsmmigfs command to globally reactivate space management for your client node. For example, enter:
dsmmigfs globalreactivate
See dsmmigfs for more information about this command.
Space management must be active on your file system to completely remove it. Ensure that you have space on your file system to recall all migrated files. Before you remove space management from your file system, ensure that all activity on your file system has stopped and that the file system is not being accessed.
On an AIX JFS workstation, if the file system is being accessed when you remove space management, the FSM cannot unmount from the file system, and the Tivoli Storage Manager backup-archive and space management functions will not work correctly. If the FSM does not unmount when you remove space management from your file system, enter the unmount command for the file system to unmount it after all processes no longer access your file system.
When you remove space management from your file system, the HSM client performs the following:
For AIX JFS file systems, if the /etc/filesystems file contains a stanza for the file system, HSM sets the mount attribute to true and removes the adsmfsm attribute.
For Solaris VxFS file systems, if the /etc/vfstab file contains an entry for the file system, the HSM client updates only the dsmmigfstab file.
To remove space management from one or more file systems, follow these steps:
Use the dsmmigfs command to remove space management. For example, to remove space management from the /home file system, enter:
dsmmigfs remove /home
See dsmmigfs for more information about this command.