TSM V6.4 UNIX and Linux Backup-Archive Client known problems and limitations

If you back up or archive data from a Tivoli Storage Manager Version 6.4 client, you cannot restore or retrieve that data using a Tivoli Storage Manager Version 6.3 or earlier client.

For more information about Tivoli Storage Manager Server/Client compatibility and upgrade considerations, go to http://www.ibm.com/support/docview.wss?&uid=swg21053218.

NAS backups performed on a Tivoli Storage Manager Version 6.3 server using the BACKUP NODE server command can only be restored by using the RESTORE NODE server command or a Tivoli Storage Manager Version 6.3 or later client.

Back to Contents

Common UNIX and Linux warnings

For option and input files in DBCS locales, ensure that all TSM client option and input files (such as dsm.opt, dsm.sys, and filelist) are created in the same SBCS locale as the TSM client.

Back to Contents

Mac OS X warnings

Case Sensitive HFS+ support

Mac OS X supports an optional Case Sensitive HFS+ file system. Before backing up any files in that file system ensure the new Case Sensitive HFS+ volume does not have the same name as any existing TSM volumes for the node. Either the TSM volumes can be renamed on the TSM server, or the CS HFS+ volume can be renamed locally.

Back to Contents

Tivoli Storage Manager Client limitations and known problems

Common Tivoli Storage Manager Client known problems and limitations

When the user attempts to modify an existing node access rule from the JGUI (Utilities -> Node Access List), the predefined path is not selected in the tree. (Internal reference #94088)
Java GUI: Adding Access Rule to the Node Access List fails when the file is not under a directory but in the root of the file system. (Internal reference #94256)

Workaround:

If the passwordaccess generate is set in the client option file and if an invalid password is given at the first attempt during open registration with the client, the second attempt will fail as well.
(Internal reference #94896)

Workaround:
Retry open registration.
GPFS only: If a file is moved from one storage pool to another, the attributes of the files change and the ctime will be set. The ctime change causes a new backup of the file in the next backup run. (Internal reference #93633)
The 'capicmd' GSKit application (gsk8capicmd.exe, gsk8capicmd_64.exe on Windows; gsk8capicmd, gsk8capicmd_64 on Unix) used to create SSL key database files fails with the "File already exists" error if one of the files of the key database has been deleted. A workaround for this issue is to delete all key database files and re-try key generation. Key database files match the following pattern: "dsmcert.*". The recommended way to delete the key database is to use following command:

"gsk8capicmd -keydb -delete -db dsmcert.kdb -pw password" (Internal reference 5571)

Initial configuration wizard reports a protocol violation

During a first time configuration, the configuration wizard will report a protocol violation. Additionally, the Domain panel will not list the file systems to select for the domain option.

The error dialog can be dismissed and the configuration wizard can be used without further error.

After the intial configuration, use the preference editor to set the Domain option.

Interrupt with CTRL-C

During a command line client operation, pressing CTRL-C might result in a TSM client program exception or other unexpected behavior. To abort a command line client operation, press the 'Q' key instead of CTRL-C.

Web Client tree expansion

When using the Web Client, the browser can crash if you do the following, using the view menu item in the backup or restore tree window:

Click on "Expand Entire Branch"
Then click on "Collapse Entire Branch"
Web Client font requirements for non-English file names

Logfile out of space handling

If any log file (dsmerror.log, dsmsched.log, or dsmwebcl.log) runs out of space during a session, writing to that log ceases, but other processing continues. End of processing return codes will reflect all errors and conditions, not just those we were able to log.

To prevent this problem, set the ERRORLOGMAXSIZE (for dsmerror.log) or SCHEDLOGMAXSIZE (for dsmsched.log and dsmwebcl.log) options to limit the log size to available space. Note that using these options causes the log data that would exceed the maximum to be written at the beginning of the log, overwriting the oldest entries.

ERRORLOGMAX and SCHEDLOGMAX behavior in out-of-space conditions:

ANS1521E Failure writing to a Tivoli Storage Manager log or log-related file:

<LOG FILE NAME>

, errno = 28, There is not enough space in the file system

This is correct behavior. However, the log header record might be incomplete or there may be no "END OF DATA" text marker at the end of the error log. After space has been made available, the TSM client will subsequently treat the log as unwrapped because a valid header record is not found. A new log will be created and this partial log will be written to the prune file.

If there is insufficient space in the file system to append an entry to the log, the TSM client will continue to run, but the error message will not be logged.

Some keywords have not been translated into Chinese for the backup-archive client:

Some keywords, such as "Snapshot", have not been translated into Chinese. This is most notably evident in the backup-archive client GUI.

If client encryption is used and you choose to abort a command line backup or archive operation when prompted for an encryption key, the entire operation will immediately end with return code (RC) 12, even if there are other files eligible for backup that do not require encryption.
When regional settings are updated in the Java GUI preference editor, the GUI needs to be restarted before the changes become effective.
There are minor errors displaying ANS1036S in foreign languages: On all platforms Polish has an extra period on the second line, and on several platforms the format is incorrect.
Handling spaces in file names in schedule definitions

objects

options

objects

For UNIX and Linux clients:

This will ensure that /home/proj1/Some file.doc is treated as a single file name, as opposed to two separate files (/home/proj1/Some, and file.doc). Some examples of options parameter usage:

options='-preschedulecmd="/home/me/my files/bin/myscript" -postschedulecmd="/home/me/my files/bin/mypostscript" -quiet'
options='-presched="/home/me/my files/bin/precmd" -postsched=finish'

For Windows clients:

objects=’"c:\home\proj1\Some file.doc"’
objects=’"c:\home\proj1\Some file.doc" "c:\home\Another file.txt" c:\home\noblanks.txt’
objects=’"c:\home\My Directory With Blank Spaces\"’
objects=’"c:\Users\user1\Documents\Some file.doc"’
objects=’"c:\Users\user1\Documents\Some file.doc" "c:\Users\user5\Documents\ Another file.txt" c:\Users\user3\Documents\noblanks.txt’ objects=’"c:\Users\user1\My Directory With Blank Spaces\"’

This will ensure that c:\home\proj1\Some file.doc is treated as a single file name, as opposed to two separate files (c:\home\proj1\Some, and file.doc). Some examples of options parameter usage:

options='-preschedulecmd="c:\home\me\my files\bin\myscript" -postschedulecmd="c:\home\me\my files\bin\mypostscript" -quiet'
options='-presched="c:\home\me\my files\bin\precmd" -postsched=finish'

You can also refer to the objects and options parameter information for the define schedule and update schedule commands in the appropriate IBM Tivoli Storage Manager Administrator’s Reference.

Java GUI does not start with invalid option in options file

In rare cases, when there is an error in the dsm.opt file, the Java GUI will not start properly. The Java GUI will attempt to start, but at some point after 70%, the title screen disappears and nothing appears to happen.

To work around the problem, refer to the dsmerror.log or dsmj.log file in the installation directory or from where the command to start the Java GUI was issued for the offending option, and remove it manually from the options file. You should be able to restart the client without problems.

Reset button does not function properly in Preference Editor Communications Panel

On certain systems, the Communications panel in the Preference editor will not display correctly when the Reset button of the preference editor is invoked. When you click the Reset button, the list of communication methods might not display the proper list after the reset is complete.

If there is a need to reset the communication panel back to the original settings, exit out of the preference editor and restart it.

IMAGEGAPSIZE Option

The value for the IMAGEGAPSIZE option should not be specified outside the listed ranges for MB and GB units. If it is, the client will not report it as an error.

Select all after applying filter from Backup window

After applying a filter to a directory in the Backup window, when you check the check box for this filtered directory, the corresponding files in the file view on the right are not checked (not selected). To select the files in that filtered directory, click on the name of the checked directory and the boxes next to the filtered files are checked.

Connection Information dialog might not appear in certain situations

The Connection Information dialog might not appear correctly in certain situations. An exception will be logged in the dsmj.log file when this occurs. This will occur when you do the following:

Dsmagent can crash if it is killed during restore operation

On Unix and Mac platforms, when TSM GUI (both Java and Web) is used to perform restore or retrieve operation and dsmj or dsmagent process is killed while the operation is still in progress, dsmagent can crash. There is no negative impact other than creating a core file.

The recommended ways to terminate an application while an operation is in progress are either pressing the Stop button or cancelling the session by TSM administrator.

Back to Contents

Common UNIX and Linux known problems and limitations

The -absolute option does not work while doing a Journal base backup. The option is accepted, but all files are not backed up. The user will see this error in the dsmerror.log:

ANS7559E The absolute option requires specifying the NoJournal option when performing a Journal Based Backup for backing up fs \\xxxxx\x$.

(Internal reference #95289)

Workaround:
Add the -nojournal option to the command.
NFSTIMEOUT option

This known problem applies to all UNIX and Linux platforms except for Mac OS. The nfstimeout option can fail if the NFS file system is hard mounted. If a hang occurs, remove the nfstimeout option from your options file and soft mount the NFS file system as follows:

mount -o soft,timeo=5,retry=5 machine:/filesystem /mountpoint

The parameters are defined as follows:

soft: Generates a soft mount of the NFS file system. If an error occurs, the stat() function returns with an error. If the option hard is used, stat() does not return until the file system is available.
timeo=n Sets the time out for a soft mount error to n seconds
retry=n Set the internal retries and the mount retries to n, the default is 10000.

TSM editor in various terminal environments

The TSM interactive-mode command line editor is designed to work in the following terminal environments:

xterm --- This is the default if terminal type not recognized.
xterm (Linux KDE)
aixterm
dtterm
VT-100

Client interoperability

Data that has been backed up from Mac OS versions prior to Mac OS X will not have the correct file owner or permissions when restored on Mac OS X. After the restore is complete, use the "sudo chown" and "sudo chmod" commands to set them.
Files backed up or archived by the z/OS Unix System Services client cannot be restored or retrieved by any other TSM Unix client. Files backed up or archived by any other TSM Unix client cannot be restored or retrieved by the z/OS Unix System Services client.
Files backed up or archived with DES-56 encryption on Intel platforms such as Linux86 cannot be restored or retrieved on non-Intel platforms (for example, AIX), and vice versa. AES-128 is the recommended encryption method.

File names containing characters with code > 127 (does not apply to Mac OS)

If you want to back up files with names containing characters with a code > 127, ensure that you have chosen either an SBCS character set locale or a DBCS character set locale corresponding to the encoding of the file names. The default code page C or the code page POSIX supports characters up to 127 only. Files whose names contain special characters that are not supported by the current locale will be skipped.

It is strongly recommended that you perform a system backup by using a SBCS character set to prevent any file or directory from being skipped. This action ensures that all files are backed up independent of their character set encoding.

Restore and retrieve using FOLLOWSYMBOLIC option

If you want to restore or retrieve to a destination path that contains a symbolic link to a directory, use the "FOLLOWSYMBOLIC Yes" option setting either during the operation or in your client options file (dsm.opt). This setting allows you to restore or retrieve the original directory tree underneath the destination path. Otherwise, you will get the ANS4029E error during restore or retrieve.

Option and input files in DBCS locales

If you create TSM client option and input files (such as dsm.opt, dsm.sys, and file lists) using different SBCS locales or using DBCS locales, the TSM client might not correctly process the file names in those files. You might receive error messages ANS1228E or ANS4005E.

Ensure that all TSM client option and input files, such as dsm.opt, dsm.sys, and file lists, are created in the same locale as the TSM client.

Java GUI and Web client known problems and limitations

The Web and Java GUI client will fail to restore a directory if a filespace with the same name exists on the Tivoli Storage Manager server. You might see the error message "No Objects on the server match query" or "ANS1395E The destination filespace or drive letter is unavailable. The following object is not processed: Filespace: '/'".

For example, '/opt' is a file system backed up by Tivoli Storage Manager. After reinstalling the machine, '/opt' is a directory below file system '/'. File system '/opt' no longer exists on the Tivoli Storage Manager client machine. The file system '/' and all of its subdirectories, including directory '/opt', is backed up by Tivoli Storage Manager. At this point there are both '/' and '/opt' file spaces on the Tivoli Storage Manager server for this Tivoli Storage Manager client node. Trying to restore files out of filespace '/', directory '/opt' fails using the Web/Java GUI (dsmj) client.

If you experience this problem, the workaround is to use the command line Tivoli Storage Manager client to restore the files. Put the file space name in curly brackets.
Example: dsmc restore "{/}opt/subdirectory/*" -subdir=yes

The maximum trace file size (tracemax) for the Java GUI preference editor is 2147483647 MB. This is a temporary workaround until the preference editor can accommodate larger numbers.
Web Client Security Exception

If you get a Security Exception error when running the Web Client in your browser due to the Web Client trying to open a TCP/IP socket to a socks server, disable the proxy server via your browser settings or Java Plugin control panel. See the "Starting a Web Client Session" section in Chapter 3 "Getting Started" of the "IBM Tivoli Storage Manager for UNIX and Linux Backup-Archive Clients Installation and User's Guide" for more information.

Web Client concurrent restore

Running more than one restore or retrieve operation at the same time from the Web Client might cause the browser to hang when destination or message windows from the different restore/retrieve processes appear on the screen at the same time. If this happens, close the Web Client browser windows, stop the TSM Agent service, then retry the operations using only one Web Client session.

Insufficient space in Javaheap (Java GUI)

If you receive the following messages from Java Virtual Machine (JVM) while running DSMJ:

JVMST109: Insufficient space in Javaheap to satisfy allocation request JVMDG217: Dump Handler is Processing OutOfMemory - Please Wait. JVMDG315: JVM Requesting Heap dump file

These errors indicate that your JVM maximum heap size is too small. The default value is 64M. To increase the maximum heap size, add the option -Xmxn (where n is size in MB) in the DSMJ script, such as the following example:

java -Xmx256m -DDSM_LANG=$LANG -DDSM_CONFIG=$DSM_CONFIG -DDSM_DIR=$DSM_DIR \ -DDSM_LOG=$DSM_LOG -DDSM_OPTIONS="$OPTIONS" -DDSM_ROOT=$PWD \ ${JAVA_XARGS} -jar dsm.jar

This example increases the maximum heap size to 256MB.

Search/Filter (Web and Java GUI)

The Search/Filter function might appear to be unresponsive when searching through very large file systems. This is caused by the Java Virtual Machine (JVM) running out of memory. You should see a "java.lang.OutOfMemoryError" in the Java console.

Problems with different JRE versions

There are several different Java Runtime Environment (JRE) versions available. Depending on the version and the operating system, there are some known problems related to JRE. See the Software Requirements section for the supported and required JRE version to be used with the TSM Java GUI for your operating system.
Some JRE versions have problems transferring the cursor focus to a component. For example, accessing shortcuts in the menu (e.g., pressing ALT-F to open the file menu) and combo-boxes in a modal dialog by clicking the mouse button might not get the cursor focus correctly. To solve this problem, transfer the cursor focus manually to that component by pressing the TAB key (or CTRL-TAB) several times.

Preference Editor limitations (Java GUI)

The domain list in the "Backup" tab of the preferences editor does not reflect any domains excluded via the dash operator after "domain ALL_LOCAL" in the domain statement of the client options file.
When any option is changed in the preferences editor of the Java GUI, the domain list is rewritten to the client options file.
When using the preferences editor to modify the Include-Exclude list, and the selected rule contains a space (for example, /test dir/inclexcl.txt), the rule will not be correctly displayed in the "Filename or Pattern" field. To work around this problem, you can either manually update this field via the "Browse" button, or manually type in the correct path. This will be fixed in a future release.
The Preferences Editor invoked from the UNIX Backup-Archive Client Java GUI uses the DSM_CONFIG variable instead of the passed parameter "optfile".
The warning in the Diagnostic tab is not removed from the filename box when tracing is disabled and an error exists.

Backup Window

When you select an item in the filelist tree of Backup window, the drop-down box for backup type will be permanently disabled. To re-enable this, exit the Backup window and reload the window.

Web client performance

Disable Java caching from the Java Control Panel to help improve performance on the Web client on the machine running the browser.

Client-side data deduplicaton memory requirements

Client-side data deduplication might require additional memory during backup or archive processing. You might need to increase the system limit on the size of the data area (segment). To check current setting of the limit, run the following command:

ulimit -d

ulimit -d <new value>

or ulimit -d unlimited

When using the TSM for Space Management (also known as UNIX HSM) client:

Back to Contents

Common AIX known problems and limitations

NQR (no query restore) does not work for HSM (space managed) file systems, when the RESTOREMIGSTATE option is set to YES. Only classical restore is possible in this case.
The Tivoli Storage Manager client will fail to perform snapshot-based image backup when mirror pool is enabled. The Tivoli Storage Manager client does not support mirror pools at the moment.
Snapshot difference incremental backups are not supported for vFiler volumes mounted using AIX NFS version 4. This problem is documented by NetApp BURT 630200. Apply the fix for this once it is available from NetApp. Specify "testflag snapdiffenablevfilernfs4" in dsm.opt file and retry the snapshot difference incremental backup.
If you perform snapshot difference incremental backup of N-Series or NetApp volumes, make sure that the volumes are defined with a security style of UNIX. TSM backups will complete successfully even if the security style is NTFS or MIXED. However, it is recommended that you use a security style of UNIX.
During snapshot difference incremental backup of N-Series or NetApp volumes, if files are skipped due to any reason, such as TSM server media mount being unavailable, these files may not be backed up during subsequent backups either, unless the files have been modified. Such files will be logged to the TSM error log. You may back them up manually or perform a snapshot difference backup with the option "createnewbase" set to "yes".
On AIX 6.1, if the EFS user keystore password is different from the user login password, the EFS keystore is not automatically opened when the user logs in. In such a situation, the TSM client might fail to restore a non-EFS file into an EFS file system. If this happens, the user can either launch the TSM client with the efskeymgr -o command or synchronize the keystore password with the user login password with the efskeymgr -n command. For example:

efskeymgr -o ./dsmj efskeymgr -n

When the NFS server corresponding to an NFS file system mounted on AIX is not reachable due to a network error or some other problem, the following error message is displayed on the console while performing TSM client operations:

NFS getattr failed for server

nfs_sever_name

: error 3 (RPC: 1832-006 Unable to send)

In order to prevent this error message from being displayed by the AIX operating system, unmount the NFS file system having the problem and retry the TSM client operation.

If you use the memoryefficient=diskcachemethod option for full incremental backup, you might need large amounts of disk space. Ensure that large file support is enabled on the file system that is being used for the disk cache file.
AIX 6.1 Encrypted File System (EFS) restore to an alternate destination might hang if the destination EFS file system runs out of disk space. To prevent this problem, apply the following AIX 6.1 APAR:

610 (6.1.0): IZ42302
61B (6.1.1): IZ42301
61D (6.1.2): IZ42300

Incremental by date backups on AIX 6.1 might back up files that have been previously backed up and are unchanged. Upgrade to the latest service pack for your AIX technology level as this problem has been resolved by AIX APAR IZ22211.
Resolving errors during AIX JFS2 snapshot based backup/archive and image backup:

The TSM client supports snapshot based backup and archive operations to provide a consistent backup of files and online image backup by exploiting the integrated snapshot capability provided by JFS2.

The TSM client deletes the AIX JFS2 snapshot created during the backup process during TSM termination. However, there are situations in which AIX may fail the snapshot delete request made by TSM.

The following are some cases when a snapshot delete request might fail:

Ctrl-C

dsmc backup image /fs1

/fs1

For example: dsmc incr /fs1/level1/dir1 from one console and dsmc incr /fs1/level2/level3/dir3 from a second console concurrently.

The reason for this is that AIX expects the snapshot delete requests to be issued in a particular order with the oldest snapshot deletion being requested first, the next oldest snapshot deletion being requested next, and so on. If TSM is not able to follow this sequence due to concurrent processes creating snapshots for the same file system, AIX will fail the delete requests. In the above situations, TSM will log a warning message asking the user to delete the snapshots manually. In some cases, TSM might even fail the backup. If this happens, the user will have to delete the snapshots manually and retry the operation.

To avoid these scenarios, do not start concurrent TSM snapshot operations on the same file system that may lead to snapshot deletion failures.

You can use the following steps to perform a manual snapshot deletion:

snapshot -q -c' '

srcFS

df -k

umount -f /tsmxxxxxxxxxx

rmdir /tsmxxxxxxxxxx

snapshot -d /dev/tsmxxxxxxxxxx

Device Busy

umount -f

srcFS

ls -l /dev/tsmxxxxxxxxxx

/dev/tsmxxxxxxxxxx

rmlv -f /tsmxxxxxxxxxx

mount

srcFS

If there are any snapshots that failed to be deleted during a previous TSM execution, TSM attempts to delete them during its next invocation. The reason for this is while older snapshots remain, AIX will fail deletion requests for newer snapshots for a given file system.

There are some cases where TSM will not attempt to delete older snapshots:

tsm

In all such cases, the user might have to perform a manual deletion since leaving unused older snapshots around will cause subsequent TSM backups to fail to delete snapshots.

NOTE: AIX has some defect fixes related to JFS2 snapshots in AIX 6.1 or later. If these fixes are not applied, it may cause either an AIX system crash or TSM to hang during snapshot deletion and snapshot query processes. It may also cause data corruption during used block image backup. Thus, TSM will not perform snapshot monitoring, the snapshot deletion feature described above, and used block image backup unless AIX is at the levels mentioned above. In order to exploit these features, ensure that your OS level is at AIX 6.1 or later.

Used Block Image backup:

The TSM client will perform used block image backup for both snapshot and static image backups by default, unless:

imagegapsize

AIX JFS2

Symbolic link as a virtual mount point:

Do not define symbolic links as virtual mount points, when performing snapshot based file backup or archive operations. In scenarios similar to the following, make sure that you set snapshotproviderfs=NONE or remove the snapshotproviderfs option.

Example:

/fs1 is a file system with a directory name dir1 that has one or more files and subdirectories.
Create a symbolic link to /fs1/dir1 under a second file system /fs2 as follows: ln -s /fs1/dir1 /fs2/dir2
In the dsm.opt file, set: followsymbolic yes
In the dsm.sys file, set:
snapshotproviderfs NONE
virtualmountpoint /fs2/dir2
Issue this command: dsmc incr /fs2/dir2
Performing snapshot backup of multiple virtual mount points for a given file system:

dsmc incr /fs1/level1/dir1 /fs1/level1/level2/dir2 /fs1/level1/level2/level3/dir3

The TSM client will take a single snapshot of file system /fs1 to back up all three virtual mount points.

In this case, make sure that you specify a single presnapshotcmd and a single postsnapshotcmd for the entire command as shown below:

dsmc incr /fs1/level1/dir1 /fs1/level1/level2/dir2 /fs1/level1/level2/level3/dir3 -presnapshotcmd=

pre-script

-postsnapshotcmd=

post-script

Using ssh localhost dsmc command:

When running ssh localhost dsmc command, the TSM client uses the server stanza defined in the dsm.opt file in the default installation directory, /usr/tivoli/tsm/client/ba/bin64 for the 64-bit client, ignoring the environment variable DSM_CONFIG, if it is set. Ensure that the dsm.opt file in the default installation directory has the correct server stanza when you invoke dsmc with ssh.

Restoring a file as a non-root TSM authorized user:

After restoring a file as a non-root TSM authorized user, the restored file is not owned by the TSM authorized user who restored the file, if the restore overwrites the original file.

Set up a non-root TSM authorized user, such as tsmuser.
Create a directory structure such as: /fs1/testdir with file file1.
Set the permissions on the directory testdir as 755 and owned by root, the permissions on the file file1 as 777 and owned by a different user than the TSM authorized user, say testusr.
Backup the testdir directory as the TSM authorized user: dsmc sel "/fs1/testdir/*" -su=y
Modify the contents of the original file file1
As the TSM authorized user, without removing the original file, restore the testdir directory to original location: dsmc rest "/svt1/testdir/*" -su=y -repl=y

tsmuser

file1

testusr

In order to avoid this problem, restore the file either as root or as the original owner of the file.

VxFS Sparse File Restore

When a sparse file is restored in the Veritas file system it can occupy more disk space than it did originally at the time of backup. This is a limitation caused by the inner workings of the space allocation algorithm used by the file system.

Because of the limited functionality of the dtterm application, not all function keys of the command line clients operate as expected. The Control-Left and Control-Right key combinations and the Home and End keys do not work. In the Aixterm environment the keys operate as specified.
Use NDMP directory level backup with valid filespace mapping.
If the COMMMethod or LANFREECommmethod options are set to SHAREdmem, the AIX environment variable EXTSHM needs to be turned on to enable extended shared memory. In order to turn on EXTSHM setting, do one of the following:

To make this a permanent change, add the following line to the /etc/environment file:

EXTSHM=ON

To make the change in the immediate shell, enter:

EXTSHM=ON export EXTSHM

Note:

Connecting to server SSL port by mistake

If a client connects to an SSL port on the server but the client does not have the SSL option enabled, the client would seem to hang until the session times out on the server.

The client does not have a way to tell whether the server is expecting an SSL session or not. If you forget to enable SSL on the client, it will try to connect using a regular (non-SSL) session to server's SSL port. The connection will "hang" because both the client and server will wait for expected responses from each other.

To avoid this situation, make sure that you set the SSL option to "yes" along with setting the TCPPort to server's SSL port.

Snapshot-based image backup of file systems or logical volumes in a big or scalable volume group

TSM Client will fail to perform snapshot-based image backup of two or more file systems or logical volumes in parallel for a big or scalable volume group. It is an AIX OS limitation. To avoid this situation, run snapshot-based image backups for big or scalable volume group sequentially, ensuring that the first backup is complete before starting the next one.

For example avoid invoking image backup as follows:

1. dsmc backup image /fs1 /fs2 /fs2

2. Start image backup for /fs1. Start image backup for /fs2 while the backup of /fs1 is still in progress.

This problem will be fixed in an upcoming AIX maintenance level with the APAR IZ74114.

During restore of NFS file system TSM client can failed to change modification or access time of the directories. And the following message can be displayed:

ANS4007E Error processing 'directory name': access to the object is denied ** Unsuccessful **

If the multi-line text contains symbols which occupy more than one column position (like Japanese or Chinese characters) TSM client command line can behave incorrectly when a character from any line except first is backspaced over. In this case the first line from bottom can be multiplied 3 times. Backspace works correctly during deletion of any character from the first line.

Back to Contents

AIX known problems and limitations

Problem with NFSV4 ACL
In the case of NFSV4 ACL, a change to the standard UNIX permissions will also change the ACLs. Since TSM stores ACLs along with the file data, a change to UNIX permissions will cause a corresponding change to NFSV4 ACLs resulting in the backup of the entire file. You can set the skipaclupdatecheck option to yes to avoid this problem.
The skipaclupdatecheck option disables checksum and size comparisons of ACL data.
When set to yes (default is no), the Tivoli Storage Manager client will not perform checksum and size comparisons before or after backup and during incremental processing (ACL checksum from previous backup and current ACL) to detect ACL updates.
However, current ACL data will be backed up if the file is selected for backup due to other reasons. If only ACLs are updated on a file, the next incremental backup will not recognize this ACL update, and the file will not be backed up.
Journaled limitation
On AIX, it is possible to create path names greater than the defined limit (1023).
The journal will not detect any changes when the path name is greater than the limit, and therefore,
no error message will be generated.

Back to Contents

AIX GPFS known problems and limitations

When using the TSM for Space Management (also known as UNIX HSM) client:

An "umount" of a managed GPFS file system does not trigger failover.
Stub files that are restored using the TSM backup-archive client are restored to resident state if the option RESTOREMIGSTATE = NO is set.
If the stub file that has to be restored holds ACLs, the RESTOREMIGSTATE will be set to NO implicitly. This occurs because ACL data is stored along with the file data on the server.
If GPFS is restarted on the node where a file is being recalled with the dsmrecall command, the DMAPI session named 'dsmrecall' might still exist after GPFS restart. The extra DMAPI sessions do not affect HSM operations unless the DMAPI session limit (4000) is exceeded. If this happens, all further HSM operations will not be possible until the GPFS cluster is restarted completely.

GPFS only: GPFS ACL support has been modified to back up ACL data if a file's
change time (ctime) has changed since the last backup. This is done so that TSM
can accurately detect changes in ACLs and other GPFS extended attributes.
A file's ctime will change whenever the owner, mode, or ACLs change. When any
of these changes occur, TSM will back up the entire file even if the file's data has
not changed. This is because ACL data is stored along with the file data on
the server.
GPFS only: The size returned for extended file attributes
can be non-zero regardless of the presence of extended ACLs. Such files are processed
as if they have ACLs.

Back to Contents

AIX JFS2 known problems and limitations

When using the TSM Hierarchical Storage Management Client:
If you use the RESTOREMIGSTATE = NO option, stub files that are restored using the
Tivoli Storage Manager Backup-Archive Client are restored to resident state.

Back to Contents

HP-UX known problems and limitations

VxFS sparse file restore
When a sparse file is restored in the Veritas file system, it can occupy more disk space than it did originally at the time of backup. This is a limitation caused by the inner workings of the space allocation algorithm used by the file system.
Raw logical volume backup does not support devices other than logical volumes, such as /dev/dsk/c0t0d1. Logical volume devices usually take the form /dev/vgXY/lvolAB.
Because of the limited functionality of the dtterm application, not all function keys of the command line Clients operate as expected. The Control-Left and Control-Right key combinations and the Home and End keys do not work.
The default limit of the data segment size of a process in HP-UX (11.23) is 64 MB. When backing up large file systems, the TSM Client might exceed this limit and thus run out of memory. To increase this limit, the kernel needs to be modified in the following way:

For image backup, file systems are umounted and re-mounted read-only in order to prohibit access during the backup. However, you might still be able to change the size of the file system by using a volume manager. Such operations must be avoided during image backup.
The tape prompt option is not supported for the Web GUI.
The Veritas file system (VxFS) versions 3.5, 4.1 and 5.0, HP and Veritas volume managers are supported. Image backup support is restricted to logical volumes managed by the HP-UX or HP Itanium volume manager only (see the table below).

Image backup support
HP-UX version / volume manager	VxFS 4.1	VxFS 5.0
11i v2 / HP VM	supported	supported
11i v2 / VxVM	unsupported	unsupported
11i v3 / HP VM	supported	unsupported
11i v3 / VxVM	unsupported	unsupported

When doing image backup directly to tape, the RESOURCEUTILIZATION option value cannot exceed the value of the MAXNUMMP on the server for that node. If it does,the backup can fail with an Unknown System Error message.
During image backup and restore over LANFree, pressing Ctrl-C can lead to a core dump of the client. Such operations must be avoided during image backup or restore.

Back to Contents

Common Linux known problems and limitations

When completing a snapshot image backup, the client might be unable to remove the snapshot due to a known problem with Linux kernel version 3.0 and LVM. The problem is documented at https://bugzilla.novell.com/show_bug.cgi?id=642296. Use the "lvdisplay" command to list the active snapshots and the "lvremove" command to remove the snapshot manually.
Snapdiff backup run on NFS4 mounted file system on Linux backs up all, not only changed files, after successful first Snapdiff backup. The problem has been reported to NetApp under BURT 629745.
When restoring to NFS-mounted file systems, remount them using the mount option "noac" to prevent NFS-related errors that are not detectable by the client.

Note: The performance can be raised for NFS by using larger values for the options rsize and wsize (for example 8192).

Backupset restore from tape is not supported in this release of the Linux client.
If you get warnings about missing "standard symbols l" fonts when starting the Java GUI with dsmj, try replacing the "standard symbols l" string with the word "symbol" in the Java font.properties file. The font name "standard symbols l" might be different, because not all Linux distributions use the same name for this font.
Image restore to another location for XFS file system

When restoring an image with XFS file system to another location, you might see the message "ANS1066E The restore operation completed successfully, but the file system could not be remounted" appear on the screen. The mount of XFS file system is not possible because by default XFS does not allow the mount of several XFS file systems with the same FS UUID simultaneously. Use the XFS mount option "nouuid" to mount XFS file systems with the same UUID or manually change FS UUID using xfs_admin utility.

File system and ACL support

For Ext2/Ext3/Ext4/XFS/JFS ACL support on Linux, the Tivoli Storage Manager client uses the libacl.so library, so it is searched for in the following locations:

A colon-separated list of directories in the user's LD_LIBRARY_PATH environment variable.
The list of libraries cached in /etc/ld.so.cache. /usr/lib, followed by /lib.

Ensure there is a valid library or symbolic link present in one of the specified locations.

If the host name shown with the hostname command is not specified in the /etc/hosts file, there will be no GUID support available. If you want to have GUID support, you must add the host name into the /etc/hosts file.
In the Java GUI the preview of the Include-Exclude List might hang. This can be avoided by setting the following before starting the GUI:

export MALLOC_CHECK_=0

Option SKIPACL has to be set to 'yes' when reading block has to be aligned on sector size and server version less than 6.1
Mount recovery of an accidentally dismounted volume on the Windows proxy machine fails continuously on the Linux mount.

This problem is being investigated. In order to clear this problem, unmount the volume and remount it again.

Back to Contents

Linux86 known problems and limitations

If "dsmc rest <files> -su=yes -rep=yes" is running in LAN free environment and
kill the process in first few seconds with kill -15, the client could abort with a coredump.
(internal reference #96339)

=>dsmc rest /SMSVT/mmfs1/ -su=yes -rep=yes
IBM Tivoli Storage Manager
Command Line Backup-Archive Client Interface
Client Version 6, Release 4, Level 3.0
Client date/time: 05/25/2015 12:19:02
(c) Copyright by IBM Corporation and other(s) 1990, 2015. All Rights Reserved.
....
Restore function invoked.

ANS1247I Waiting for files from the server...
Restoring         131,072 /SMSVT/mmfs1/ [Done]
Restoring          10,240 /SMSVT/mmfs1/tf0 [Done]
....
Restoring         606,208 /SMSVT/mmfs1/tf1164 [Done]
ANS2820E An interrupt has occurred. The current operation will end and the
client will shut down.
Abort(coredump)
NQR (no query restore) does not work for HSM (space managed) file systems, when the RESTOREMIGSTATE option is set to YES. Only classical restore is possible in this case.
Templates and virtual machines deployed as vApps

The Tivoli Storage Manager Backup-Archive Client cannot back up vCenter virtual machine templates. Also, virtual machines that are deployed as vApps are not included in full VM backups.

During snapshot difference incremental backup of N-Series or NetApp volumes, if files are skipped due to any reason, such as TSM server media mount being unavailable, these files may not be backed up during subsequent backups either, unless the files have been modified. Such files will be logged to the TSM error log. You can back them up manually or perform a snapshot difference backup with the option "createnewbase" set to "yes".
VMware Transports(SAN, hotadd, NBDSSL, NBD) - VMVSTORTRANSPORT option

Use the VMVSTORTRANSPORT option with the backup VM or restore VM command to specify the transport to be used with the vStorage API for Data Protection (VADP). The transport determines how VADP accesses virtual disk data. Valid transports include any order or combination of san, hotadd, nbdssl, and nbd separated by a colon. The first transport in the list that is available in the environment will be used. NBD or network based data transfer is the LAN transport and should be available in all environments. It is not necessary to set this option. The default value is to use the VADP order that is currently defined as "san:hotadd:nbdssl:nbd". This option is passed directly to the VADP API.

Place the VMVSTORTRANSPORT option in the client options file (dsm.opt). Some common examples are shown below:

Current default order of transports when none is specified. VMVSTORTRANSPORT san:hotadd:nbdssl:nbd When SAN path is temporary unavailable fail the backup so as not to impact LAN. VMVSTORTRANSPORT san Disable the use of hotAdd when running backup server inside VM. VMVSTORTRANSPORT nbdssl:nbd Select NBD even when NBDSSL is available for better performance. VMVSTORTRANSPORT nbd

If you want to start the TSM scheduler using inittab, ensure that the required language environment is set. You can do this by starting a script that first sets the environment and then starts the scheduler. See also the section that handles SBCS in this file.
Image backup on raw devices

Only certified image backup from raw devices, where their corresponding partitions are marked with ID x83, are shown in the partition table. To see the IDs of all partitions of the partition table, use the command "fdisk -l" and look for the column marked with "Id".

When using option 'COMMMETHOD Sharedmem', you might get the error message 'ANR8294W Shared Memory Session unable to initialize' on the server or storage agent console. By default, Linux is not set up with sufficient system resources to create the message queues. You must increase the kernel parameter MSGMNI to 128 (default is 16). You can modify this parameter by performing the following command:

echo 128 > /proc/sys/kernel/msgmni

kernel.msgmni=128

ipcs -l

and look at the "max queues system wide" value. The default is 16.

Running the TSM client GUI by issuing the 'dsmj' command on RedHat 5.0 on x86_64 (AMD64 or Xeon) platform might result in the following error message:

-jar option not found

When attempting to perform TSM BA backup|restore|show vm operations on a RHEL or SLES system, you may run into errors like:

ANS4152E Failure initializing VMware virtual machine environment. RC=-303. Refer to client dsmerror.log for detailed error messages.

The fuse library package (libfuse2 on SUSE Linux, fuse-libs-2 on Red Hat Linux) is a required prerequisite for the VMware virtual machine environment and must be installed prior to performing any VMware virtual machine operations.

If you perform snapshot difference incremental backup of N-Series or NetApp volumes, make sure that the volumes are defined with a security style of UNIX. TSM backups will complete successfully even if the security style is NTFS or MIXED. However, it is recommended that you use a security style of UNIX.
When using the Tivoli Storage Manager for Space Management (HSM) client:

If GPFS is restarted on the node where a file is being recalled with the dsmrecall command, the DMAPI session named 'dsmrecall' might still exist after the GPFS restart. The extra DMAPI sessions do not affect HSM operations unless the DMAPI session limit (4000) is exceeded. If this happens, all further HSM operations will not be possible until GPFS is restarted on the quorum node.
An "umount" of a managed GPFS file system does not trigger failover.
Stub files that are restored using the TSM backup-archive client are restored to resident state if the option RESTOREMIGSTATE = NO is set.
If the stub file that has to be restored holds ACLs, the RESTOREMIGSTATE will be set to NO implicitly. This occurs because ACL data is stored along with the file data on the server.
If GPFS is restarted on the node where a file is being recalled with the dsmrecall command, the DMAPI session named 'dsmrecall' might still exist after GPFS restart. The extra DMAPI sessions do not affect HSM operations unless the DMAPI session limit (4000) is exceeded. If this happens, all further HSM operations will not be possible until GPFS is restarted completely.

If GPFS software is uninstalled while the TSM client is still installed, the TSM client will fail to load because of missing links to GPFS libraries. In this case you can either reinstall the TSM client or recreate the missing links manually. To recreate the links manually, issue the following commands:

ln -s /opt/tivoli/tsm/client/api/bin64/libgpfs.so /usr/lib64/libgpfs.so ln -s /opt/tivoli/tsm/client/api/bin64/libdmapi.so /usr/lib64/libdmapi.so

GPFS only: GPFS ACL support has been modified to back up ACL data if a file's change time (ctime) has changed since the last backup. This is done so that TSM can accurately detect changes in ACLs and other GPFS extended attributes. A file's ctime will change whenever the owner, mode, or ACLs change. When any of these changes occur, TSM will back up the entire file even if the file's data has not changed. This is because ACL data is stored along with the file data on the server.
GPFS only: The size returned for extended file attributes can be non-zero regardless of the presence of extended ACLs. Such files are processed as if they have ACL
GPFS 3.5 only: When SELINUX is used in enforcing mode, after an GPFS upgrade to the 3.5 level, the first incremental backup will be a full backup. For details, see also APAR IC88793 and the GPFS wiki onhttps://www.ibm.com/developerworks/mydeveloperworks/wikis/home?lang=en#/wiki/General%20Parallel%20File%20System%20%28GPFS%29/page/Using%20GPFS%20with%20SElinux

Back to Contents

LinuxPPC known problems and limitations

If you want to start the TSM scheduler using inittab, ensure that the required language environment is set. You can do this by starting a script first sets the environment and then starts the scheduler.
When using option 'COMMMETHOD Sharedmem', you might get error message 'ANR8294W Shared Memory Session unable to initialize' on the server or storage agent console. By default, Linux is not set up with sufficient system resources to create the message queues. You must increase the kernel parameter MSGMNI to 128 (default is 16). You can modify this parameter by performing the following command:

echo 128 > /proc/sys/kernel/msgmni

kernel.msgmni=128

ipcs -l

Back to Contents

LinuxZ known problems and limitations

If SELinux is installed on the system but is disabled, installation of the GSKit crypto component (gskcrypt64-8.0.14.11.linux.s390x.rpm) can produce multiple messages as follows:

libsepol.context_from_record: MLS is enabled, but no MLS context found
libsepol.context_from_record: could not create context structure
libsemanage.validate_handler: invalid context system_u:object_r:textrel_shlib_t specified for /usr/local/ibm/gsk 8/lib/C/icc/osslib/libcrypto.so.0.9.7 [all files]
libsemanage.dbase_llist_iterate: could not iterate over records

This is related to a problem that has been identified and resolved by Redhat. The problem is fixed in policycoreutils-1.33.12-14.4.el5. See this report for more information.

These installation errors do not affect TSM functionality and can be ignored. To work around the problem, you can bypass the SELinux check by setting the following environment variable before installation:

export GSK_DISABLE_SELINUX_TEST=1

The problem does not occur if SELINUX=enforcing is set.

Back to Contents

VMware off-host Full VM backup using vStorage APIs

Transport mode does not failover

FULL VM vstor backup (IFFULL, IFINCR, FULL, INCR) failures
Consistent backup failures have been experienced with some Tivoli Storage Manager version 6.4 Backup-Archive Client Linux installations on all supported Linux platforms, that are used as the data mover for VMware backups. IC88237 has been opened for the issue. A fix is included in the 6.4.0.1 interim fix available since January 31, 2013
Powering on a restored Linux Guest VM with EFI boot enabled may fail. Manual steps are required in order to configure the guest VM to properly boot from the restored Linux EFI boot disk.

The following steps describe the procedure to successfully boot from a restored Linux EFI guest VM:

TSM Linux vStorage Server - SAN transport prerequisites

SAN transport for full virtual machine backup and recovery from a physical Linux TSM vStorage backup server is only supported with the following configuration

Host hypervisor ESXi 5.x
Datastore must be formatted with VMFS 5
Datastore must reside on storage hardware that fully supports the vStorage APIs for Array Integration (VAAI), specifically the VAAI Atomic Test and Set (ATS) feature which is also known as Hardware Accelerated Locking

If the environment does not meet these prerequisite versions, normal operations on a host server such as power on/off of a virtual machine, snapshot creation, etc. could cause the backup from the vStorage backup server to fail. This problem is not seen when using a Windows vStorage backup server or a Linux vStorage backup server running in a guest VM since the hotadd transport is not effected. When running on a physical vStorage backup server, the problem can be avoided by setting one of the following options:

vmvstortransport "nbdssl:nbd" vmvstortransport "nbdssl" vmvstortransport "nbd"

If the storage hardware VAAI ATS feature is not being used the Tivoli Storage Manager Backup-Archive client processing on the Linux TSM vStorage backup server may fail with the following message:

ANS9365E VMware vStorage API error. TSM function name : vddksdkRead TSM file : vmvddksdk.cpp (2382) API return code : 16000 API error message : One of the parameters supplied is invalid ANS0361I DIAG: ANS1111I VmProcessExtent(): vddksdkRead() rc=-1, startSector=4480000, numSectorsToRead=512 ANS1228E Sending of object 'suse10vm04' failed ANS5283E The operation was unsuccessful.

This problem is the result of SCSI reservation conflicts. As such, SCSI reservation conflict errors will be reported in the /var/log/messages file on the Linux vStorage backup server.

The TSM BA Linux proxy requires full VAAI support

See VMware KB for description of vStorage for Array Integration (VAAI) support required for TSM BA Linux proxy SAN transport:

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1021976

VMware Change Block Tracking (CBT) support for content-aware backups

TSM Full VM backup will use VMware Change Block Tracking (CBT) support if available to enable content-aware (used-block only) backups. VMware CBT requires ESX 4.0 or later host (with virtual hardware 7). TSM will turn on CBT on the first TSM backup of the virtual machine. TSM Full VM backup does support virtual machines that do not support CBT. In this case both used and un-used areas of the disk will be backed up and a informational message will be logged to the dsmerror.log. The TSM GUI Backup dialog or command line 'dsmc show vm all' will display the CBT status.

RDM Physical Compatibility Mode and Independent disks

Backups are not supported for virtual machines that have either independent disks (either persistent or non-persistent), or that have raw device mapping (RDM) disks configured in physical compatibility mode. RDM disk configured in virtual compatibility mode are supported.

In previous releases, attempting to backup a Virtual Machine which included one of these disks would fail. Two options have been added to allow the Virtual Machine backup to succeed, but skip the backup of these disks. Also on restore, these disks will not be restored.

The new options are "VMPROCESSVMWITHPRDM=NO/YES" and "VMPROCESSVMWITHINDEPENDENTDISK=NO/YES". The default for these options is "NO", meaning the Virtual Machine will fail to backup. If both types of disks are included in the Virtual Machine, then both options must be specified and set to "YES" for the backup to succeed.

Client-side deduplication performance and return code=254

VMware off-host Full VM backups with client-side deduplication enabled fails with return code=254.

Error message: ANS7899E The client referenced a deduplicated extent that does not exist on the TSM server.

This error can occur when a TSM server expiration or similar process is either in the process of removing a deduplicated extent or has an extent locked during the backup. A subsequent backup should succeed.

Fix: When running heavy backup loads on the TSM server, if a VM backup fails multiple times with RC=254 or an increase backup run time is noticed when client-side deduplication is enabled, disable client-side deduplication for the VMware backup proxy node until a permeant fix is available. This issue is not seen when using server-side deduplication.

Incorrect level of the VMware Tools

When a TSM Backup VM fails consistently with the following error in the dsmerror.log, an incorrect level of the VMware Tools may be installed inside the guest OS:

ANS9365E VMware vStorage API error. TSM function name : visdkWaitForTask TSM file : vmvisdk.cpp (2811) API return code : 78

API error message : Cannot create a quiesced snapshot because the create snapshot operation exceeded the time limit for holding off I/O in the frozen virtual machine.

Fix: It may be necessary to uninstall and re-install the latest version of VMware Tools on the guest OS virtual machine.

If the problem persist, refer to the following VMware Knowledge Base (KB) article for detailed information and solution:

Titled: "Cannot create a quiesced snapshot because the snapshot operation exceeded the time limit for holding off I/O in the frozen virtual machine".

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1018194

Long running snapshot delete

A TSM Backup VM may fail if the delete snapshot of a failed prior run requires longer than the timeout value set by VMware vCenter. The following error may be seen in dsmerror.log as a result of this timeout value that is exceeded by vStorage API:

ANS9365E VMware vStorage API error, API error message: "server refused connection", API return code:14009,

Fix: The snapshot delete will continue to be processed. The next backup attempt should succeed.

Change Block Tracking (CBT) needs reset

When a TSM Backup VM fails with one of the following warning messages in the dsmerror.log:

ANS9365E VMware vStorage API error. TSM function name : visdkQueryChangedDiskAreas TSM file : vmvisdk.cpp (3592) API return code : 12 API error message : SOAP 1.1 fault: "":ServerFaultCode[no subcode]"A specified parameter was not correct. deviceKey" ANS9365E VMware vStorage API error. TSM function name : visdkPrintSOAPError TSM file : vmvisdk.cpp (885) API return code : 12 API error message : SOAP 1.1 fault: "":ServerFaultCode[no subcode]"Error caused by file /vmfs/volumes/4ade85fd-81f49624-57f5-000e0cdd0d21/winxp-32/winxp-32.vmdk"

It may be necessary to reset the VMware Change Block Tracking (CBT) for the virtual machine.

This can occur if the ESX server is shutdown or rebooted without first entering maintenance mode. It may also occur when a backup is attempted through both the ESX Server and vCenter at the same time for the same VM. Change Block Tracking is needed by TSM to backup content-aware (used-block only areas of the disk). If CBT is in this failed state, TSM will continue with the backup, and backup the used and unused areas of the disk.

Fix: Run a single TSM backup with the testflag vmbackup_cbt_reset:

Example: dsmc backup vm 'myvm' -testflag=vmbackup_cbt_reset.

This testflag will require a snapshot and snapshot delete to turn off CBT and a second snapshot to turn on CBT so should not be set permanently as it could impact backup performance.

Restore to ESX server instead of to the vCenter

Restoring a virtual machine to ESX server, when that VM was backed up from vCenter requires -datacenter=ha-datacenter parameter on the restore command.

"Clear lazy zero" is repeated in VMware vCenter task during VM Restore

When performing a virtual machine restore using vSphere 4.0 and when using SAN transport, the message "Clear lazy zero" is repeated in the "recent tasks" section of the VMware vSphere console, and the performance of the restore drops. VMware, instead of zeroing a VMDK as it is created, instead sets the lazy zero bit on each block in the VMDK so that it can be cleared at a later time. To work around the performance issue, a restore to the ESX server instead of a vCenter can be used.

Slow restore using SAN transport

Testing has revealed that although VM backups using SAN as transport is very fast, VM restore using SAN is usually the slowest of all transports. This problem is more prevalent when thin disks are used, and VMware recommends using NBDSSL for restores when thin disks are involved.

See VMware KB for more detailed information (Best practices when using SAN transport for backup and restore):

http://kb.vmware.com/kb/1035096

Thin-provisioned disk

There is a know problem in VMware vSphere 4 where a snapshot disk is not showing the correct thin- or thick- provisioned trait based on the actual disk's property. As a result, when a backup of a virtual machine with a thin-provisioned disk set in the disk properties is performed, TSM is unable to save the thin-provisioned disk attribute. Since this attribute is not saved, TSM is unable to restore the disk as thin-provisioned, but instead will restore the disk using the thick-provisioned disk attribute.

See VMware KB for detailed information:

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1020137

Update: VMware vSphere 5 now provides the ability to obtain the thin or thick-provision trait, so this information will be saved during a backup operation. The default restore behavior is to still restore all disks as thick-provisioned, even if they were thin at the time the virtual machine was backed up. This behavior can be overridden by specifying the test flag VMRESTOR_DONT_FORCE_THICK. Specifying this test flag will retain the original thin- or thick-provisioned trait, but only if Tivoli Storage Manager was able to obtain that attribute at the time of backup.

There is also a known issue with the VMware VDDK where thin disks being restored to a local datastore might incorrectly try to use the SAN transport. To restore thin disks in this situation, and also preserve the thin attribute, the "VMVSTORTRANSPORT nbdssl:nbd" (or any acceptable values excluding "san") should be used. Since "san" is not in the list of transports to use, it will not be incorrectly chosen when restoring thin disks to a local datastore.

English/7bit ASCII Virtual Machine names

There are some limitation of the VMware vStorage APIs and in TSM processing virtual machines with non-English/7bit ASCII characters in the name. Tivoli Storage Manager support for VMWare backup and restore is limited to English Virtual Machine names. Names using other languages is not supported at this time.

VMware Datastore space concerns for backups

When backing up a VM, a snapshot of the VM is taken and usually stored in the same datastore as the virtual hard disk being backed up. If there is not enough space in the datastore for this snapshot, an error will occur. If the VM is not running, the error returned will be something similar to -

'API error message : File testVM/testVM.vmx is larger than the maximum size supported by datastore[DataStore1_LUN1_500GB]'.

However if the VM is running, the error returned may not be as obvious and similar to -

'API error message : Cannot create a quiesced snapshot because the create snapshot operation exceeded the time limit for holding off I/O in the frozen virtual machine.'

See VMware KB for more detailed information:

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1012384

Virtual Machine names containing commas, semi-colon characters, or other special characters

The following characters have special meaning to Tivoli Storage Manager, and therefore virtual machine names containing these characters cannot be backed up using the FULL VM backup commands. These special characters are:

" ' : ; * ? , < > / \ |

Backing up the backup proxy

It is possible for a backup proxy running in a virtual machine to back itself up, but there are still some known issues where the backup may fail because a snapshot cannot be created. The backup will fail with the message:

"Cannot create a quiesced snapshot because the snapshot operation exceeded the time limit for holding off I/O in the frozen virtual machine".

It may be possible to prevent this message by modifying how part of the virtual machine is quiesced. It is possible to configure the virtual machine so that application data is not quiesced, while still allowing the filesystem to be quiesced. The implication to this is that application data may not be flushed to disk before the snapshot is taken; Filesystem data will still be flushed, and the I/O held while the snapshot is being taken.

See VMware KB for more detailed information:

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1031298

Backing up Windows 2008 and Windows 2008 R2 Virtual Machines Containing Dynamic Disks

Backing up Windows 2008 and Windows 2008 R2 virtual machines that contain dynamic disks may result in the dynamic disks being detached from Windows at the end of the backup. The dynamic disks are still intact and can be reconnected to Windows.

This problem occurs because application-consistent quiescing is performed when the snapshot is taken during the backup process. For application-consistent quiescing to be available on Windows 2008 and Windows 2008 R2 one of the conditions that Windows requires is that the virtual machine must not use dynamic disks. This is because VMware uses a hardware provider to perform the VSS snapshot, and shadow copies created by hardware providers of volumes that reside on dynamic disks have a specific requirement that they cannot be imported onto the same system.

See Microsoft VSS topic for more detailed information:

http://msdn.microsoft.com/en-us/library/aa384594%28v=vs.85%29.aspx

See VMware KB for more detailed information:

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1031298

Restoring Virtual Machines backed up with TSM 6.4 from down-level clients not supported

Virtual machines backed up using the 6.4 level of the TSM backup-archive client cannot be restored from down-level clients. This includes prior levels of the TSM backup-archive client, as well as prior levels of the TSM for Virtual Environments clients.

Backing up a Virtual Machine that has a TSM for Virtual Environments volume mounted

"Cannot create a quiesced snapshot because the snapshot operation exceeded the time limit for holding off I/O in the frozen virtual machine".

To allow the backup of this virtual machine to continue, unmount the volume from TSM for VE and retry the operation.

Avoid uses of / (slash), \ (backslash) or % (percent) in named elements (virtual machine names, datacenter names, folder names, etc)

VMware allows the uses of (/, \, %) but internally VMware stores each of this characters as an escape sequence. A slash is escaped as %2F or %2f, a backslash is escaped as %5C or %5c, and a percent is escaped as %25.

If one is using these special characters then the escape sequence must be used to reference them in TSM client.

e.g. OurDatacenter\LevelOne becomes OurDatacenter%5cLevelOne

Message "You do not have access rights to this file"

When a TSM Backup VM fails with the following error message in the dsmerror.log:

ANS9365E VMware vStorage API error. TSM function name : VixDiskLib_Open TSM file : vmvddksdk.cpp (1404) API return code : 13 API error message : You do not have access rights to this file

The following should be checked

Message "NBD_ERR_NETWORK_CONNECT

When using NBD transport, the following error in the dsmerror.log may indicate that a VMware limit of ESX host connections has been exceeded. User action is to reduce the vmmaxparallel and vmlimitperhost options to limit the number of TSM connections to the ESX host.

ANS9365E VMware vStorage API error. TSM function name : VixDiskLib_Open TSM file : vmvddksdk.cpp (1748) API return code : 14009 API error message : NBD_ERR_NETWORK_CONNECT

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1022543

Restore using a HotAdd mode, may not clean up target VMDKs from the DataMover vmx file.

The problem is being investigated. In order to clear it, turn off the DataMover VM, edit the vmx to remove the redundant vmdks.

Back to Contents

VMware VI API known problems and limitations

Tivoli Storage Manager client does not display virtual machines in Virtual Center linked mode. Tivoli Storage Manager development believes that this is a VMware issue. VI API do not support Virtual Center in linked mode. User has to setup backup proxy as it does today connecting each vCenter server. Tivoli Storage Manager client supports connection to one vCenter server at one time.

Back to Contents

Mac OS known problems and limitations

TSM Client GUI does not start if Oracle JRE 7 installed on MacOS 10.7 and above. JRE 6 has to be installed to run Java GUI. Workaround: double click on the TSM Client GUI icon and follow the prompts to install JRE 6.
Objects on a Xsan 2.0 file system have their modification, change, and attribute times stored in a high resolution format--to the nanosecond. Tivoli Storage Manager will not back up or restore these fractions of a second of times. The fractional time will not be used to determine if a file should be backed up and it will always be set to zero during restore.
TSM treats symbolic links as files. When an input path to the command line contains a symbolic link, TSM is unable to process the path. This only occurs when the path entered on the command line. This is due to limitations in Mac OS X.
Due to limitations in Mac OS X, files with certain characters will not be managed by TSM correctly. Some examples are:

0131 LATIN SMALL LETTER DOTLESS I 01C8 LATIN CAPITAL LETTER L WITH SMALL LETTER J 1FB2 GREEK SMALL LETTER ALPHA WITH VARIA AND YPOGEGRAMMENI 03D0 GREEK BETA SYMBOL 03D1 GREEK THETA SYMBOL

The recommendation is to either rename these files or use a case-sensitive file system.

Due to limitations in Mac OS X, Tivoli Storage Manager will crash if it encounters any of the following Unicode symbols as part of a name:

D800 - <Non Private Use High Surrogate, First> D87F - <Non Private Use High Surrogate, Last> DB80 - <Private Use High Surrogate, First> DBFF - <Private Use High Surrogate, Last> DC00 - <Low Surrogate, First>

Note: These are special Unicode flag symbols. They are not normally used in file names.

Back to Contents

Solaris known problems and limitations

The following are supported devices for the Solaris SPARC client:

Normal disk slices
Metadevices created with Solaris Volume Manager
Metadevices created with Veritas Volume Manager v5.x

NOTE: Meta devices created by the Veritas Volume Manager must be listed including its disk group in the /etc/vfstab to be recognized by the TSM backup-archive client for an image backup of file systems (ANS1134E). Raw devices should not be listed in the /etc/vfstab.

EXAMPLE: the following would be not recognized correctly:

/dev/vx/dsk/<meta device name>

/dev/vx/dsk/<disk group>/<meta device name>

The overlap device of the root disk (c0t0d0s2) is not supported from raw device backup. Avoid using this feature on disks or slices that are used as swapping device.
Disk slices containing cylinder 0 should not be backed up or restored. If this happens, the VTOC will be overwritten. If you need to back up the first disk slice, exclude cylinder 0 by starting the disk slice from cylinder 1 (using the format utility). During a restore, the backup-archive client does not check whether cylinder 0 is contained in the device that is overwritten.
Performance Degradation

If the DISABLENQR YES (Classic Restore) option is set, the restore performance of the Solaris client is slower when compared to previous releases. Depending on how much load you put on the system, the restore may take up to 25% more time.

For image backup, the file systems are unmounted and re-mounted read-only if you use COPYSERIALIZATION=STATIC in order to prohibit access during the backup. However, you might still be able to change the size of the file system by using a volume manager for all values of copy serialization. Such operations must be avoided during image backup.
Image backup of large volumes

The current client version can only back up volumes created on disks with the VTOC label. Volumes larger than 1TB require EFI disk label which is currently not supported by the TSM client. As a result image backup is not possible for volumes larger than 1 TB.

When doing image backup directly to tape, the RESOURCEUTILIZATION option value cannot exceed the value of the MAXNUMMP on the server for that node. If it does,the backup can fail with an Unknown System Error message.
Using the backupset and portable media functionality requires tapes that are fully compliant with the Sun generic device driver standards. Note that several non-Sun tapes are not fully compatible with all required I/O operations, although simple tape commands such as tar work without problems.
Sparse file handling

Sparse files (files whose block size and the number of physical disk blocks allocated to store file data are not equal) are handled as sparse files on the Tivoli Storage Manager Server. If a block of a file consists only of bytes with value zero, this block is not restored as a physical disk block. For sparse files with large holes in the address space this will lead to an increase of performance. If files have been backed up as sparse files and need to be restored as normal files (non-sparse files), this should be done by the internal option MAKESPARSEFILE NO in dsm.opt or makesparsefile=no which is supported by the command line Client only. The option is only necessary for files where the existence of physical disk blocks is required. This is the case in some rare situations for system files like ufsboot which is needed during boot time. The boot file loader of the operating system accesses physical disk blocks directly and does not support sparse files.

VxFS sparse file restore

When a sparse file is restored in the Veritas file system, it can occupy more disk space than it did originally at the time of backup. This is a limitation caused by the inner workings of the space allocation algorithm used by the file system.

Because of limitations of the standard system settings, the TSM Client does not use large communication buffers by default. This can be changed by specifying "LARGECOMMBUFFERS YES" in the Client system options file (dsm.sys). Read the following section for information about how to set up your system to let ordinary users run TSM with large communication buffers.

If you specify "LARGECOMMBUFFERS YES" in the dsm.sys file to enable large communication buffers, non-root users may get the following error message when starting an TSM Client or API application:

"ANS1030E System ran out of memory. Process ended."

set shmsys:shminfo_shmmax=2097152

Because of the limited functionality of the dtterm application, not all function keys of the command line Clients operate as expected. The Control-Left and Control-Right key combinations and the Home and End keys do not work.
During image backup and restore over LANFree, pressing the Ctrl-C keys can lead to a core dump. You must avoid such operations during image backup or restore.
Limitations of the TSM Solaris backup-archive client functionality regarding QFS:

The package LSCqfs 3.5.0 is the standalone version of QFS. Only this version of QFS is supported.
Image backup of QFS file systems is not supported.
A QFS file system contains two hidden system files and a system directory that cannot be backed up. A backup of these files is not needed. They contain internal data to manage the file system. This data will be automatically excluded from a backup and recreated automatically by the file system itself if a restore of files is invoked.
The TSM Solaris backup-archive client does not support the combination of QFS and SAM to archive files onto a tertiary background storage such as tapes. Files are recalled from tape onto disk automatically if migrated files are found during the backup.

When the TSM Remote Agent runs on Solaris, you might encounter communication failure from the browser while monitoring the backup or restore of multiple NAS volumes, even though the backup or restore operations complete normally. To avoid such monitoring problems, wait until one backup or restore operation is finished before starting another, or use the TSM Server Scheduler to back up or restore multiple NAS volumes.
Use NDMP directory level backup with valid filespace mapping.
Limitations of the TSM Solaris backup-archive client functionality regarding Solaris Zones:

It is not strictly required to install the TSM backup-archive client manually on each non-global zone, because it is automatically installed as usual when the TSM client is installed in the global zone, or when the non-global zones are created or installed. This is valid if the TSM packages are installed without using the -G option of the pkgadd command.
On sparse-root non-global zones the /usr file system is usually mounted as LOFS read-only. In this case the TSM installation procedure would not be able to create the required links for TSM libraries and TSM option files in the /usr file system. If TSM is already installed in the global zone these links are already present and no further actions are required, otherwise a Warning message is produced asking to create the required links manually from the administrator of the global zone. The exact command required to creates these links is printed out in the Warning message.
When performing a backup with the domain option set to "all-lofs" in a Solaris 10 global zone, all explicit loopback file systems (LOFS) are backed up including all loopback file systems (LOFS) of each non-global zones.
When performing an incremental backup of the local zones from the global zone, the administrator must take special care about which files from the zonepath should be included or excluded in the backup (for example, dev-directory, system files, kernel files, and so on, which are not automatically excluded).
Any image backup for a file system that is mounted within a non-global zone should be performed within the non-global zone where the file system is mounted.

Back to Contents

[{"Product":{"code":"SSGSG7","label":"Tivoli Storage Manager"},"Business Unit":{"code":"BU058","label":"IBM Infrastructure w\/TPS"},"Component":"Client","Platform":[{"code":"PF002","label":"AIX"},{"code":"PF010","label":"HP-UX"},{"code":"PF016","label":"Linux"},{"code":"PF014","label":"iOS"},{"code":"PF027","label":"Solaris"}],"Version":"6.4","Edition":"","Line of Business":{"code":"LOB26","label":"Storage"}}]

Tips

TSM V6.4 UNIX and Linux Backup-Archive Client known problems and limitations

Question & Answer

Question

Answer

Was this topic helpful?

Document Information

UID

Share your feedback

Need support?