IC84946: HSM FOR WINDOWS NEW RECALL OPTION CONCURRENTTHREADLIMIT

APAR status

• Closed as program error.

Error description

• New option in HSM for Windows to control the number of
  concurrent recall threads
  
  Problem Description:
  When an application is accessing a migrated file through a CIFS
  share, a system thread is utilized on the file server to send
  back the file content to the CIFS client. The system thread is
  blocked until the file is recalled by the HSM Recall service.
  Depending on the number of concurrent recalls and the time
  needed to recall each file, several system threads may be
  blocked at a time. Additionally, more than one system thread may
  wait for the same file, if multiple requests recalling the same
  file are issued. However the number of system threads is limited
  and depends on the operating system configuration and hardware.
  The file server might become unresponsive with a high number of
  blocked system threads until some or all files are recalled and
  the system threads are unblocked.
  
  Affected Products:
  HSM for Windows 5.5.x.x, 5.4.x.x, 6.1.0.0 - 6.1.5.0 (inclusive),
  6.3.0.0 - 6.3.0.1 (inclusive)
  
  Customer/L2 diagnostics (if applicable): No errors are reported
  by HSM for Windows, but file server is unresponsive for some
  period of time. Cluster environments may initiate failovers.
  
  Additional Keywords: HSM for Windows, Cluster failover
  
  Initial Impact: File server is hanging
  

Local fix

• Increase the number of Windows system threads. Increase recall
  performance, e.g. by using more tape drives or using disk
  storage. Find out why that many recalls are initiated at once,
  reduce the number of concurrent recalls if possible. You can set
  HSM for windows recall option ConcurrentRecallLimit, IC72637
  (NEW HSM FOR WINDOWS RECALL OPTION CONCURRENTRECALLLIMIT)
  available in version 6.1.4.2 and later versions.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: TSM HSM for Windows Client Versions          *
  * 5.4.x.x, 5.5.x.x and 6.1.0.0 to 6.1.5.0 (inclusive),         *
  * 6.3.0.0 - 6.3.0.1 (inclusive)                                *
  *                                                              *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * See ERROR DESCRIPTION                                        *
  ****************************************************************
  * RECOMMENDATION: Apply fixing level when available.           *
  * The new options are not available for versions 5.4.x.x and   *
  * 5.5.x.x.                                                     *
  ****************************************************************
  

Problem conclusion

• PROBLEM CONCLUSION
  The new option ConcurrentThreadLimit has been implemented to
  control the maximum number of concurrently recalling Windows
  system threads. During recalls Windows system threads are
  blocked. Because the number of system threads is fixed the
  Windows system may become unresponsive.
  If a new recall request is initiated such that the number of
  concurrently recalling system threads exceeds the value of
  ConcurrentThreadLimit, the HSM for Windows filter driver will
  not initiate the recall of the file, but immediately return the
  recall request with an error. The error code can be configured
  with the option ConcurrentThreadLimitCode. The return code
  received by the application is the only indication that the
  ConcurrentThreadLimit might have been exceeded. No errors are
  logged by HSM for Windows.
  The value for ConcurrentThreadLimit applies to the total of all
  concurrently recalling system threads regardless of whether they
  were triggered through CIFS access or local file system access.
  This option is different from the configurable amount of
  concurrent connections of the HSM recall service (see
  http://publib.boulder.ibm.com/infocenter/tsminfo/v6/topic/com.
  ibm.itsm.hsmwin.doc/c_recall_stngs.html). This option is
  different from HSM for Windows recall option
  ConcurrentRecallLimit IC72637 (NEW HSM FOR WINDOWS RECALL OPTION
  CONCURRENTRECALLLIMIT), as one recall may block several Windows
  system threads waiting for the same stub file being recalled.
  
  In order to activate the concurrent thread limit the
  administrator has to manually create one DWORD value in the
  Windows registry under
  HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\ithsmdrv
  \Parameters :
  Name:  ConcurrentThreadLimit
  Type:  REG_DWORD
  Value: Maximum number of allowed concurrent recall threads. The
  minimum value is "1" and the maximum value is 0x7FFFFFFF. If the
  option is not set, the maximum value is used. Every concurrent
  recall thread request exceeding this number is immediately
  returned with an error. The error code returned is by default
  0xC0000267 (STATUS_FILE_IS_OFFLINE).
  The system administrator can optionally change the error code by
  manually creating this DWORD value in the Windows registry at
  the same location:
  Name:  ConcurrentThreadLimitCode
  Type:  REG_DWORD
  Value: Error code returned to the application when
  ConcurrentThreadLimit is exceeded. If the option is not set, the
  default value of 0xC0000267 (STATUS_FILE_IS_OFFLINE) is
  returned. Consult with your Microsoft Windows system
  administrators or Microsoft Windows technical support before
  changing this value.
  After setting or changing any of these values the system must be
  rebooted in order to apply the changes.
  When the HSM file system filter driver recognizes and applies
  the registry value ConcurrentThreadLimit during boot time it
  reports feedback by writing the registry key
  ConcurrentThreadLimitStamp under
  HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\ithsmdrv
  \Parameters :
  Name:  ConcurrentThreadLimitStamp
  Type:  REG_SZ
  Value: UTC date and time, maximum number of concurrent threads,
  return code for recalls that are denied due to the concurrent
  threads limit. (Sample: "12-06-18 08:33:42.062: 16,
  0xc0000267").
  If the HSM administrator or supporter recognizes this registry
  key he can be sure that the registry was configured correctly.
  
  Note: File access errors can occur for any number of reasons. It
  is possible for applications to receive 0xC0000267 for reasons
  other than exceeding the ConcurrentThreadLimit. It is also
  possible for applications to receive other file access return
  codes unrelated to ConcurrentThreadLimit.
  The error message presented to the user depends on how the
  application handles the error code to an error message.
  
  These options should not be applied or changed without thorough
  testing. The behavior of the applications receiving errors
  cannot be predicted nor controlled by HSM for Windows.
  HSM for Windows is not responsible for any application or
  operating system malfunction as result of applying these
  options.
  

Temporary fix

• See LOCAL FIX
  

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  TSM HSM FOR WIN

• Fixed component ID

  5698ICNCL

Applicable component levels

• R610 PSY

     UP

• R630 PSY

     UP