PM31517: ARCHIVE TAPE READ REQUEST CAN HANG IN A DATA SHARING ENVIRONMENT WHEN MULTIPLE, CONCURRENT REQUESTS ARE ACTIVE

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• Many concurrent IFI for IFCID306 log-merge requests are
  contending for archive tape volumes in a 2 member data sharing
  environment.  When a minimum of 3 requestors are contending for
  2 archive tape volumes, there is a timing window in DSNJR206
  where a archive tape volume request can be sent to the wrong
  archive read service task.  The owner of the task will not hold
  the volume lock for the archive volume 'owned' by the reader
  service task which can result in a dead-lock when a subsequent
  request is queued for that service task.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All data sharing users of DB2 that have      *
  *                 archive log data sets on tape volumes.       *
  ****************************************************************
  * PROBLEM DESCRIPTION: Threads reading log records from        *
  *                      archive tape volumes in a data sharing  *
  *                      environment may hang or deadlock.       *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  When there are at least three threads merging log records
  from tape volumes in a data sharing environment, a hang
  or deadlock situation may occur when the three threads
  contend for two tape volumes.  There is a timing window
  that can result in an archive read request getting scheduled
  to the wrong ARDE by the archive read scheduler.  If this
  occurs, the owner of the ARDE reader task may not hold the
  archive tape volser lock, which can result in a hang or
  deadlock situation.  The -CANCEL THREAD command can be
  used to cancel one of the log read threads to break the
  deadlock.
  

Problem conclusion

• The archive read scheduler has been changed to not schedule a
  request to an ARDE with queued requests, as the queued requests
  may be for a different archive tape volume.
  Note: it is not recommended to have the COPY1 archive log
  data sets on tape volumes in a data sharing environment.
  Since only one request can be processed at a time for a tape
  volume, this can lead to lengthy tape volume lock delays
  and IRLM deadlocks.  It is recommended that at least the
  COPY1 archive logs be placed on DASD volumes to avoid these
  delays.
  

Temporary fix

Comments

APAR Information

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

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

  UK65556 UK65558 UK65557

Modules/Macros

• DSNJR206
  

Fix information

• Fixed component name

  DB2 OS/390 & Z/

• Fixed component ID

  5740XYR00

Applicable component levels

• RA10 PSY UK65556

     UP11/03/23 P F103

• R810 PSY UK65557

     UP11/03/23 P F103

• R910 PSY UK65558

     UP11/03/23 P F103

Fix is available

• Select the PTF appropriate for your component level. You will be required to sign in. Distribution on physical media is not available in all countries.