OA51641: IXGLOGR STAGING DATA SET LCB LATCH DEADLOCK STATUS STOP HANG OA50362 17/01/10 PTF PECHANGE

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• This problem was introduced by APAR OA50362. If a fix for the
  symptoms for OA50362 is desirable, the recommendation is to
  install the OA50362 PTFs only after the fix for OA51641 is
  installed.
  
  IXGLOGR was hung because SRBs and TCBs were status stopped per
  ASCBDSP1. The only dispatchable TCB was running IXGA1TSK and
  waiting for an LCB latch held by an SRB running IXGF1TRX. The
  IXGF1TRX SRB had run to process a transition from using the
  staging dataset to using local buffers. The SRB scheduled an
  allocation request to synchronously delete the staging data set
  and then suspended. The IXGA1TSK TCB picks up the allocation
  request rqe and issues an IOS purge which causes the STATUS
  STOP. The TCB then made the LCB latch request which caused the
  deadlock.
  
  One trigger for the scenario is the completion of duplexed
  rebuild of a coupling facilty structure. The transition to a
  duplexed structure for a logstream configured STG_DUPLEX(YES)
  DUPLEXMODE(COND) can cause System Logger to switch duplexing
  methods. For example, given the DUPLEXMODE(COND) policy
  specification, when a CF structure is simplex-mode and a
  system-managed rebuild occurs that results in a duplex-mode
  structure, the new configuration can modify the non-volatility
  of the CF structures or change the failure-isolated
  relationship between the z/OS images and the composite view of
  the two (2) CF structures. This causes Logger to stop duplexing
  to staging data sets and a request to delete the staging data
  set. The structure duplex transition is evident with IXG219I:
  IXG219I SYSTEM LOGGER PROCESSED TRANSITION TO DUPLEX MODE FOR
  STRUCTURE xxxxx
  
  In one case, XCF took an ABEND026 RSN08118001 dump on another
  system because of the IXGLOGR hang.
  

Local fix

• -The client can consider configuring DUPLEXMODE(UNCOND) and
  LOGGERDUPLEX(UNCOND) for logstreams connected to structures
  to be rebuilt. This will avoid the delete of the staging
  dataset for the CF structure change.
  
  -The hung logger address space can prevent responses to
  subsequent XES structure events. Msg IXC360I in
  response to a d xcf,str command indicates the hung connectors.
  
  D XCF,STR,STRNAME=OPERLOG
  
  IXC360I  16.16.49  DISPLAY XCF
  STRNAME: OPERLOG
  .
  .
  .
  
  * ASTERISK DENOTES OUTSTANDING REBUILD PROCESSING FOR CONNECTOR
   CONNECTION NAME  ID VERSION  SYSNAME  JOBNAME  ASID STATE
   ---------------- -- -------- -------- -------- ----
  
  *IXGLOGR_BMPB     0D 000D0006 BMPB     IXGLOGR  001B ACTIVE
  *IXGLOGR_BMPY     06 00060035 BMPY     IXGLOGR  001B ACTIVE
  *IXGLOGR_BMP2     07 00070027 BMP2     IXGLOGR  001B ACTIVE
  *IXGLOGR_BMP3     04 0004004C BMP3     IXGLOGR  001B ACTIVE
  *IXGLOGR_BMP7     05 00050039 BMP7     IXGLOGR  001B ACTIVE
  *IXGLOGR_BMP9     08 00080015 BMP9     IXGLOGR  001B ACTIVE
  
  Forcing IXGLOGR on the systems that are hung will allow the
  structure event to complete.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: Users of HBB7780 and above supported         *
  *                 releases using System Logger Services        *
  *                 with PTFs UA81931, UA81948 or UA81949        *
  *                 for OA50362 applied.                         *
  ****************************************************************
  * PROBLEM DESCRIPTION: IXGLOGR hung waiting on logstream       *
  *                      latch while managing staging data       *
  *                      set.                                    *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  At the end of transitioning from a staging data set to local
  buffers the staging data set must be deleted synchronously.
  The staging data set is deleted while holding the logstream
  latch.  During the delete of the staging data set, all I/O
  will be purged.  The purge thread will attempt to obtain
  the logstream latch, causing a deadlock condition.
  

Problem conclusion

• Logic has been corrected to remove the deadlock condition.
  
  This fix may be applied to one system at a time in the sysplex.
  
  Systems with and without the fix activated can coexist in the
  same sysplex, although the problem may persist on those systems
  running without the fix.
  
  Additional Searchable keywords:
  SYSPLEXDS
  

Temporary fix

• *********
  * HIPER *
  *********
  

Comments

APAR Information

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

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

  UA83932 UA83933 UA83934

Modules/Macros

• IXGA1MM  IXGA1MTX IXGC4STE IXGF1TRX
  

Fix information

• Fixed component name

  SYSTEM LOGGER

• Fixed component ID

  5752SCLOG

Applicable component levels

• R7A0 PSY UA83932

     UP17/02/15 P F702 «

• R780 PSY UA83933

     UP17/02/15 P F702 «

• R790 PSY UA83934

     UP17/02/15 P F702 «

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.