IJ30878: AFM GATEWAY CRASH DURING THE READ

APAR status

• Closed as program error.

Error description

• If the read on the file is in progress , next read on the
  same file will bypass read if the gap is more than 64MB.
  This is causing the problem if the original read is
  stuck in the kernel and  if its gets killed due
  to afmSyncOpWaitTimeout , bypass read causes kernel
  crash as the original read frees up the allocated mount
  structure.  readbypass is undocumented config option
  and is not persistent across daemon restarts. so when
  user restarts
  daemon during the efix installation, this change
  will be lost and user does not need to set to back to
  default value.
  

Local fix

• work around to avoid crash:  mmfsadm afm  readbypass -1
  

Problem summary

• AFM gateway node crashes if the home is not
  responding and multiple threads are trying to
  read the same file.
  

Problem conclusion

• Benefits of the solution:
  AFM gateway node no longer crashes
  
  Work around:
  Use undocumented config option
  "mmfsadm afm readbypass -1" at the gateway node.
  
  Problem trigger:
  Reading the same file from multiple threads
  when the home is not responding
  
  Symptom: Crash
  
  Platforms affected: All Linux OS environments
  
  Functional Area affected: AFM
  
  Customer Impact: HiPER
  

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:

  IJ31044

Fix information

• Fixed component name

  SPEC SCALE STD

• Fixed component ID

  5737F33AP

Applicable component levels