IJ03271: HA: CL_MIRRORSET MAY FAIL TO RETRIEVE LV PARTITION DATA

PowerHA-7.2.2 SP1

APAR status

• Closed as program error.

Error description

• A LVM_IO_FAIL entry is not only logged in errpt,
  when a logical partition is no longer accessable
  (no mirror copy is available), but also when one
  copy is still accessable.
  
  A LVM_IO_FAIL entry triggeres the PowerHA
  clreserror error notification script, which
  is supposed to verify if a full copy of each
  shared LV is available using the cl_mirrorset
  script.
  
  A fallover should only be triggered in
  case not all data are accessable (for at
  least one shared LV a full copy of the
  data is not available).
  
  In case of high IO the cl_mirrorset script
  may fail to read LVM data, because LVM is
  writing many updates VGSA for missing disks
  and stale PPs.
  
  This can lead to:
  - an incorrectly triggered fallover
  as well as
  - no fallover when a fallover is needed
  

Local fix

• To avoid incorrect selective fallovers, when a full
  copy of data is available:
  
  Add
  HACMP_MIRRORSET_MAXTRIES=X
  to /etc/environment with X > 3 (e.g. 10).
  
  Note:
  This setting does not help to prevent cases,
  where no fallover has been triggered even
  if it would have been correct to do so.
  

Problem summary

• LVM_IO_FAIL in PowerHA may incorrectly triggers
  fallover of the RG where it is not necessary.
  

Problem conclusion

• Added retry logic to read LV's to check to complete
  copy of LV's are available in the available disks of the VG
  

Temporary fix

Comments

APAR Information

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

  IJ00926

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

Fix information

• Fixed component name

  POWERHA SYSMIR

• Fixed component ID

  5765H3900

Applicable component levels