IV89070: POTENTIAL DATA LOSS IN VIOS SSP WITH AN UN-INITIALIZED THICK LU

APAR status

• Closed as program error.

Error description

• **************************************************************
  * USERS AFFECTED:
    * Systems running the AIX 6100-09 Technology Level or
    * VIOS 2.2.4.2x
    * with pool.basic.rte between and including the 6.1.9.101 and
    * 6.1.9.103 levels.
    **************************************************************
    * PROBLEM DESCRIPTION:
    * When using VIOS SSP with thick LUs it is possible to have
    * data loss if the thick LU is accessed very soon
    * after creation, while SSP is still initializing the thick
    * blocks.
    *
    * We've seen this happen in two very rare scenarios:
    *
    * If data is written to a thick LU while it is being
    * initialized, and the pool server node (MFS manager) happens
    * to encounter a problem and stops suddenly, the write
    * may be lost.  And further, the pool will fail to find a new
    * server node and become unusable.
    *
    * Also, if a thick LU is cloned with PowerVC while the
    * LU is still being initialized, a block may be temporarily
    * shared between both LUs.  During this short time, writes
    * to base and cloned LU may overwrite each other, causing
    * data loss.
    **************************************************************
    * RECOMMENDATION:
    * Install APAR IV89070.
    * Prior to fix availability, an interim fix is available from
    * either
    * ftp://aix.software.ibm.com/aix/ifixes/iv89070/
    * https://aix.software.ibm.com/aix/ifixes/iv89070/
    * Installation of the ifix requires a reboot.
    **************************************************************
  

Local fix

• This problem can be avoided by waiting to use a thick LU until
  it has been completely initialized.
  

Problem summary

•   **************************************************************
    * USERS AFFECTED:
    * Systems running the AIX 6100-09 Technology Level or
    * VIOS 2.2.4.2x
    * with pool.basic.rte between and including the 6.1.9.101 and
    * 6.1.9.103 levels.
    **************************************************************
    * PROBLEM DESCRIPTION:
    * When using VIOS SSP with thick LUs it is possible to have
    * data loss if the thick LU is accessed very soon
    * after creation, while SSP is still initializing the thick
    * blocks.
    *
    * We've seen this happen in two very rare scenarios:
    *
    * If data is written to a thick LU while it is being
    * initialized, and the pool server node (MFS manager) happens
    * to encounter a problem and stops suddenly, the write
    * may be lost.  And further, the pool will fail to find a new
    * server node and become unusable.
    *
    * Also, if a thick LU is cloned with PowerVC while the
    * LU is still being initialized, a block may be temporarily
    * shared between both LUs.  During this short time, writes
    * to base and cloned LU may overwrite each other, causing
    * data loss.
    **************************************************************
    * RECOMMENDATION:
    * Install APAR IV89070.
    * Prior to fix availability, an interim fix is available from
    * either
    * ftp://aix.software.ibm.com/aix/ifixes/iv89070/
    * https://aix.software.ibm.com/aix/ifixes/iv89070/
    * Installation of the ifix requires a reboot.
    **************************************************************
  

Problem conclusion

• Detect that thick block extent is updated before MFS failure
  and do not terminate it so that blocks will be in two
  locations.
  

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:

Fix information

• Fixed component name

  AIX 610 STD EDI

• Fixed component ID

  5765G6200

Applicable component levels

• R610 PSY U872800

     UP17/01/19 I 1000 Ž