IJ58225: THE CODE INCORRECTLY USED THE RAW MS-BASED VALUE (0X1E) DIRECTLY

APAR status

• Closed as program error.

Error description

• Based on the trace analysis, the following sequence of
  events was observed:
  
  * A command timeout occurred (likely after 30 seconds,
    consistent with rw_timeout = 30 seconds).
  * The command subsequently entered the retry logic.
  * After a new path was selected, we observed that PCM
    had PCM_QUIESCE_IO set.
  * The code then used retry_delay_code with a value of
    0x1E, which represents 30 ? 100ms = 3000ms (3 seconds).
  * However, mpio_delay is expected to be specified in
    seconds, but the code incorrectly used the raw ms-based
    value (0x1E) directly as seconds.
  * After setting the timeout along with SST_MPIO_DELAY,
    scsidisk_start() was invoked.
  * Under SST_MPIO_DELAY, a 30‑second watchdog timer
  is
    scheduled.
    -> Since the command had already spent 30 seconds,
       an additional 30 seconds were added, resulting in
       a total latency of ~60 seconds.
  

Local fix

• correcting the code behaviour
  

Problem summary

• When I/O is scheduled on a path that has just recovered from a
  failure, the operation may experience additional delay. In HA
  cluster environments, this unexpected delay can contribute to
  node timeouts and may result in a system crash
  

Problem conclusion

• The code was updated to enforce an appropriate delay when a
  path is in the recovery phase
  

Temporary fix

Comments

APAR Information

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

  IJ57772

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

Fix information

• Fixed component name

  AIX V7.2

• Fixed component ID

  5765CD200

Applicable component levels