After a SP/SPP failover, it's possible that the DIMMs can be reported inconsistenly if one or more CMIOUs were replaced with CMIOUs of different set of DIMMs prior to the SP/SPP failover. As shown in the following example for a SPARC M8-4 server, ereports for faulted DIMMs result following the fail over of DCU1.
-> set /SYS/DCU1 initiate_sp_failover=true Are you sure you want to set initiate_sp_failover=true (y/n)? y Set 'initiate_sp_failover' to 'true' 2017-07-27/11:48:36 ereport.fruid.inaccessible@/SYS/CMIOU4/CM/CMP/BOB11/CH1/DIMM 2017-07-27/11:49:18 ereport.fruid.inaccessible@/SYS/CMIOU4/CM/CMP/BOB21/CH1/DIMM 2017-07-27/11:49:39 ereport.fruid.inaccessible@/SYS/CMIOU4/CM/CMP/BOB00/CH1/DIMM 2017-07-27/11:50:07 ...
Recovery: Only when the HOST for the replaced CMIOU is off, complete these steps to reboot the SPM that manages the CMIOU:
Run the show command on the host for the replaced CMIOU and check the output from the power_state property to verify that it reports the host off.
For SPARC M8-8 servers, use the show /HOST0 command.
For SPARC M8-4 servers, use the show command for the correct host, where CMIOU[0-3] belong to /SYS/DCU0 on HOST0, and CMIOU[4-7] belong to /SYS/DCU1 on HOST1.
For example:
-> show /HOST1 ... power_state = Off send_break_action = (Cannot show property) sp_name = /SYS/SP0/SPM0 state_capture_mode = default state_capture_on_error = enabled state_capture_status = enabled status = OpenBoot Running ... ->
If the host is off, proceed to Step 3. If the HOST reports that is on, power off the host and proceed to Step 3.
Now, identify the SPM that manages the CMIOU that reports DIMM issues. You can determine the SPM from the output for the sp_name property as follows:
-> show -t -l 2 sp_name dcus_assigned Target | Property | Value -------------------+-------------------+-------------- /HOST0 | dcus_assigned | /SYS/DCU0 /HOST0 | sp_name | /SYS/SP1/SPM0 /HOST1 | dcus_assigned | /SYS/DCU1 /HOST1 | sp_name | /SYS/SP1/SPM1
Reboot the SPMs identified in the previous step. For example:
For SPARC M8-8 servers, use the reset /SYS/SP0/SPM0 and reset /SYS/SP1/SPM0 commands.
For SPARC M8-4 servers, if CMIOU 0-3 have issues, reset /SYS/SP0/SPM0 and reset /SYS/SP1/SPM0 commands, and if CMIOU 4-7 have issues, reset /SYS/SP0/SPM1 and reset /SYS/SP1/SPM1 commands.