Edit mcelog File if Faults Are Not Present in the Fault Management Database

If the entry raw = yes in the mcelog.conf file is commented out, the Oracle Linux Fault Management software cannot obtain the information it needs to create a fault case. If that happens, fault cases for machine check events processed by mcelog are not added to the fault management database.

1. Confirm that the format of mcelog file messages are in the raw format by opening /var/log/mcelog in a text editor.

   The following table shows an example of a default message and a "raw" message (required by Oracle Linux FMA).

   Default Format

   Raw Format

   Hardware event. This is not a software error. MCE 0 CPU 0 BANK 8 MISC 7 ADDR 102bfc0368 TIME 1383171020 Wed Oct 30 18:10:20 2013 MCG status:EIPV MCIP MCi status: Corrected error Error enabled MCi_MISC register valid MCi_ADDR register valid MCA: MEMORY CONTROLLER RD_CHANNEL0_ERR Transaction: Memory read error STATUS 9c00000000000090 MCGSTATUS 6 MCGCAP 1000c14 APICID 20 SOCKETID 1 CPUID Vendor Intel Family 6 Model 45

   CPU 0 BANK 8 TSC 0 RIP 00:0 MISC 0x85 ADDR 0x102bfc0368 STATUS 0x9c00000000000090 MCGSTATUS 0x6 PROCESSOR 0:0x306f1 TIME 1383171020 SOCKETID 1 APICID 20 MCGCAP 0x1000c14

2. If the messages in the mcelog file are in the default format, edit the /etc/mcelog/mcelog.conf file to uncomment the “raw = yes” entry.
3. Delete the old mcelog file that was in the default format.

   rm /var/log/mcelog

4. Then restart the mcelog daemon, as follows:

   service mcelogd restart