Re: Missing MCA error messages for bad ECC

On 2/1/2026 15:30, Bob Bishop wrote:
> Hi,
>
>> On 1 Feb 2026, at 16:35, G. Paul Ziemba<[email protected]> wrote:
>>
>> OS: 14.2-STABLE as of 250403
>>
>> I seem to have at least one bad ECC DIMM
> Check the power supply voltages are within tolerance if you haven’t already.
>
>> and  was expecting to see MCA
>> messages in /var/log/messages or to the console (which I have recently
>> redirected to /var/log/console.log via syslog.conf:
>>
>>     console.info /var/log/console.log
>>
>> but I can't find anything in any of my logs. Why am I not seeing them?
> If you have the -F variant of the board that supports IPMI, it may be that the BMC is capturing the errors so check the BMC event log. Possibly there is a setting on the BMC to control what gets passed to MCA.
>
> Also check the BIOS event logging; I don’t see settings in the BIOS to control MCA events.
>
> And check the BIOS version is up to date.
>
>> Background:
>>
>> Motherboard: Supermicro X11SCA
>> CPU: Xeon E-2176G
>> Chipset: C246
>> Memory: 4x SK Hynix HMA82GU7CJR8N-VK (16GB ECC)
>>
>> Bios reports ECC on its startup screen and dmidecode reports
>>
>>     Total Width: 72 bits
>>     Data Width: 64 bits
>>
>> for each of the dimms.
>>
>> Amanda started reporting checksum errors on large backup files in its
>> holding disk. I discovered that a large file (200GB) on any of three
>> disks on this system yields different sha512sum values every time I
>> run it on the same file. SMART data looks OK on all disks.
>>
>> memtest86+ finds three bad spots in memory, at 42G, 47G and 53G. I have
>> 4x16GB dimms installed, so I think that corresponds to two bad dimms.
>>
>>     % sysctl hw.mca
>>     hw.mca.cmc_throttle: 60
>>     hw.mca.force_scan: 0
>>     hw.mca.interval: 300
>>     hw.mca.maxcount: -1
>>     hw.mca.count: 0
>>     hw.mca.erratum383: 0
>>     hw.mca.intel6h_HSD131: 0
>>     hw.mca.amd10h_L1TP: 1
>>     hw.mca.log_corrected: 1
>>     hw.mca.enabled: 1
>>
>> Thanks for any insights.
>> -- 
>> G. Paul Ziemba
>> FreeBSD unix:
>> 8:31AM  up 2 days, 14:38, 11 users, load averages: 0.71, 0.43, 0.39
>>
I have one of these boards in a server here:

Platform Firmware Information
         Vendor: American Megatrends Inc.
         Version: 2.5
         Release Date: 06/14/2024
         Address: 0xF0000
         Runtime Size: 64 KiB
         ROM Size: 32 MiB
....

Base Board Information
         Manufacturer: Supermicro
         Product Name: X11SCA-F
         Version: 1.01A

....

Handle 0x002F, DMI type 17, 84 bytes
Memory Device
         Array Handle: 0x001F
         Error Information Handle: Not Provided
         Total Width: 72 bits
         Data Width: 64 bits
         Size: 16 GiB
         Form Factor: DIMM
         Set: None
         Locator: DIMMA1
         Bank Locator: P0_Node0_Channel0_Dimm0
         Type: DDR4
         Type Detail: Synchronous
         Speed: 2667 MT/s
         Manufacturer: SK Hynix
         Serial Number: 7474963A
         Asset Tag: 9876543210
         Part Number: HMA82GU7CJR8N-VK
         Rank: 2
         Configured Memory Speed: 2667 MT/s
         Minimum Voltage: 1.2 V
         Maximum Voltage: 1.2 V
         Configured Voltage: 1.2 V
         Memory Technology: DRAM
         Memory Operating Mode Capability: Volatile memory
         Firmware Version: Not Specified
         Module Manufacturer ID: Bank 1, Hex 0xAD
         Module Product ID: Unknown
         Memory Subsystem Controller Manufacturer ID: Unknown
         Memory Subsystem Controller Product ID: Unknown
         Non-Volatile Size: None
         Volatile Size: 16 GiB
         Cache Size: None
         Logical Size: None

.... (and for the other 3, 64Gb total)

I have a Mellanox card in this one for 10g/40g networking along with a 
bunch of SAS/SATA expansion as well; it gets fairly heavy use as a "hot 
standby" Postgres database box, general file service, video server and 
as a build system as well for various distributions I use in other 
contexts.

Check the RAM part numbers against what Supermicro specifies as 
"approved"; these boards are extremely picky in that regard but 
typically if you have the wrong part numbers compared with what they 
want they will refuse to POST straight up rather than do hinky stuff.

This board has been in service here for quite a long time (many years); 
the most-recent BIOS is what I am running I believe; I've not had any 
trouble with ECC errors being logged nor any sort of data corruption, 
crashes or other misbehavior that could be attributed to that sort of issue.

I'm extremely allergic to RAM issues due to the machine supporting a 
quite-large amount of storage on ZFS and the need for it to be 
rock-solid reliable.  It is.

The CPU I have in mine is a E-2146G.

I'm on 14.3-STABLE (compiled from source fairly recently as I keep up 
with security and potential driver issues that might impact me.)

-- 
Karl Denninger
[email protected]
/The Market Ticker/
/[S/MIME encrypted email preferred]/