memory errors via APEI

[kuznetcoff777]

Hi!
Exporter has some memory error monitoring

# TYPE node_edac_correctable_errors_total counter
# TYPE node_edac_csrow_correctable_errors_total counter
# TYPE node_edac_csrow_uncorrectable_errors_total counter
# TYPE node_edac_uncorrectable_errors_total counter

But some systems has no edac counters, for example with APEI (ACPI Platform Error Interface)

Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]: Hardware error from APEI Generic Hardware Error Source: 4
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]: It has been corrected by h/w and requires no further action
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]: event severity: corrected
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]:  Error 0, type: corrected
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]:  fru_text: D2
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]:   section_type: memory error
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]:    error_status: Storage error in DRAM memory (0x0000000000000400)
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]:   physical_address: 0x0000006a47d1f540
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]:   node:6 card:1 module:0 rank:1 bank:2 device:11 row:40948 column:560
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]:   error_type: 2, single-bit ECC
Dec 29 05:24:23 proxmox kernel: {35}[Hardware Error]:   DIMM location: not present. DMI handle: 0x0000
Dec 29 05:24:23 proxmox kernel: mce: [Hardware Error]: Machine check events logged

What about to add counter that discovered from journactl? At least that some "mce: [Hardware Error]:" appeared?

[SuperQ]

No, the node_exporter doesn't produce metrics from dmesg/logs.

If these errors can be read from /proc or /sys, or some other system call we can expose them.

[kuznetcoff777]

As i undestand APEI doesnt write to metrics, except of journal, its a feature, not a bug)