Re: NetBSD 9.1 upgrade and file system crash - reboot fails

[mlelstv%serpens.de@localhost (Michael van Elst)]

riccardo.mottola%libero.it@localhost (Riccardo Mottola) writes:

>> Also use atactl to check the smart status of the disk.
>How reliable is that data?

It's what the manufacturer tells you.

>   1  58   34     yes online  positive    Raw read error rate         27218486
>195  58    0     no  online  positive    Hardware ECC Recovered       27218486

>   1  59   34     yes online  positive    Raw read error rate         232650323
>195  59    0     no  online  positive    Hardware ECC Recovered       232650323

>   1  60   34     yes online  positive    Raw read error rate         73875073
>195  60    0     no  online  positive    Hardware ECC Recovered       73875073

>The number of read errors skyrocketed!

There is no number of read errors given. The raw values only have
a vendor specific meaning, can be anything. Some vendors document
the raw values.

You have a value of 58, then 59, then 60 and a threshold of 34.
When the value reaches the treshold (from above!) or sinks below,
the drive will flag an error.

smartmontools can show some more details, e.g.:

ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE
  1 Raw_Read_Error_Rate     0x000f   113   099   006    Pre-fail  Always       -       58007672
195 Hardware_ECC_Recovered  0x001a   027   027   000    Old_age   Always       -       58007672

This also shows some undecipherable raw value, but also a "worst
case" datum for the drive.

So, the raw read error rate isn't a count, but some kind of average
for how frequent read errors occurred. And apparently the numbers
of read errors and ECC recovery events seem to be correlated.

Modern drives produce a huge amount of correctable ECC errors and
some report such confusing SMART data.

-- 
-- 
                                Michael van Elst
Internet: mlelstv%serpens.de@localhost
                                "A potential Snark may lurk in every tree."