RE: fsck mdadm

Hi

Thanks Grant.

I have try :
for i in c d e f g; do dd if=/dev/sd${i} of=/dev/null bs=1M count=500; done
500+0 enregistrements lus
500+0 enregistrements écrits
524288000 octets (524 MB, 500 MiB) copiés, 2,23716 s, 234 MB/s
500+0 enregistrements lus
500+0 enregistrements écrits
524288000 octets (524 MB, 500 MiB) copiés, 3,12926 s, 168 MB/s
500+0 enregistrements lus
500+0 enregistrements écrits
524288000 octets (524 MB, 500 MiB) copiés, 1,04179 s, 503 MB/s
500+0 enregistrements lus
500+0 enregistrements écrits
524288000 octets (524 MB, 500 MiB) copiés, 1,15904 s, 452 MB/s
500+0 enregistrements lus
500+0 enregistrements écrits
524288000 octets (524 MB, 500 MiB) copiés, 1,16094 s, 452 MB/s

Yes, after reboot I have the problem. My md device was degraded. Sde go out, I have change sata cable. But now, raid volume is ok :
md5 : active raid5 sde1[9] sdd1[8] sdg1[5] sdf1[6] sdc1[7]
      15627542528 blocks super 1.2 level 5, 512k chunk, algorithm 2 [5/5] [UUUUU]

But, I had to disable fsck at boot :
/dev/md5                /mnt/9To        ext4            noatime         0 0

At each reboot, I got this error :
Error writing file system info: Input/output error

/dev/mapper/VolGroupArray-lvmedia: ** FILE SYSTEM WAS MODIFIED **

I try offline fsck.

Thanks for help

Regards,

Erwan RIGOLLOT

-----Message d'origine-----
De : Grant Taylor <[email protected]> 
Envoyé : mardi 10 mars 2026 03:18
À : [email protected]
Objet : Re: [gentoo-user] fsck mdadm

On 3/9/26 6:40 AM, Erwan RIGOLLOT wrote:
> Hello all,

Hi,

> md5 : active raid5 sde1[9] sdd1[8] sdg1[5] sdf1[6] sdc1[7]
>        15627542528 blocks super 1.2 level 5, 512k chunk, algorithm 2 
> [5/5] [UUUUU]
> 
> (My problem is about MD5)

It appears as if md5 is happy (enough).

> for i in c d e f g; do /usr/sbin/smartctl -a /dev/sd$i | grep 
> "Serial\|^# 1\|

Try reading 1 MB ~ 1 GB from each of the disks.

for i in c d e f g; do dd if=/dev/sd${i} of=/dev/null bs=1M count=1; done

Raise count to read more than 1 MB from each drive.

I've had some really weird almost silent failures wherein the drive would respond to the kernel for anything but data.  Trying to read data, especially more than a few (hundred) MB almost certainly exceeds any cache and thus forces a read from disk.

If you can really read from the persistent media, then you know that you don't have that particular problem.

> So, my five disks seem to be ok.
Were you trying to do an online / mounted or offline / unmounted fsck?

I usually balk at this request, but I'd be remiss if I didn't ask; does the problem continue after a reboot?



-- 
Grant. . . .