Buffer I/O error on device sdd1, logical block, HDD failure? [SOLVED]

greenfish · 2013-10-21 10:04:23

Hello!

I'm a bit puzzled here to be honest. Granted i'm not using linux as much as I used to (not after windows 7). I have archlinux running on my HTPC. Never had any issues before that was this severe, unless I upgraded and forgot to read news section. Booted the htpc today to be greeted by "Buffer I/O error on device sdd1, logical block" with a massive wall of text, a few seconds later "welcome to emergency mode."

*This is NOT hdd where the linux kernel is residing on. What logical purpose would it serve for the kernel/userspace to abort everything just because fsck fails or something? If this was indeed my linux partition I would fully understand.

Anyways, I used parted magic, ran fsck, smart. Sure enough fsck warned me about bad/missing superblock. Restored the superblock by using e2fsck. I had over 10 000 "incorrect size" chunks. Ran 2-3 SMART after that. fsck says okay, smart gives a 100% status report with no errors.

Oh yeah, I have turned off FSCK completly in my fstab, thinking about at least turning it on my bigger hdds

Questions:

*Is SMART reliable? If it says it's alright, does that mean i'm safe? Would physical broken sectors turn up by SMART?
*I know SMART warns the user in windows 7 if hdd failure is imment. Is this possible within linux as well? Since i'm NOT using a GUI, is this possible to send through a terminal/email?

*Sometimes the HTPC have been forefully shut down (power breakage), could this be one of the causes of the I/O error?

As always, thank you for your support.

Last edited by greenfish (2013-10-23 13:23:21)

graysky · 2013-10-21 13:52:22

Any reallocated sectors in smartmontools? If you run 'e2fsck -fv /dev/sdd1' does it complete wo/ errors? Probably best to repeat for all linux partitions on that disk.

alphaniner · 2013-10-21 14:19:03

greenfish wrote:

*Is SMART reliable? If it says it's alright, does that mean i'm safe? Would physical broken sectors turn up by SMART?
*I know SMART warns the user in windows 7 if hdd failure is imment. Is this possible within linux as well? Since i'm NOT using a GUI, is this possible to send through a terminal/email?

1) Don't trust the 'SMART overall-health self-assessment test result', run the diagnostics (short, long, conveyance, offline). The short and conveyance tests are quick so start with them. If they both pass run the long test. The offline test is supposed to update SMART attributes, but it generally takes longer than the long test, so save it for last if at all. Usually when I see bad drives the short or long tests pick them up.

2) Look into smartd.service.

greenfish wrote:

What logical purpose would it serve for the kernel/userspace to abort everything just because fsck fails or something?

Systemd craps itself if an fs configured to mount during boot can't be mounted, even if the fs isn't necessary for the system to boot. Rot sure about how it handles fsck failures. This 'feature' can be disabled by putting nofail in the fstab options. I add it to every non-essential automounting fs.

Last edited by alphaniner (2013-10-21 14:20:33)

greenfish · 2013-10-23 13:23:02

graysky wrote:

Any reallocated sectors in smartmontools? If you run 'e2fsck -fv /dev/sdd1' does it complete wo/ errors? Probably best to repeat for all linux partitions on that disk.

Sorry for the late reply guys. Been busy with my other hdd that decided to screw with me. e2fsck first complained about bad sectors, and wrong size. Now it says all clean. I've decided to remove this HDD from server and mark it "damaged".

Thank you again for your help

alphaniner wrote:

greenfish wrote:
*Is SMART reliable? If it says it's alright, does that mean i'm safe? Would physical broken sectors turn up by SMART?
*I know SMART warns the user in windows 7 if hdd failure is imment. Is this possible within linux as well? Since i'm NOT using a GUI, is this possible to send through a terminal/email?
1) Don't trust the 'SMART overall-health self-assessment test result', run the diagnostics (short, long, conveyance, offline). The short and conveyance tests are quick so start with them. If they both pass run the long test. The offline test is supposed to update SMART attributes, but it generally takes longer than the long test, so save it for last if at all. Usually when I see bad drives the short or long tests pick them up.
2) Look into smartd.service.
greenfish wrote:
What logical purpose would it serve for the kernel/userspace to abort everything just because fsck fails or something?
Systemd craps itself if an fs configured to mount during boot can't be mounted, even if the fs isn't necessary for the system to boot. Rot sure about how it handles fsck failures. This 'feature' can be disabled by putting nofail in the fstab options. I add it to every non-essential automounting fs.

Thank you for the useful information. I will save this post for future references.

Will deff look into smartd.service, especially when I have so much data running 24/7.

Will also update my fstab table with "nofail" like you suggested

Thank You!

Arch Linux

#1 2013-10-21 10:04:23

Buffer I/O error on device sdd1, logical block, HDD failure? [SOLVED]

#2 2013-10-21 13:52:22

Re: Buffer I/O error on device sdd1, logical block, HDD failure? [SOLVED]

#3 2013-10-21 14:19:03

Re: Buffer I/O error on device sdd1, logical block, HDD failure? [SOLVED]

#4 2013-10-23 13:23:02

Re: Buffer I/O error on device sdd1, logical block, HDD failure? [SOLVED]

Board footer