why disk errors are more likely to happen on md super block area?

why disk errors are more likely to happen on md super block area?

[hank peng]

Hi, guys:
I found that when disk errors happened, it is more likely to locate in
md super block area. Kernel messages are like this:
end_request: I/O error, dev sda, sector 1953525000
md: super_written gets error=-5, uptodate=0
In some cases, I check if disk is really failed using 'dd' command and
found that it is OK for read and write on those sectors.

I know that md super block information is located in the end of a
disk, but is it related with this problem? Is it safer to put md super
block in the start area of a disk?


-- 
The simplest is not all best but the best is surely the simplest!

Re: why disk errors are more likely to happen on md super block area?

[Roberto Spadim]

i think you will have no problems with location of metadata, there's
some versions of metadata, at start at end.. check your mdadm --create
command
what information at metadata is/was wrong? can you get it? or just i/o error?
some months ago i had two broken harddisks, the disk start to read
from position 0 to position X without problem, but when it get at a
position where disk was crashed (real crashed, the disk hit the floor)
the array start to resync (maybe a disk report) it was a raid10 array.
maybe you have problem at physical layer (disk with problem).
i don't know what informations md update at metadata, but i think that
there's some last write counter/last write timestamp to sync what
mirrors is more recent in a broken array start, (this information can
be found with others developers in this list)

2011/3/23 hank peng <pengxihan@gmail.com>:
> Hi, guys:
> I found that when disk errors happened, it is more likely to locate in
> md super block area. Kernel messages are like this:
> end_request: I/O error, dev sda, sector 1953525000
> md: super_written gets error=-5, uptodate=0
> In some cases, I check if disk is really failed using 'dd' command and
> found that it is OK for read and write on those sectors.
>
> I know that md super block information is located in the end of a
> disk, but is it related with this problem? Is it safer to put md super
> block in the start area of a disk?
>
>
> --
> The simplest is not all best but the best is surely the simplest!
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>



-- 
Roberto Spadim
Spadim Technology / SPAEmpresarial
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: why disk errors are more likely to happen on md super block area?

[hank peng]

2011/3/24 Roberto Spadim <roberto@spadim.com.br>:
> i think you will have no problems with location of metadata, there's
> some versions of metadata, at start at end.. check your mdadm --create
> command

yes, I found that option. If I choose to put metadata on the start
area of a disk, disk's head seek time will be decreased, right?

So, is there someone in this list to tell me what is real difference
when putting meta data on start or end of a disk?

> what information at metadata is/was wrong? can you get it? or just i/o error?
> some months ago i had two broken harddisks, the disk start to read
> from position 0 to position X without problem, but when it get at a
> position where disk was crashed (real crashed, the disk hit the floor)
> the array start to resync (maybe a disk report) it was a raid10 array.
> maybe you have problem at physical layer (disk with problem).
> i don't know what informations md update at metadata, but i think that
> there's some last write counter/last write timestamp to sync what
> mirrors is more recent in a broken array start, (this information can
> be found with others developers in this list)
>
> 2011/3/23 hank peng <pengxihan@gmail.com>:
>> Hi, guys:
>> I found that when disk errors happened, it is more likely to locate in
>> md super block area. Kernel messages are like this:
>> end_request: I/O error, dev sda, sector 1953525000
>> md: super_written gets error=-5, uptodate=0
>> In some cases, I check if disk is really failed using 'dd' command and
>> found that it is OK for read and write on those sectors.
>>
>> I know that md super block information is located in the end of a
>> disk, but is it related with this problem? Is it safer to put md super
>> block in the start area of a disk?
>>
>>
>> --
>> The simplest is not all best but the best is surely the simplest!
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
>
>
>
> --
> Roberto Spadim
> Spadim Technology / SPAEmpresarial
>



-- 
The simplest is not all best but the best is surely the simplest!
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: why disk errors are more likely to happen on md super block area?

[Roberto Spadim]

humm head seek time? only testing to know if it will be decreased or not
the idea is: if md update metadata veryoften you should consider put
most i/o files near metadata sectors to don't lose time with head
positioning

real difference of metadata position?
maybe some bootloaders can't boot with md metadata at start/end of
disk/partition.
if you don't use this md device to boot, you don't need to worry about
it, any position will work, i don't know about growing array size
problems about meta data positioning, but i think you don't need to
worry about this too

2011/3/24 hank peng <pengxihan@gmail.com>:
> 2011/3/24 Roberto Spadim <roberto@spadim.com.br>:
>> i think you will have no problems with location of metadata, there's
>> some versions of metadata, at start at end.. check your mdadm --create
>> command
>
> yes, I found that option. If I choose to put metadata on the start
> area of a disk, disk's head seek time will be decreased, right?
>
> So, is there someone in this list to tell me what is real difference
> when putting meta data on start or end of a disk?
>
>> what information at metadata is/was wrong? can you get it? or just i/o error?
>> some months ago i had two broken harddisks, the disk start to read
>> from position 0 to position X without problem, but when it get at a
>> position where disk was crashed (real crashed, the disk hit the floor)
>> the array start to resync (maybe a disk report) it was a raid10 array.
>> maybe you have problem at physical layer (disk with problem).
>> i don't know what informations md update at metadata, but i think that
>> there's some last write counter/last write timestamp to sync what
>> mirrors is more recent in a broken array start, (this information can
>> be found with others developers in this list)
>>
>> 2011/3/23 hank peng <pengxihan@gmail.com>:
>>> Hi, guys:
>>> I found that when disk errors happened, it is more likely to locate in
>>> md super block area. Kernel messages are like this:
>>> end_request: I/O error, dev sda, sector 1953525000
>>> md: super_written gets error=-5, uptodate=0
>>> In some cases, I check if disk is really failed using 'dd' command and
>>> found that it is OK for read and write on those sectors.
>>>
>>> I know that md super block information is located in the end of a
>>> disk, but is it related with this problem? Is it safer to put md super
>>> block in the start area of a disk?
>>>
>>>
>>> --
>>> The simplest is not all best but the best is surely the simplest!
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
>>> the body of a message to majordomo@vger.kernel.org
>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>
>>
>>
>>
>> --
>> Roberto Spadim
>> Spadim Technology / SPAEmpresarial
>>
>
>
>
> --
> The simplest is not all best but the best is surely the simplest!
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>



-- 
Roberto Spadim
Spadim Technology / SPAEmpresarial
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html