recommending RAID6 over RAID5 when doing mdadm --create

recommending RAID6 over RAID5 when doing mdadm --create

[Mikael Abrahamsson]

Hi,

I believe this has been discussed before, but I would like to bring it up 
again.

I would like to see mdadm recommend RAID6 over RAID6 when using component 
drives that are 500GB or larger and ask if the user is really sure about 
using RAID5, plus a pointer to a webpage on the wiki informing the user 
about the trouble with RAID5 on large volumes with consumer or prosumer 
drives.

I am willing to contribute text to the webpage being pointed to.

-- 
Mikael Abrahamsson    email: swmike@swm.pp.se

Re: recommending RAID6 over RAID5 when doing mdadm --create

[Tommy Apel]

2013/11/11 Mikael Abrahamsson <swmike@swm.pp.se>:
>
> Hi,
>
> I believe this has been discussed before, but I would like to bring it up
> again.
>
> I would like to see mdadm recommend RAID6 over RAID6 when using component
> drives that are 500GB or larger and ask if the user is really sure about
> using RAID5, plus a pointer to a webpage on the wiki informing the user
> about the trouble with RAID5 on large volumes with consumer or prosumer
> drives.
>
> I am willing to contribute text to the webpage being pointed to.
>
> --
> Mikael Abrahamsson    email: swmike@swm.pp.se
> --
> 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

Hello,

May I suggest that aswell for arrays with more than 5 drives aswell ?

-- 

/Tommy

Re: recommending RAID6 over RAID5 when doing mdadm --create

[Stan Hoeppner]

On 11/11/2013 1:44 AM, Mikael Abrahamsson wrote:

> I believe this has been discussed before, but I would like to bring it
> up again.
> 
> I would like to see mdadm recommend RAID6 over RAID6 when using
> component drives that are 500GB or larger and ask if the user is really
> sure about using RAID5, plus a pointer to a webpage on the wiki
> informing the user about the trouble with RAID5 on large volumes with
> consumer or prosumer drives.
> 
> I am willing to contribute text to the webpage being pointed to.

Wikis change to regularly, too easily.  Networks can be down at times,
often the worst times.  See Murphy's Law.

It would be better to put this information in mdadm(8) and refer the
user there.

-- 
Stan

Re: recommending RAID6 over RAID5 when doing mdadm --create

[Tommy Apel]

2013/11/11 Stan Hoeppner <stan@hardwarefreak.com>:
> On 11/11/2013 1:44 AM, Mikael Abrahamsson wrote:
>
>> I believe this has been discussed before, but I would like to bring it
>> up again.
>>
>> I would like to see mdadm recommend RAID6 over RAID6 when using
>> component drives that are 500GB or larger and ask if the user is really
>> sure about using RAID5, plus a pointer to a webpage on the wiki
>> informing the user about the trouble with RAID5 on large volumes with
>> consumer or prosumer drives.
>>
>> I am willing to contribute text to the webpage being pointed to.
>
> Wikis change to regularly, too easily.  Networks can be down at times,
> often the worst times.  See Murphy's Law.
>
> It would be better to put this information in mdadm(8) and refer the
> user there.
>
> --
> Stan
>
>
>
>
>
> --
> 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

Good point Stan,
maybe even "are you sure you really want to do this? (see mdadm(8) for
further information)" when creating the array

-- 

/Tommy

Re: recommending RAID6 over RAID5 when doing mdadm --create

[joystick]

On 11/11/2013 10:37, Tommy Apel wrote:
> 2013/11/11 Stan Hoeppner <stan@hardwarefreak.com>:
>> On 11/11/2013 1:44 AM, Mikael Abrahamsson wrote:
>>
>>> I believe this has been discussed before, but I would like to bring it
>>> up again.
>>>
>>> I would like to see mdadm recommend RAID6 over RAID6 when using
>>> component drives that are 500GB or larger and ask if the user is really
>>> sure about using RAID5, plus a pointer to a webpage on the wiki
>>> informing the user about the trouble with RAID5 on large volumes with
>>> consumer or prosumer drives.
>>>
>>> I am willing to contribute text to the webpage being pointed to.
>> Wikis change to regularly, too easily.  Networks can be down at times,
>> often the worst times.  See Murphy's Law.
>>
>> It would be better to put this information in mdadm(8) and refer the
>> user there.
>>
>> --
>> Stan
> Good point Stan,
> maybe even "are you sure you really want to do this? (see mdadm(8) for
> further information)" when creating the array
>

Currently raid6 does not have an implementation for shortcut rmw, so 
there is at least one reason to do raid5, and it's not easy to guess for 
the average user.

Putting a yes/no confirmation seems exxagerate to me, that would be on 
the same warning level of "partition contains metadata of an existing 
array", which is a MUCH more serious error message.

The user would be then tempted to use "yes" in front of mdadm create (*) 
when copy-pasting lines, and then possibly blow up an existing RAID 
without chance to see the confirmation prompt.

(*) I prepare mdadm create lines on a text files and then paste them on 
systems, usually many create commands together, I think most people do 
the same, because mdadm create lines are too complex to be typed 
in-place. If the create commands are to be scripted for any reason, it's 
worse, in this case almost certainly the user would use "yes".

I think a simple warning message without confirmation prompt is fair. 
The user can repeat the command with raid6 after reading.

Regards
J.

Re: recommending RAID6 over RAID5 when doing mdadm --create

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 1016 bytes --]

On Mon, 11 Nov 2013 02:56:37 -0600 Stan Hoeppner <stan@hardwarefreak.com>
wrote:

> On 11/11/2013 1:44 AM, Mikael Abrahamsson wrote:
> 
> > I believe this has been discussed before, but I would like to bring it
> > up again.
> > 
> > I would like to see mdadm recommend RAID6 over RAID6 when using
> > component drives that are 500GB or larger and ask if the user is really
> > sure about using RAID5, plus a pointer to a webpage on the wiki
> > informing the user about the trouble with RAID5 on large volumes with
> > consumer or prosumer drives.
> > 
> > I am willing to contribute text to the webpage being pointed to.
> 
> Wikis change to regularly, too easily.  Networks can be down at times,
> often the worst times.  See Murphy's Law.
> 
> It would be better to put this information in mdadm(8) and refer the
> user there.
> 

.. or in md(4) maybe.

If anyone wants to run with this and send me a patch (for code and
documentation) I would be very likely to apply it.

NeilBrown

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: recommending RAID6 over RAID5 when doing mdadm --create

[Giovanni Tessore]

On 11/12/2013 03:52 AM, NeilBrown wrote:
> On Mon, 11 Nov 2013 02:56:37 -0600 Stan Hoeppner <stan@hardwarefreak.com>
> wrote:
>
>> On 11/11/2013 1:44 AM, Mikael Abrahamsson wrote:
>>
>>> I believe this has been discussed before, but I would like to bring it
>>> up again.
>>>
>>> I would like to see mdadm recommend RAID6 over RAID6 when using
>>> component drives that are 500GB or larger and ask if the user is really
>>> sure about using RAID5, plus a pointer to a webpage on the wiki
>>> informing the user about the trouble with RAID5 on large volumes with
>>> consumer or prosumer drives.
>>>
>>> I am willing to contribute text to the webpage being pointed to.
>> Wikis change to regularly, too easily.  Networks can be down at times,
>> often the worst times.  See Murphy's Law.
>>
>> It would be better to put this information in mdadm(8) and refer the
>> user there.
>>
> .. or in md(4) maybe.
>
> If anyone wants to run with this and send me a patch (for code and
> documentation) I would be very likely to apply it.
>
> NeilBrown

That issue (RAID5 going fully down during resync of degraded array if 
read errors occur) hasn't been addressed by the badblocks log feature 
available in later kernels and mdadm ?
Yet, it is not clear to me if the fetaure is fully implemented and 
operational.

Regards
Giovanni