Good hardware for mdadm

Good hardware for mdadm

[Paul van der Vlis]

Hello,

I like mdadm and I am using it many years. But in my opinion it has one
disadvantage: when the MBR of the boot-disk is corrupt, the machine will
not boot.

A bios could check this. Wait for some kind of signal from Grub or
Linux, and after a timeout boot from another disk. But I don't know
about a bios with that feature.

A PCIe card could do something like that, but I don't know about such a
PCIe card.

Is there such hardware?
What do you do to avoid this problem?

With regards,
Paul van der Vlis.




-- 
Paul van der Vlis Linux systeembeheer, Groningen
http://www.vandervlis.nl/

Re: Good hardware for mdadm

[Roberto Spadim]

i use two or more boot disks, if the first raid1 disk don't boot bios
go to second boot disk, third, etc etc, you must write grub to mbr of
each disk
i'm using dell server r410 if i'm not wrong

2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
> Hello,
>
> I like mdadm and I am using it many years. But in my opinion it has one
> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
> not boot.
>
> A bios could check this. Wait for some kind of signal from Grub or
> Linux, and after a timeout boot from another disk. But I don't know
> about a bios with that feature.
>
> A PCIe card could do something like that, but I don't know about such a
> PCIe card.
>
> Is there such hardware?
> What do you do to avoid this problem?
>
> With regards,
> Paul van der Vlis.
>
>
>
>
> --
> Paul van der Vlis Linux systeembeheer, Groningen
> http://www.vandervlis.nl/
>
> --
> 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

Re: Good hardware for mdadm

[Paul van der Vlis]

Hi Roberto,

op 29-08-14 22:44, Roberto Spadim schreef:
> i use two or more boot disks, if the first raid1 disk don't boot bios
> go to second boot disk, third, etc etc, 

In my opinion this only works when the boot-disk is completely defect or
removed. Not when the data in the MBR on that boot-disk is corrupt.

Or did you test this, or do you have other reasons to believe that your
bios will handle this correct?

With the boot-disk I mean the disk what's in the bios the first disk. So
this could also be the second raid1 disk. Or an USB stick.

> you must write grub to mbr of each disk

Of course.

With regards,
Paul van der Vlis.

> i'm using dell server r410 if i'm not wrong
> 
> 2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>> Hello,
>>
>> I like mdadm and I am using it many years. But in my opinion it has one
>> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
>> not boot.
>>
>> A bios could check this. Wait for some kind of signal from Grub or
>> Linux, and after a timeout boot from another disk. But I don't know
>> about a bios with that feature.
>>
>> A PCIe card could do something like that, but I don't know about such a
>> PCIe card.
>>
>> Is there such hardware?
>> What do you do to avoid this problem?
>>
>> With regards,
>> Paul van der Vlis.
>>
>>
>>
>>
>> --
>> Paul van der Vlis Linux systeembeheer, Groningen
>> http://www.vandervlis.nl/
>>
>> --
>> 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
> 
> 
> 





-- 
Paul van der Vlis Linux systeembeheer, Groningen
http://www.vandervlis.nl/

Re: Good hardware for mdadm

[Adam Talbot]

I have found placing /boot + MBR on a DOM (Disk on module) to be very
effective.

Perhaps we should be asking how the MBR got broken in the first place?

On Fri, Aug 29, 2014 at 2:47 PM, Paul van der Vlis <paul@vandervlis.nl> wrote:
> Hi Roberto,
>
> op 29-08-14 22:44, Roberto Spadim schreef:
>> i use two or more boot disks, if the first raid1 disk don't boot bios
>> go to second boot disk, third, etc etc,
>
> In my opinion this only works when the boot-disk is completely defect or
> removed. Not when the data in the MBR on that boot-disk is corrupt.
>
> Or did you test this, or do you have other reasons to believe that your
> bios will handle this correct?
>
> With the boot-disk I mean the disk what's in the bios the first disk. So
> this could also be the second raid1 disk. Or an USB stick.
>
>> you must write grub to mbr of each disk
>
> Of course.
>
> With regards,
> Paul van der Vlis.
>
>> i'm using dell server r410 if i'm not wrong
>>
>> 2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>>> Hello,
>>>
>>> I like mdadm and I am using it many years. But in my opinion it has one
>>> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
>>> not boot.
>>>
>>> A bios could check this. Wait for some kind of signal from Grub or
>>> Linux, and after a timeout boot from another disk. But I don't know
>>> about a bios with that feature.
>>>
>>> A PCIe card could do something like that, but I don't know about such a
>>> PCIe card.
>>>
>>> Is there such hardware?
>>> What do you do to avoid this problem?
>>>
>>> With regards,
>>> Paul van der Vlis.
>>>
>>>
>>>
>>>
>>> --
>>> Paul van der Vlis Linux systeembeheer, Groningen
>>> http://www.vandervlis.nl/
>>>
>>> --
>>> 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
>>
>>
>>
>
>
>
>
>
> --
> Paul van der Vlis Linux systeembeheer, Groningen
> http://www.vandervlis.nl/
>
> --
> 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: Good hardware for mdadm

[Roger Heflin]

No bios I have seen automatically handles it unless the disks are
total dead, some of the hardware raid controlers may properly deal
with a disk returning an error and go to the other mirror, the low end
quality raid controllers start around $200.

Outside of the raid controllers I don't believe any of the bioses
themselves will do it, so you are unlikely to find any hardware that
does that.

Likely he is talking about putting the boot block on multiple disks
and and use the bbr menu or the bios boot disk selection to pick
another disk.

On Fri, Aug 29, 2014 at 4:47 PM, Paul van der Vlis <paul@vandervlis.nl> wrote:
> Hi Roberto,
>
> op 29-08-14 22:44, Roberto Spadim schreef:
>> i use two or more boot disks, if the first raid1 disk don't boot bios
>> go to second boot disk, third, etc etc,
>
> In my opinion this only works when the boot-disk is completely defect or
> removed. Not when the data in the MBR on that boot-disk is corrupt.
>
> Or did you test this, or do you have other reasons to believe that your
> bios will handle this correct?
>
> With the boot-disk I mean the disk what's in the bios the first disk. So
> this could also be the second raid1 disk. Or an USB stick.
>
>> you must write grub to mbr of each disk
>
> Of course.
>
> With regards,
> Paul van der Vlis.
>
>> i'm using dell server r410 if i'm not wrong
>>
>> 2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>>> Hello,
>>>
>>> I like mdadm and I am using it many years. But in my opinion it has one
>>> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
>>> not boot.
>>>
>>> A bios could check this. Wait for some kind of signal from Grub or
>>> Linux, and after a timeout boot from another disk. But I don't know
>>> about a bios with that feature.
>>>
>>> A PCIe card could do something like that, but I don't know about such a
>>> PCIe card.
>>>
>>> Is there such hardware?
>>> What do you do to avoid this problem?
>>>
>>> With regards,
>>> Paul van der Vlis.
>>>
>>>
>>>
>>>
>>> --
>>> Paul van der Vlis Linux systeembeheer, Groningen
>>> http://www.vandervlis.nl/
>>>
>>> --
>>> 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
>>
>>
>>
>
>
>
>
>
> --
> Paul van der Vlis Linux systeembeheer, Groningen
> http://www.vandervlis.nl/
>
> --
> 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: Good hardware for mdadm

[Roberto Spadim]

i'm considering that you are using x86, x86_64 hardware, what's your
application?
there's no bios test (must check if the newer version of this server
have uefi and work differente), it just don't boot the disk and go to
next boot
but for your question: yes... maybe with 'lucky' it can't boot, a
"bad" (crashed) grub could stop server startup

maybe others questions that you could check before solving this boot
problem is why use a "bad" disk?
why shutdown a server?
why remote startup a server?
why use a 'bad' disk?
if it's a ha system, include a human to check monthly, or something
like it, if the server is ok clean,something like it, since ha
probably include a real time application or a security or risk
application, or anything like it

probably you must check others solutions at hardware level, some
computers have dual bios (yes if one bios is lost you have the other
to boot the computer), maybe an openbios could help, or a uefi bios,
must check what's better


2014-08-29 18:47 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
> Hi Roberto,
>
> op 29-08-14 22:44, Roberto Spadim schreef:
>> i use two or more boot disks, if the first raid1 disk don't boot bios
>> go to second boot disk, third, etc etc,
>
> In my opinion this only works when the boot-disk is completely defect or
> removed. Not when the data in the MBR on that boot-disk is corrupt.
>
> Or did you test this, or do you have other reasons to believe that your
> bios will handle this correct?
>
> With the boot-disk I mean the disk what's in the bios the first disk. So
> this could also be the second raid1 disk. Or an USB stick.
>
>> you must write grub to mbr of each disk
>
> Of course.
>
> With regards,
> Paul van der Vlis.
>
>> i'm using dell server r410 if i'm not wrong
>>
>> 2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>>> Hello,
>>>
>>> I like mdadm and I am using it many years. But in my opinion it has one
>>> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
>>> not boot.
>>>
>>> A bios could check this. Wait for some kind of signal from Grub or
>>> Linux, and after a timeout boot from another disk. But I don't know
>>> about a bios with that feature.
>>>
>>> A PCIe card could do something like that, but I don't know about such a
>>> PCIe card.
>>>
>>> Is there such hardware?
>>> What do you do to avoid this problem?
>>>
>>> With regards,
>>> Paul van der Vlis.
>>>
>>>
>>>
>>>
>>> --
>>> Paul van der Vlis Linux systeembeheer, Groningen
>>> http://www.vandervlis.nl/
>>>
>>> --
>>> 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
>>
>>
>>
>
>
>
>
>
> --
> Paul van der Vlis Linux systeembeheer, Groningen
> http://www.vandervlis.nl/
>
> --
> 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

Re: Good hardware for mdadm

[Paul van der Vlis]

op 29-08-14 23:52, Adam Talbot schreef:
> I have found placing /boot + MBR on a DOM (Disk on module) to be very
> effective.

Maybe a good DOM is more reliable then a good harddisk. But it's not
redundant.

> Perhaps we should be asking how the MBR got broken in the first place?

I am using mdadm about 12 years, and I've seen many times that a system
did not come up after booting. And when I removed the boot-disk, it did
come up.

I must say it's less then for a few years, maybe I am using better
disks. But I am looking for something what's better.

What helps, is that there is something like IPMI. So when there is a
problem and IPMI is working fine, you can boot by hand from another disk.

With regards,
Paul van der Vlis.


> On Fri, Aug 29, 2014 at 2:47 PM, Paul van der Vlis <paul@vandervlis.nl> wrote:
>> Hi Roberto,
>>
>> op 29-08-14 22:44, Roberto Spadim schreef:
>>> i use two or more boot disks, if the first raid1 disk don't boot bios
>>> go to second boot disk, third, etc etc,
>>
>> In my opinion this only works when the boot-disk is completely defect or
>> removed. Not when the data in the MBR on that boot-disk is corrupt.
>>
>> Or did you test this, or do you have other reasons to believe that your
>> bios will handle this correct?
>>
>> With the boot-disk I mean the disk what's in the bios the first disk. So
>> this could also be the second raid1 disk. Or an USB stick.
>>
>>> you must write grub to mbr of each disk
>>
>> Of course.
>>
>> With regards,
>> Paul van der Vlis.
>>
>>> i'm using dell server r410 if i'm not wrong
>>>
>>> 2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>>>> Hello,
>>>>
>>>> I like mdadm and I am using it many years. But in my opinion it has one
>>>> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
>>>> not boot.
>>>>
>>>> A bios could check this. Wait for some kind of signal from Grub or
>>>> Linux, and after a timeout boot from another disk. But I don't know
>>>> about a bios with that feature.
>>>>
>>>> A PCIe card could do something like that, but I don't know about such a
>>>> PCIe card.
>>>>
>>>> Is there such hardware?
>>>> What do you do to avoid this problem?
>>>>
>>>> With regards,
>>>> Paul van der Vlis.
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Paul van der Vlis Linux systeembeheer, Groningen
>>>> http://www.vandervlis.nl/
>>>>
>>>> --
>>>> 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
>>>
>>>
>>>
>>
>>
>>
>>
>>
>> --
>> Paul van der Vlis Linux systeembeheer, Groningen
>> http://www.vandervlis.nl/
>>
>> --
>> 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
> --
> 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
> 





-- 
Paul van der Vlis Linux systeembeheer, Groningen
http://www.vandervlis.nl/

Re: Good hardware for mdadm

[Paul van der Vlis]

op 30-08-14 00:38, Roger Heflin schreef:
> No bios I have seen automatically handles it unless the disks are
> total dead, 

That's my experience too.

> some of the hardware raid controlers may properly deal
> with a disk returning an error and go to the other mirror, the low end
> quality raid controllers start around $200.

Hardware raid controllers have other disadvantages, like using a
proprietary format for storing information on disk. And in many cases
you have to use tools from the manufacturer. That are reasons for me to
like mdadm.

Maybe it could be done with a hardware raid controller in JBOD mode.

( And hardware controllers are not perfect too, I've seen simular
problems with an HP smart array. First when I've removed the first
harddisk the system did boot from the second. )

> Outside of the raid controllers I don't believe any of the bioses
> themselves will do it, so you are unlikely to find any hardware that
> does that.

It could be done with a PCIe card too. Some kind of fake-raid adapter
what uses mdadm.

> Likely he is talking about putting the boot block on multiple disks
> and and use the bbr menu or the bios boot disk selection to pick
> another disk.

Not sure what a "bbr menu" is. But realize that the servers I use are
remote.

With regards,
Paul van der Vlis.

> On Fri, Aug 29, 2014 at 4:47 PM, Paul van der Vlis <paul@vandervlis.nl> wrote:
>> Hi Roberto,
>>
>> op 29-08-14 22:44, Roberto Spadim schreef:
>>> i use two or more boot disks, if the first raid1 disk don't boot bios
>>> go to second boot disk, third, etc etc,
>>
>> In my opinion this only works when the boot-disk is completely defect or
>> removed. Not when the data in the MBR on that boot-disk is corrupt.
>>
>> Or did you test this, or do you have other reasons to believe that your
>> bios will handle this correct?
>>
>> With the boot-disk I mean the disk what's in the bios the first disk. So
>> this could also be the second raid1 disk. Or an USB stick.
>>
>>> you must write grub to mbr of each disk
>>
>> Of course.
>>
>> With regards,
>> Paul van der Vlis.
>>
>>> i'm using dell server r410 if i'm not wrong
>>>
>>> 2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>>>> Hello,
>>>>
>>>> I like mdadm and I am using it many years. But in my opinion it has one
>>>> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
>>>> not boot.
>>>>
>>>> A bios could check this. Wait for some kind of signal from Grub or
>>>> Linux, and after a timeout boot from another disk. But I don't know
>>>> about a bios with that feature.
>>>>
>>>> A PCIe card could do something like that, but I don't know about such a
>>>> PCIe card.
>>>>
>>>> Is there such hardware?
>>>> What do you do to avoid this problem?
>>>>
>>>> With regards,
>>>> Paul van der Vlis.
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Paul van der Vlis Linux systeembeheer, Groningen
>>>> http://www.vandervlis.nl/
>>>>
>>>> --
>>>> 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
>>>
>>>
>>>
>>
>>
>>
>>
>>
>> --
>> Paul van der Vlis Linux systeembeheer, Groningen
>> http://www.vandervlis.nl/
>>
>> --
>> 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
> --
> 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
> 





-- 
Paul van der Vlis Linux systeembeheer, Groningen
http://www.vandervlis.nl/

Re: Good hardware for mdadm

[Paul van der Vlis]

op 30-08-14 02:44, Roberto Spadim schreef:
> i'm considering that you are using x86, x86_64 hardware, what's your
> application?

I am a sysadmin and I administrate many machines: webservers and
mailservers located in a datacenter, fileservers located at firms, etc.

> there's no bios test (must check if the newer version of this server
> have uefi and work differente), it just don't boot the disk and go to
> next boot
> but for your question: yes... maybe with 'lucky' it can't boot, a
> "bad" (crashed) grub could stop server startup
> 
> maybe others questions that you could check before solving this boot
> problem is why use a "bad" disk?

I would replace it when I would know it's bad.

> why shutdown a server?

Kernel updates.

> why remote startup a server?

Because it's work to go to such a location.

> why use a 'bad' disk?

I would replace it when I would know it's bad.

> if it's a ha system, 

Not really. Maybe.

> include a human to check monthly, or something
> like it, if the server is ok clean,something like it, 

You can't check that way if a MBR boots.

> since ha
> probably include a real time application or a security or risk
> application, or anything like it

I don't know about such an application. But maybe it would be possible
to read the MBR and check it with a copy. I was looking for a redundant
solution, maybe I must forget that.

> probably you must check others solutions at hardware level, some
> computers have dual bios (yes if one bios is lost you have the other
> to boot the computer), 

That's not the problem. I did not see often problems with a bad bios on
a good-running system.

> maybe an openbios could help

It's difficult to buy hardware for coreboot. And I don't think coreboot
can check the booting process.

>, or a uefi bios, must check what's better

Do you know about UEFI bios-es who can check the process?

With regards,
Paul van der Vlis.

> 
> 
> 2014-08-29 18:47 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>> Hi Roberto,
>>
>> op 29-08-14 22:44, Roberto Spadim schreef:
>>> i use two or more boot disks, if the first raid1 disk don't boot bios
>>> go to second boot disk, third, etc etc,
>>
>> In my opinion this only works when the boot-disk is completely defect or
>> removed. Not when the data in the MBR on that boot-disk is corrupt.
>>
>> Or did you test this, or do you have other reasons to believe that your
>> bios will handle this correct?
>>
>> With the boot-disk I mean the disk what's in the bios the first disk. So
>> this could also be the second raid1 disk. Or an USB stick.
>>
>>> you must write grub to mbr of each disk
>>
>> Of course.
>>
>> With regards,
>> Paul van der Vlis.
>>
>>> i'm using dell server r410 if i'm not wrong
>>>
>>> 2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>>>> Hello,
>>>>
>>>> I like mdadm and I am using it many years. But in my opinion it has one
>>>> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
>>>> not boot.
>>>>
>>>> A bios could check this. Wait for some kind of signal from Grub or
>>>> Linux, and after a timeout boot from another disk. But I don't know
>>>> about a bios with that feature.
>>>>
>>>> A PCIe card could do something like that, but I don't know about such a
>>>> PCIe card.
>>>>
>>>> Is there such hardware?
>>>> What do you do to avoid this problem?
>>>>
>>>> With regards,
>>>> Paul van der Vlis.
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Paul van der Vlis Linux systeembeheer, Groningen
>>>> http://www.vandervlis.nl/
>>>>
>>>> --
>>>> 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
>>>
>>>
>>>
>>
>>
>>
>>
>>
>> --
>> Paul van der Vlis Linux systeembeheer, Groningen
>> http://www.vandervlis.nl/
>>
>> --
>> 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
> 
> 
> 





-- 
Paul van der Vlis Linux systeembeheer, Groningen
http://www.vandervlis.nl/

Re: Good hardware for mdadm

[Roger Heflin]

That means you need an external config tool  (serial, ilo or IMM that
allows remote console access), I rarely go into the data center that
all of the hw I run is in, and because of the remote console access
can fix all but truly broken hw without having hands in the DC).
For most low end enterprise hw this option adds $200-300 to the price.
   If using the simplest (serial) the given mb bios must support
serial and all add-on bioses must also properly support serial.   I
have messed with machines that when using serial you cannot see/access
some of the add on bioses which limits what you can do.   ILO/IMM
(full graphics console remote type setups) don't have this deficiency
and can in some cases actually allow one to present a remote boot cd
to a device and rescue boot it.

The MBR did not get broken, it was broken like this as originally
designed 30 years ago, it was always setup to be only able to deal
with totally dead drivers (ie drive not showing up at all).

On Sat, Aug 30, 2014 at 3:35 AM, Paul van der Vlis <paul@vandervlis.nl> wrote:
> op 30-08-14 00:38, Roger Heflin schreef:
>> No bios I have seen automatically handles it unless the disks are
>> total dead,
>
> That's my experience too.
>
>> some of the hardware raid controlers may properly deal
>> with a disk returning an error and go to the other mirror, the low end
>> quality raid controllers start around $200.
>
> Hardware raid controllers have other disadvantages, like using a
> proprietary format for storing information on disk. And in many cases
> you have to use tools from the manufacturer. That are reasons for me to
> like mdadm.
>
> Maybe it could be done with a hardware raid controller in JBOD mode.
>
> ( And hardware controllers are not perfect too, I've seen simular
> problems with an HP smart array. First when I've removed the first
> harddisk the system did boot from the second. )
>
>> Outside of the raid controllers I don't believe any of the bioses
>> themselves will do it, so you are unlikely to find any hardware that
>> does that.
>
> It could be done with a PCIe card too. Some kind of fake-raid adapter
> what uses mdadm.
>
>> Likely he is talking about putting the boot block on multiple disks
>> and and use the bbr menu or the bios boot disk selection to pick
>> another disk.
>
> Not sure what a "bbr menu" is. But realize that the servers I use are
> remote.
>
> With regards,
> Paul van der Vlis.
>
>> On Fri, Aug 29, 2014 at 4:47 PM, Paul van der Vlis <paul@vandervlis.nl> wrote:
>>> Hi Roberto,
>>>
>>> op 29-08-14 22:44, Roberto Spadim schreef:
>>>> i use two or more boot disks, if the first raid1 disk don't boot bios
>>>> go to second boot disk, third, etc etc,
>>>
>>> In my opinion this only works when the boot-disk is completely defect or
>>> removed. Not when the data in the MBR on that boot-disk is corrupt.
>>>
>>> Or did you test this, or do you have other reasons to believe that your
>>> bios will handle this correct?
>>>
>>> With the boot-disk I mean the disk what's in the bios the first disk. So
>>> this could also be the second raid1 disk. Or an USB stick.
>>>
>>>> you must write grub to mbr of each disk
>>>
>>> Of course.
>>>
>>> With regards,
>>> Paul van der Vlis.
>>>
>>>> i'm using dell server r410 if i'm not wrong
>>>>
>>>> 2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>>>>> Hello,
>>>>>
>>>>> I like mdadm and I am using it many years. But in my opinion it has one
>>>>> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
>>>>> not boot.
>>>>>
>>>>> A bios could check this. Wait for some kind of signal from Grub or
>>>>> Linux, and after a timeout boot from another disk. But I don't know
>>>>> about a bios with that feature.
>>>>>
>>>>> A PCIe card could do something like that, but I don't know about such a
>>>>> PCIe card.
>>>>>
>>>>> Is there such hardware?
>>>>> What do you do to avoid this problem?
>>>>>
>>>>> With regards,
>>>>> Paul van der Vlis.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Paul van der Vlis Linux systeembeheer, Groningen
>>>>> http://www.vandervlis.nl/
>>>>>
>>>>> --
>>>>> 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
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>>
>>>
>>> --
>>> Paul van der Vlis Linux systeembeheer, Groningen
>>> http://www.vandervlis.nl/
>>>
>>> --
>>> 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
>> --
>> 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
>>
>
>
>
>
>
> --
> Paul van der Vlis Linux systeembeheer, Groningen
> http://www.vandervlis.nl/
>
> --
> 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: Good hardware for mdadm

[Paul van der Vlis]

op 30-08-14 11:53, Roger Heflin schreef:
> That means you need an external config tool  (serial, ilo or IMM that
> allows remote console access), I rarely go into the data center that
> all of the hw I run is in, and because of the remote console access
> can fix all but truly broken hw without having hands in the DC).

That's a way to repear something what's broken. Not a more redundant
solution I was looking for. But you are right, it helps a lot. I use
IPMI for remote console. And on some machines a serial console.

What I don't like about IPMI is that the part on the server is -so far I
know- not really open source and I cannot update it for security, so I
can only use it from a secure subnet initiated by another machine.

A serial connection can be done OSS, but you need another machine in the
direct neighborhood, and it's not really nice.

So in both cases you need 2 machines with at least to IP's. Many firms
here have DSL with only one IP.

> For most low end enterprise hw this option adds $200-300 to the price.

It's standard on the machines I buy (Supermicro).

>    If using the simplest (serial) the given mb bios must support
> serial and all add-on bioses must also properly support serial.   I
> have messed with machines that when using serial you cannot see/access
> some of the add on bioses which limits what you can do.   ILO/IMM
> (full graphics console remote type setups) don't have this deficiency
> and can in some cases actually allow one to present a remote boot cd
> to a device and rescue boot it.

Not sure this is possible with OSS applications.

> The MBR did not get broken, it was broken like this as originally
> designed 30 years ago, it was always setup to be only able to deal
> with totally dead drivers (ie drive not showing up at all).

I think you mean the bios. Some bios-es can do fakeraid, and I expect
they can boot from another disk when the first disk is corrupt. But you
cannot use mdadm then, what I would like.

With regards,
Paul van der Vlis.

> On Sat, Aug 30, 2014 at 3:35 AM, Paul van der Vlis <paul@vandervlis.nl> wrote:
>> op 30-08-14 00:38, Roger Heflin schreef:
>>> No bios I have seen automatically handles it unless the disks are
>>> total dead,
>>
>> That's my experience too.
>>
>>> some of the hardware raid controlers may properly deal
>>> with a disk returning an error and go to the other mirror, the low end
>>> quality raid controllers start around $200.
>>
>> Hardware raid controllers have other disadvantages, like using a
>> proprietary format for storing information on disk. And in many cases
>> you have to use tools from the manufacturer. That are reasons for me to
>> like mdadm.
>>
>> Maybe it could be done with a hardware raid controller in JBOD mode.
>>
>> ( And hardware controllers are not perfect too, I've seen simular
>> problems with an HP smart array. First when I've removed the first
>> harddisk the system did boot from the second. )
>>
>>> Outside of the raid controllers I don't believe any of the bioses
>>> themselves will do it, so you are unlikely to find any hardware that
>>> does that.
>>
>> It could be done with a PCIe card too. Some kind of fake-raid adapter
>> what uses mdadm.
>>
>>> Likely he is talking about putting the boot block on multiple disks
>>> and and use the bbr menu or the bios boot disk selection to pick
>>> another disk.
>>
>> Not sure what a "bbr menu" is. But realize that the servers I use are
>> remote.
>>
>> With regards,
>> Paul van der Vlis.
>>
>>> On Fri, Aug 29, 2014 at 4:47 PM, Paul van der Vlis <paul@vandervlis.nl> wrote:
>>>> Hi Roberto,
>>>>
>>>> op 29-08-14 22:44, Roberto Spadim schreef:
>>>>> i use two or more boot disks, if the first raid1 disk don't boot bios
>>>>> go to second boot disk, third, etc etc,
>>>>
>>>> In my opinion this only works when the boot-disk is completely defect or
>>>> removed. Not when the data in the MBR on that boot-disk is corrupt.
>>>>
>>>> Or did you test this, or do you have other reasons to believe that your
>>>> bios will handle this correct?
>>>>
>>>> With the boot-disk I mean the disk what's in the bios the first disk. So
>>>> this could also be the second raid1 disk. Or an USB stick.
>>>>
>>>>> you must write grub to mbr of each disk
>>>>
>>>> Of course.
>>>>
>>>> With regards,
>>>> Paul van der Vlis.
>>>>
>>>>> i'm using dell server r410 if i'm not wrong
>>>>>
>>>>> 2014-08-29 17:31 GMT-03:00 Paul van der Vlis <paul@vandervlis.nl>:
>>>>>> Hello,
>>>>>>
>>>>>> I like mdadm and I am using it many years. But in my opinion it has one
>>>>>> disadvantage: when the MBR of the boot-disk is corrupt, the machine will
>>>>>> not boot.
>>>>>>
>>>>>> A bios could check this. Wait for some kind of signal from Grub or
>>>>>> Linux, and after a timeout boot from another disk. But I don't know
>>>>>> about a bios with that feature.
>>>>>>
>>>>>> A PCIe card could do something like that, but I don't know about such a
>>>>>> PCIe card.
>>>>>>
>>>>>> Is there such hardware?
>>>>>> What do you do to avoid this problem?
>>>>>>
>>>>>> With regards,
>>>>>> Paul van der Vlis.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Paul van der Vlis Linux systeembeheer, Groningen
>>>>>> http://www.vandervlis.nl/
>>>>>>
>>>>>> --
>>>>>> 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
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Paul van der Vlis Linux systeembeheer, Groningen
>>>> http://www.vandervlis.nl/
>>>>
>>>> --
>>>> 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
>>> --
>>> 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
>>>
>>
>>
>>
>>
>>
>> --
>> Paul van der Vlis Linux systeembeheer, Groningen
>> http://www.vandervlis.nl/
>>
>> --
>> 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
> --
> 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
> 





-- 
Paul van der Vlis Linux systeembeheer, Groningen
http://www.vandervlis.nl/

Re: Good hardware for mdadm

[Brad Campbell]

On 30/08/14 18:37, Paul van der Vlis wrote:
> op 30-08-14 11:53, Roger Heflin schreef:
>> That means you need an external config tool  (serial, ilo or IMM that
>> allows remote console access), I rarely go into the data center that
>> all of the hw I run is in, and because of the remote console access
>> can fix all but truly broken hw without having hands in the DC).
>
> That's a way to repear something what's broken. Not a more redundant
> solution I was looking for. But you are right, it helps a lot. I use
> IPMI for remote console. And on some machines a serial console.
>

You do realise that the IPMI is also not redundant and uses exactly the 
same memory type that the DOM offers (which you pointed out was not 
redundant).

Disks die. Anything else is a hands on fix. If you want more reliable 
you go dual-redundant (possibly with a hot standby).

Regards,
Brad
-- 
Dolphins are so intelligent that within a few weeks they can
train Americans to stand at the edge of the pool and throw them
fish.