software raid and bcache

software raid and bcache

[Vasiliy Tolstov]

Hello. I'm newbie with bcache and stuck at drive configuration.
Now i have 1 sata 3Tb disk and one 240Gb ssd eMLC disk. I want to
create from 1 sata disk raid1 with missing 1 drive (i want to able to
sync pre failed sata disk with new by remove ssd disk and add sata)

What steps i need to do?



--
Vasiliy Tolstov,
e-mail: v.tolstov-+9FY0jupvH6HXe+LvDLADg@public.gmane.org
jabber: vase-+9FY0jupvH6HXe+LvDLADg@public.gmane.org

Re: software raid and bcache

[Vasiliy Tolstov]

Hmm. Sorry i'm not fully understand.

I think :

create /dev/md0 with /dev/sda (sata) level=1 raid-devices=1
format /dev/md0 with bcache
format /dev/sdb (ssd) with brache

create bcache0 with /dev/md0 and /dev/sdb

why this is not possible?


2013/7/3 matthew patton <pattonme-/E1597aS9LQAvxtiuMwx3w@public.gmane.org>:
> short answer you can't. you can unbind the SSD from bcache and otherwise remove the bcache config but you can't turn a live disk into a MD raid member without destroying all data on it. Introduce the 2nd big hard disk and create a raid1 on it with an absent mirror pair. DD or RSYNC the data from your first disk to the new MD device. reconfigure first disk to be MD member and let MD re-sync the (same) data back once more.
>
> add SSD and MD device to bcache config and it's now accelerated.



--
Vasiliy Tolstov,
e-mail: v.tolstov-+9FY0jupvH6HXe+LvDLADg@public.gmane.org
jabber: vase-+9FY0jupvH6HXe+LvDLADg@public.gmane.org

Re: software raid and bcache

[Gabriel de Perthuis]

On Wed, 03 Jul 2013 19:43:55 +0400, Vasiliy Tolstov wrote:
> Hello. I'm newbie with bcache and stuck at drive configuration.
> Now i have 1 sata 3Tb disk and one 240Gb ssd eMLC disk. I want to
> create from 1 sata disk raid1 with missing 1 drive (i want to able to
> sync pre failed sata disk with new by remove ssd disk and add sata)
> 
> What steps i need to do?

Hello,

You could use `blocks to-lvm`[1] to convert your large disk to lvm,
and then `blocks to-bcache` to convert the lvm volume to bcache.
The lvm conversion will allow you to migrate the data off the
prefailed disk later, either by moving it (pvmove) or by converting
to raid1 (lvconvert).

The blocks README should guide you through converting.  Right now it
doesn't include instructions on attaching the slow backing device to
the SSD cache device, I'll see about adding that, or you can refer
to the ATTACHING section of [2].

[1] https://github.com/g2p/blocks
[2] http://atlas.evilpiepirate.org/git/linux-bcache.git/tree/Documentation/bcache.txt?h=bcache-dev

Re: software raid and bcache

[Vasiliy Tolstov]

2013/7/3 Gabriel de Perthuis <g2p.code-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>:
> Hello,
>
> You could use `blocks to-lvm`[1] to convert your large disk to lvm,
> and then `blocks to-bcache` to convert the lvm volume to bcache.
> The lvm conversion will allow you to migrate the data off the
> prefailed disk later, either by moving it (pvmove) or by converting
> to raid1 (lvconvert).


Why i can't do :

create /dev/md0 with /dev/sda (sata) level=1 raid-devices=1
format /dev/md0 with bcache
format /dev/sdb (ssd) with brache

create bcache0 with /dev/md0 and /dev/sdb

?



--
Vasiliy Tolstov,
e-mail: v.tolstov-+9FY0jupvH6HXe+LvDLADg@public.gmane.org
jabber: vase-+9FY0jupvH6HXe+LvDLADg@public.gmane.org

Re: software raid and bcache

[Gabriel de Perthuis]

On Wed, 03 Jul 2013 20:50:59 +0400, Vasiliy Tolstov wrote:
> 2013/7/3 Gabriel de Perthuis:
>> Hello,
>>
>> You could use `blocks to-lvm`[1] to convert your large disk to lvm,
>> and then `blocks to-bcache` to convert the lvm volume to bcache.
>> The lvm conversion will allow you to migrate the data off the
>> prefailed disk later, either by moving it (pvmove) or by converting
>> to raid1 (lvconvert).
> 
> Why i can't do :
> 
> create /dev/md0 with /dev/sda (sata) level=1 raid-devices=1
> format /dev/md0 with bcache
> format /dev/sdb (ssd) with brache
> 
> create bcache0 with /dev/md0 and /dev/sdb

That's reasonable, if you don't have any data you need to keep.
You can use

    make-bcache -C /deb/sdb -B /dev/md0

to format the devices and attach them at the same time.

Re: software raid and bcache

[Jason Warr]

On 07/03/2013 11:50 AM, Vasiliy Tolstov wrote:
> 2013/7/3 Gabriel de Perthuis <g2p.code-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>:
>> Hello,
>>
>> You could use `blocks to-lvm`[1] to convert your large disk to lvm,
>> and then `blocks to-bcache` to convert the lvm volume to bcache.
>> The lvm conversion will allow you to migrate the data off the
>> prefailed disk later, either by moving it (pvmove) or by converting
>> to raid1 (lvconvert).
>
> Why i can't do :
>
> create /dev/md0 with /dev/sda (sata) level=1 raid-devices=1
> format /dev/md0 with bcache
> format /dev/sdb (ssd) with brache
>
> create bcache0 with /dev/md0 and /dev/sdb
>
> ?
If you want md0 to be a raid1 device in the future you need to create it
as a 2 device array with a missing drive:

mdadm -C /dev/md0 -l1 -n2 /dev/sda missing

You can then add in the other device whenever you can/want.  The key
though would be to make sure the size in blocks of the first device is =
or smaller than the second device will be.  What I would do is create a
fd type partition that is 1% smaller than total raw device size on
/dev/sda just to give your self a small margin of error.

parted /dev/sda mkpart primary 1.0 99% --align optimal
parted /dev/sda set 1 raid on
mdadm -C /dev/md0 -l1 -n2 /dev/sda1 missing

That way down the road you can add in a disk of about the same size to
the existing raid1 completely transparent of bcache with a single command.

mdadm /dev/md1 -a /dev/sdX1

>
>
> --
> Vasiliy Tolstov,
> e-mail: v.tolstov-+9FY0jupvH6HXe+LvDLADg@public.gmane.org
> jabber: vase-+9FY0jupvH6HXe+LvDLADg@public.gmane.org
> --
> To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: software raid and bcache

[Vasiliy Tolstov]

2013/7/3 Jason Warr <jason-/cow75dQlsI@public.gmane.org>:
> If you want md0 to be a raid1 device in the future you need to create it
> as a 2 device array with a missing drive:
>
> mdadm -C /dev/md0 -l1 -n2 /dev/sda missing
>
> You can then add in the other device whenever you can/want.  The key
> though would be to make sure the size in blocks of the first device is =
> or smaller than the second device will be.  What I would do is create a
> fd type partition that is 1% smaller than total raw device size on
> /dev/sda just to give your self a small margin of error.
>
> parted /dev/sda mkpart primary 1.0 99% --align optimal
> parted /dev/sda set 1 raid on
> mdadm -C /dev/md0 -l1 -n2 /dev/sda1 missing
>
> That way down the road you can add in a disk of about the same size to
> the existing raid1 completely transparent of bcache with a single command.
>
> mdadm /dev/md1 -a /dev/sdX1


Thanks. But as i understand if the second drive bigger then first i
can sync raid1 and after thet grow /dev/md0 to new size...

--
Vasiliy Tolstov,
e-mail: v.tolstov-+9FY0jupvH6HXe+LvDLADg@public.gmane.org
jabber: vase-+9FY0jupvH6HXe+LvDLADg@public.gmane.org

Re: software raid and bcache

[Jason Warr]

On 07/03/2013 12:15 PM, Vasiliy Tolstov wrote:
> 2013/7/3 Jason Warr <jason-/cow75dQlsI@public.gmane.org>:
>> If you want md0 to be a raid1 device in the future you need to create it
>> as a 2 device array with a missing drive:
>>
>> mdadm -C /dev/md0 -l1 -n2 /dev/sda missing
>>
>> You can then add in the other device whenever you can/want.  The key
>> though would be to make sure the size in blocks of the first device is =
>> or smaller than the second device will be.  What I would do is create a
>> fd type partition that is 1% smaller than total raw device size on
>> /dev/sda just to give your self a small margin of error.
>>
>> parted /dev/sda mkpart primary 1.0 99% --align optimal
>> parted /dev/sda set 1 raid on
>> mdadm -C /dev/md0 -l1 -n2 /dev/sda1 missing
>>
>> That way down the road you can add in a disk of about the same size to
>> the existing raid1 completely transparent of bcache with a single command.
>>
>> mdadm /dev/md1 -a /dev/sdX1
>
> Thanks. But as i understand if the second drive bigger then first i
> can sync raid1 and after thet grow /dev/md0 to new size...
In a mirror your raid device can only be as big as the smallest device. 
So if you put in a secondary device that is larger than the first you
still cannot go beyond the size of the first unless you then replace it
after syncing the second.

> --
> Vasiliy Tolstov,
> e-mail: v.tolstov-+9FY0jupvH6HXe+LvDLADg@public.gmane.org
> jabber: vase-+9FY0jupvH6HXe+LvDLADg@public.gmane.org

Re: software raid and bcache

[Vasiliy Tolstov]

2013/7/3 Jason Warr <jason-/cow75dQlsI@public.gmane.org>:
> In a mirror your raid device can only be as big as the smallest device.
> So if you put in a secondary device that is larger than the first you
> still cannot go beyond the size of the first unless you then replace it
> after syncing the second.


Yes i'm understand this. But normally i want to work with one device
in raid1 =). Thanks all for help!

--
Vasiliy Tolstov,
e-mail: v.tolstov-+9FY0jupvH6HXe+LvDLADg@public.gmane.org
jabber: vase-+9FY0jupvH6HXe+LvDLADg@public.gmane.org

Re: software raid and bcache

[Gabriel de Perthuis]

On Wed, 03 Jul 2013 21:57:54 +0400, Vasiliy Tolstov wrote:
> 2013/7/3 Jason Warr <jason-/cow75dQlsI@public.gmane.org>:
>> In a mirror your raid device can only be as big as the smallest device.
>> So if you put in a secondary device that is larger than the first you
>> still cannot go beyond the size of the first unless you then replace it
>> after syncing the second.
> 
> Yes i'm understand this. But normally i want to work with one device
> in raid1 =). Thanks all for help!

LVM would give you more flexibility for the use case of
moving data around, you could use any number of disks
regardless of their sizes.  Though I suppose you're
sticking with what you know.

Re: software raid and bcache

[Vasiliy Tolstov]

2013/7/3 Gabriel de Perthuis <g2p.code-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>:
> LVM would give you more flexibility for the use case of
> moving data around, you could use any number of disks
> regardless of their sizes.  Though I suppose you're
> sticking with what you know.


In my case i have server with only two slots for hdd =) But thanks for
help, i think that lvm in my case not so bad.

--
Vasiliy Tolstov,
e-mail: v.tolstov-+9FY0jupvH6HXe+LvDLADg@public.gmane.org
jabber: vase-+9FY0jupvH6HXe+LvDLADg@public.gmane.org