Re: component devices too large

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Saturday February 18, hardyjm@xxxxxxxxxxx wrote:
> > The amount of each component drive this is actually in use is stored -
> > in the default metadata - in a 32bit number as kilobytes.
> > This sets a limit of 4TB.
> > 
> > The version-1 metadata format has a 64bit field.
> > 
> > If you use mdadm 2.3, it will automatically select version-1 metadata
> > if you choose a size larger than 2TB (I think).
> > You might also need a very recent (16-rc) kernel for v.large raid1 to
> > work.  There were bugs related to this usage pattern that were fixed
> > only recently.
> > 
> > To avoid resyncing >4TB, use of a write-intent-bitmap is strongly
> > recommended. 
> 
> Thanks for the quick reply.  I see now that mdadm supported the new
> version-1 metadata format as of 1.8.1, though that was labelled a devel
> release in its Announcement (also where the new metadata is mentioned).
> I do not see any mention of this in the Changelog however until 2.0,
> which allows you to specify the metadata type with the '-e' option.

Indeed, 1.8.1 was a devel release - an early prerelease of what was to
become 2.0.

> 
> Also, the man pages for the 2.x releases state that the default metadata
> "limits arrays to 28 componenet devices and limits component devices of
> levels 1 and greater to 2 terabytes."
> 
> So, I admit I am still confused.  Either:
> 
> 1) As you indicated, and I have seen myself, I can go up to 4TB with the
> default metadata on 1.11.0 which means the man pages for 2.x, the first
> to actually mention metadata, are incorrect.

The man page is probably incorrect... but don't bet on it.
The version 0.90 metadata has an unsigned 32 field for kilobytes,
allowing up to 4TB.  However the ioctl call for creating an array has
a *signed* 32 bit size, allowing on 2TB.

mdadm-1.* uses the ioctl to create arrays.  mdadm-2 doesn't.  It
writes the metadata directly and then assembles the array.
So while I haven't tested the various versions with different sizes
very much, it could be that mdadm-1.* is limited to 2TB, and mdadm-2
allows upto 4TB with v0.90 metadata, and virtually unlimited with v1.
However >2TB is only known to work with mdadm-2.3. and later.

> 
> 2) mdadm-1.11.0 (which I am using), and possibly earlier, supported
> version-1 metadata, but it was not a selectable option within mdadm, but
> rather mdadm auto-selected metadata based on what options and sizes one
> threw at it.

No.  mdadm-1 (with the exception of 1.8.1) do not support version-1
metadata.

> 
> 3) Neither of the above, and I was just bloody lucky to be able to
> create an array that size, and anything beyond 2TB on mdadm 1.x is not a
> very good idea for production use.

Definitely >= 2TB not recommended with anything before 2.3.

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

[Index of Archives]     [Linux RAID Wiki]     [ATA RAID]     [Linux SCSI Target Infrastructure]     [Linux Block]     [Linux IDE]     [Linux SCSI]     [Linux Hams]     [Device Mapper]     [Device Mapper Cryptographics]     [Kernel]     [Linux Admin]     [Linux Net]     [GFS]     [RPM]     [git]     [Yosemite Forum]


  Powered by Linux