[linux-lvm] calculating size of metadata.

[linux-lvm] calculating size of metadata.

[Jan-Hendrik Palic]

Hi there,

we planning to create a pv and vg on a 19,95TB partition running on 
raid6. One issue I have is the size of the metadata. I am looking for a 
way to calculate the this size, but unfortunatly, I cannot find a some.
Which size should I use?

Thanks in advance.

Jan

Re: [linux-lvm] calculating size of metadata.

[Lars Ellenberg]

On Tue, Apr 24, 2012 at 02:23:31PM +0200, Jan-Hendrik Palic wrote:
> Hi there,
> 
> we planning to create a pv and vg on a 19,95TB partition running on
> raid6. One issue I have is the size of the metadata. I am looking
> for a way to calculate the this size, but unfortunatly, I cannot
> find a some.
> Which size should I use?

Depends on how many LVs you intend to have,
and how much LV fragmentation you expect.

I'd estimate 1 to 2 kByte LVM meta data per expected LV,
that should be on the safe side.

-- 
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

Re: [linux-lvm] calculating size of metadata.

[Ray Morris]

It's good think about this because running out of metadata space is no
fun. We use 181KB for about 300 LVs, where each snapshot is counted as
two LVs because of the way it's stored.  That's based on the size of the
backup. vgs shows 8 MB of metadata used, though:

# vgs --units k -o vg_mda_count,vg_mda_free,vg_mda_size clones
vgs  #VMda VMdaFree  VMdaSize 
vgs      3  7973.50k 16444.00k

In any event, on a 19,950,000 MB partition, even 64 MB of metadata
would represent 0.0006% of the storage. To think of it another way, 64
MB costs roughly 0.03 cents, so allocate plenty.
-- 
Ray Morris
support@bettercgi.com

Strongbox - The next generation in site security:
http://www.bettercgi.com/strongbox/

Throttlebox - Intelligent Bandwidth Control
http://www.bettercgi.com/throttlebox/

Strongbox / Throttlebox affiliate program:
http://www.bettercgi.com/affiliates/user/register.php




On Tue, 24 Apr 2012 14:23:31 +0200
Jan-Hendrik Palic <billgotchy@ki.tng.de> wrote:

> Hi there,
> 
> we planning to create a pv and vg on a 19,95TB partition running on 
> raid6. One issue I have is the size of the metadata. I am looking for
> a way to calculate the this size, but unfortunatly, I cannot find a
> some. Which size should I use?
> 
> Thanks in advance.
> 
> Jan
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
> 

Re: [linux-lvm] calculating size of metadata.

[Jan-Hendrik Palic]

Hi,

On 24.04.2012 16:52, Ray Morris wrote:
> It's good think about this because running out of metadata space is no
> fun.

We had this problem already on an other system. We had to migrate to 
another storage with pain. The new storage has a larger space for the 
metadata for sure. .-)

But I want to understand and not only set this size to an high value. 
Therefore this question here.

> We use 181KB for about 300 LVs, where each snapshot is counted as
> two LVs because of the way it's stored.  That's based on the size of the
> backup. vgs shows 8 MB of metadata used, though:
>
> # vgs --units k -o vg_mda_count,vg_mda_free,vg_mda_size clones
> vgs  #VMda VMdaFree  VMdaSize
> vgs      3  7973.50k 16444.00k

Thank you and Lars for your replies. If I understand you right, I have 
to look to the count of LVs I will have in the future and not only on 
the size of my volumegroup.

Ok.

> In any event, on a 19,950,000 MB partition, even 64 MB of metadata
> would represent 0.0006% of the storage. To think of it another way, 64
> MB costs roughly 0.03 cents, so allocate plenty.

Lars stated in his mail to this list, that one entry for a LV should be 
less than 1kb. So if I calculate 2kb for this entry, I will be safe. I 
checked this here and I can confirm this.

If I have a metadata space of 64MB, I am able to create 32000 different 
LV (without snapshots). Right?

Well, that is enought, even with snapshots,

Anyway ... thank you for my enlightment. .-)

Best regards

Jan