[linux-lvm] bcache as a new i/o layer?

[linux-lvm] bcache as a new i/o layer?

[Gionatan Danti]

Hi list,
in BZ 1643651 I read:

"In 7.6 (and 8.0), lvm began using a new i/o layer (bcache)
to read and write data blocks."

Last time I checked, bcache was a completely different caching layer, 
unrelated from LVM. The above quote, instead, implies that bcache is now 
actively used by LVM.

Am I missing something? Can you explain how they interact/behave?
Thanks.

-- 
Danti Gionatan
Supporto Tecnico
Assyoma S.r.l. - www.assyoma.it
email: g.danti@assyoma.it - info@assyoma.it
GPG public key ID: FF5F32A8

Re: [linux-lvm] bcache as a new i/o layer?

[Zdenek Kabelac]

Dne 30. 11. 18 v 10:35 Gionatan Danti napsal(a):
> Hi list,
> in BZ 1643651 I read:
> 
> "In 7.6 (and 8.0), lvm began using a new i/o layer (bcache)
> to read and write data blocks."
> 
> Last time I checked, bcache was a completely different caching layer, 
> unrelated from LVM. The above quote, instead, implies that bcache is now 
> actively used by LVM.
> 
> Am I missing something? Can you explain how they interact/behave?



Hi


The name of i/o layer bcache is only internal to lvm2 code for caching
reads form disks during disk processing - the name comes from usage
of  bTree and caching - thus the name bcache.

It's not a dm target - so nothing you could use for LVs.
And has nothing in common with exiting  bcache kernel block device.

The only purpose is to slightly accelerate device scanning and probing - and 
helps mostly in cases your system has tons a devices that needs to be scanned,
since this bcache is also using async disk reads for its filling....


Regards

Zdenek

Re: [linux-lvm] bcache as a new i/o layer?

[Gionatan Danti]

On 30/11/2018 10:52, Zdenek Kabelac wrote:
> Hi
> 
> 
> The name of i/o layer bcache is only internal to lvm2 code for caching
> reads form disks during disk processing - the name comes from usage
> of� bTree and caching - thus the name bcache.
> 
> It's not a dm target - so nothing you could use for LVs.
> And has nothing in common with exiting� bcache kernel block device.
> 
> The only purpose is to slightly accelerate device scanning and probing - 
> and helps mostly in cases your system has tons a devices that needs to 
> be scanned,
> since this bcache is also using async disk reads for its filling....
> 

Hi Zdenek, thanks for the information.
I can imagine bcache being instrumental to the deprecation of lvmetad, 
right?

Thanks.

-- 
Danti Gionatan
Supporto Tecnico
Assyoma S.r.l. - www.assyoma.it
email: g.danti@assyoma.it - info@assyoma.it
GPG public key ID: FF5F32A8

Re: [linux-lvm] bcache as a new i/o layer?

[Zdenek Kabelac]

Dne 30. 11. 18 v 11:03 Gionatan Danti napsal(a):
> On 30/11/2018 10:52, Zdenek Kabelac wrote:
>> Hi
>>
>>
>> The name of i/o layer bcache is only internal to lvm2 code for caching
>> reads form disks during disk processing - the name comes from usage
>> of� bTree and caching - thus the name bcache.
>>
>> It's not a dm target - so nothing you could use for LVs.
>> And has nothing in common with exiting� bcache kernel block device.
>>
>> The only purpose is to slightly accelerate device scanning and probing - and 
>> helps mostly in cases your system has tons a devices that needs to be scanned,
>> since this bcache is also using async disk reads for its filling....
>>
> 
> Hi Zdenek, thanks for the information.
> I can imagine bcache being instrumental to the deprecation of lvmetad, right?

Hi

Well lvmetad was dropped from release  2.03  (upstream HEAD)

But for different reason - it's had some core design issues which were not 
fixable with exist state of code ATM, so the whole lvmetad was dropped and for 
now replaced with storage of some data in /run/lvm/pvs_online  so at least 
auto-activation still works in some way.

IMHO internal bcache is a bit to heavy weapon for its purpose, but it will 
evolve into a self contained io-manager.


Zdenek