[linux-lvm] swap on thin-provisioning

[linux-lvm] swap on thin-provisioning

[Chris Murphy]

Hi,

Is it expected that swap on a thin volume should work? I'm guessing
the answer is no because mkswap doesn't come with any sort of
allocation request that would set the LE to PE mapping.

What about a swapfile (on ext4 or XFS) onr a thin volume? In this
case, I'd expect fallocate would set the LE to PE mapping, and it
should work. But does it work for both paging and hibernation files?

And what's the implication if the thin volume were to be snapshot?

Thanks,

-- 
Chris Murphy

_______________________________________________
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] swap on thin-provisioning

[Gionatan Danti]

Il 2020-11-29 01:18 Chris Murphy ha scritto:
> What about a swapfile (on ext4 or XFS) onr a thin volume? In this
> case, I'd expect fallocate would set the LE to PE mapping, and it
> should work. But does it work for both paging and hibernation files?

If things did not change, fallocate does *not* allocate PE space (ie: 
they are not passed down to the underlying thin pool).

Moreover, I expect for a memory-starved machine to page out pool 
metadata (right?) so I would not use a thin volume for swap.

Regards.

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

_______________________________________________
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] swap on thin-provisioning

[Sreyan Chakravarty]

[-- Attachment #1.1: Type: text/plain, Size: 662 bytes --]

On Sun, 29 Nov 2020, 6:23 pm Gionatan Danti, <g.danti@assyoma.it> wrote:

> Il 2020-11-29 01:18 Chris Murphy ha scritto:
> > What about a swapfile (on ext4 or XFS) onr a thin volume? In this
> > case, I'd expect fallocate would set the LE to PE mapping, and it
> > should work. But does it work for both paging and hibernation files?
>
> If things did not change, fallocate does *not* allocate PE space (ie:
> they are not passed down to the underlying thin pool).
>
> Moreover, I expect for a memory-starved machine to page out pool
> metadata (right?) so I would not use a thin volume for swap.
>
> Regards.
>


What about a swap file allocated with dd??

>
>

[-- Attachment #1.2: Type: text/html, Size: 1231 bytes --]

[-- Attachment #2: Type: text/plain, Size: 197 bytes --]

_______________________________________________
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] swap on thin-provisioning

[Gionatan Danti]

Il 2020-11-29 16:12 Sreyan Chakravarty ha scritto:
> 
> What about a swap file allocated with dd??

If thinpool metadata can be paged out (which I don't know for sure), 
swap on thinvol can potentially deadlock.

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

_______________________________________________
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] swap on thin-provisioning

[Nicholas Geovanis]

[-- Attachment #1.1: Type: text/plain, Size: 1185 bytes --]

IIRC at VMWare vSphere 5 and before, swap could not be located on their
type of thin logical volume. Nor could it swap to a linux LVM volume on top
of that.
It just couldnt perform well. That later changed.

On Sun, Nov 29, 2020, 6:53 AM Gionatan Danti <g.danti@assyoma.it> wrote:

> Il 2020-11-29 01:18 Chris Murphy ha scritto:
> > What about a swapfile (on ext4 or XFS) onr a thin volume? In this
> > case, I'd expect fallocate would set the LE to PE mapping, and it
> > should work. But does it work for both paging and hibernation files?
>
> If things did not change, fallocate does *not* allocate PE space (ie:
> they are not passed down to the underlying thin pool).
>
> Moreover, I expect for a memory-starved machine to page out pool
> metadata (right?) so I would not use a thin volume for swap.
>
> Regards.
>
> --
> Danti Gionatan
> Supporto Tecnico
> Assyoma S.r.l. - www.assyoma.it
> email: g.danti@assyoma.it - info@assyoma.it
> GPG public key ID: FF5F32A8
>
> _______________________________________________
> 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/
>
>

[-- Attachment #1.2: Type: text/html, Size: 2063 bytes --]

[-- Attachment #2: Type: text/plain, Size: 197 bytes --]

_______________________________________________
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/