From: "Lukas Sichert" <l.sichert@proxmox.com>
To: "Christian Ebner" <c.ebner@proxmox.com>, <pve-devel@lists.proxmox.com>
Subject: Re: [PATCH v3 pve-qemu qemu-server 0/6] bypass host page cache for restore on zvol
Date: Thu, 20 Aug 2026 13:36:20 +0200 [thread overview]
Message-ID: <DKTQNX5190Y5.1Z3CFAINZNWK@proxmox.com> (raw)
In-Reply-To: <20260820090528.117853-1-c.ebner@proxmox.com>
On 2026-08-20 11:05, Christian Ebner <c.ebner@proxmox.com> wrote:
> This patch series skips writing to the host page cache for restores
> to zvols, with the intention to avoid potential I/O delay issues for
> other ZFS-backed VMs during page writeback on certain setups, as
> reported in enterprise support and reproduced internally.
>
> This patch series adds an optional no-cache parameter for pbs-restore
> command invocation, setting the BDRV_O_NOCACHE flag for the block
> device being restored to and sets `cache=none` for vma restores.
>
> Since this seems to affect only zvol's, conditionally set the flag only
> during restores to storages with type `zfspool`.
Tested this on a PVE VM whose disk is backed by ZFS, with another VM
acting as PBS. Inside the PVE VM, I created a Linux VM with a 200 GB
disk and wrote 150 GB of data to it. I also installed a Windows Server
2026 VM alongside the inner PVE VM.
During restore, the Dirty and Writeback page cache values from
/proc/meminfo, as reported by the outer PVE VM, were in the 1-4.5 GB
range. ioping reported delays of up to 10 seconds. The Windows VM
stalled; while I was still able to drag windows around, it was not
possible to open any new applications.
After applying the patch, the host page cache values did not increase
when starting a restore and remained basically at zero, with at most
100 KB. The Windows VM was still very slow, but during restore it was
eventually able to open applications. The ioping issue was also
resolved.
One nit:
I already mentioned in [1] that extending the version check regex to
allow suffixes would be nice. The regex in [1] only allows suffixes
starting with +, but Debian Policy [2], Section 5.6.12.2, also
mentions ~. Therefore, something like the following regex might be
better:
my $package_version_re = qr/\(pve-qemu-kvm_(\d+\.\d+(\.\d+)?-\d+)([+~][a-zA-Z0-9~]+)?\)/;
But since this works as advertized, consider:
Tested-by: Lukas Sichert <l.sichert@proxmox.com>
If no bigger changes are made, feel free to use my Tested-by in a future version.
[1] https://lore.proxmox.com/all/DKTOJDG9FQ00.31RUAI1TR35Y2@proxmox.com/T/#m38e4cd2377a01db1d090c946122b90f9e7f8ded1
[2] https://www.debian.org/doc/debian-policy/policy.pdf
next prev parent reply other threads:[~2026-08-20 11:36 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-08-20 9:05 [PATCH v3 pve-qemu qemu-server 0/6] bypass host page cache for restore on zvol Christian Ebner
2026-08-20 9:05 ` [PATCH v3 pve-qemu 1/1] pbs-restore: add optional no-cache flag to bypass host page cache Christian Ebner
2026-08-20 9:05 ` [PATCH v3 qemu-server 2/6] helpers: never cache `unknown` on failed kvm binary version check Christian Ebner
2026-08-20 9:05 ` [PATCH v3 qemu-server 3/6] helpers: optionally get package version for kvm_user_version() Christian Ebner
2026-08-20 9:05 ` [PATCH v3 qemu-server 4/6] helpers: add helper for min kvm package version comparison Christian Ebner
2026-08-20 9:05 ` [PATCH v3 qemu-server 5/6] pbs-restore: set 'no-cache' on block devices backed by zfspool Christian Ebner
2026-08-20 9:05 ` [PATCH v3 qemu-server 6/6] vma restore: skip page cache " Christian Ebner
2026-08-20 9:56 ` [PATCH v3 pve-qemu qemu-server 0/6] bypass host page cache for restore on zvol Lukas Sichert
2026-08-20 12:07 ` Thomas Lamprecht
2026-08-20 12:36 ` Christian Ebner
2026-08-20 11:36 ` Lukas Sichert [this message]
2026-08-20 13:12 ` Christian Ebner
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=DKTQNX5190Y5.1Z3CFAINZNWK@proxmox.com \
--to=l.sichert@proxmox.com \
--cc=c.ebner@proxmox.com \
--cc=pve-devel@lists.proxmox.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.