From: Nikolay Borisov <nborisov@suse.com>
To: "Martin Steigerwald" <martin@lichtvoll.de>,
"Agustín DallʼAlba" <agustin@dallalba.com.ar>,
linux-btrfs@vger.kernel.org
Subject: Re: raid10 corruption while removing failing disk
Date: Mon, 10 Aug 2020 10:51:47 +0300 [thread overview]
Message-ID: <376677b3-0e2f-3c53-c706-4362738e6d3f@suse.com> (raw)
In-Reply-To: <16328609.DpnNoz7ane@merkaba>
On 10.08.20 г. 10:38 ч., Martin Steigerwald wrote:
> Hi Nikolay.
>
> Nikolay Borisov - 10.08.20, 09:22:14 CEST:
>> On 10.08.20 г. 10:03 ч., Agustín DallʼAlba wrote:
> […]
>>> # uname -a
>>> Linux susanita 4.15.0-111-generic #112-Ubuntu SMP Thu Jul 9 20:32:34
>>> UTC 2020 x86_64 x86_64 x86_64 GNU/Linux
>>
>> This is a vendor kernel so you should ideally seek support through the
>> vendor. This kernel is not even an LTS so it's not entirely clear
>> which patches have/have not been backported. With btrfs it's
>> advisable too use the latest stable kernel as each release brings bug
>> fixes or at the very least (because always using the latest is not
>> feasible) at least stick to a supported long-term stable kernel -
>> i.e 4.14, 4.19 or 5.4 (preferably 5.4)
>
> The interesting thing with this recommendation is that it to some part
> equals:
>
> Do not use distro / vendor kernels.
On the contrary - it means to use kernels which have support for btrfs.
Namely - Suse distributes kernels with btrfs + have developers who are
familiar with the state of btrfs on their kernel. So if someone hits a
problem on a Suse kernel - they should report this to Suse and not the
upstream mailing list. Suse's (or any other vendor for that matter)
needn't look anything like the upstream kernel. Same thing with Fedora
or Ubuntu. Since time is limited I (as an upstream developer) would
prefer to spend my time where it would have the most impact - upstream
and not spend possibly hours looking at some custom kernel.
>
> Consequently vendors shall *just* exclude BTRFS from being shipped?
No, vendors should verify that what they are offering is stable and when
they diverge from upstream should provide support. Fedora for example
spent the time to test btrfs for their use cases + they have the support
from some of the developers to fix issues they find because they
themselves don't have the development expertise (yet) to deal with
btrfs. Furthermore I *think* fedora is sticking to using unadorned
upstream kernels (don't quote me on this, I have never used fedora).
>
> At one point in BTRFS development, I'd expect BTRFS to be stable enough
> to be shipped in distro kernels, like XFS, Ext4 and other filesystems.
>
> For me it appears to be. I used it on 5.8 stable kernel on this laptop,
> but on another laptop and on two server virtual machines I used the
> standard Devuan 3 aka Debian 10 kernel (4.19). Without issues so far.
>
> I am just raising this, cause I would believe that at one point it time
> it is important to say: It is *okay* to use vendor kernels. Still
> probably ask your vendor for support first, but it is basically *okay* to
> use them. On the other hand, regarding Debian, I'd expect I could reach
> way more experts regarding BTRFS issues on this mailing list than I
> would find in the Debian kernel team. So I'd probably still ask here
> first.
Sure, but what guarantees that Debian (or any other distro) have
backported whatever patches are necessary too reach certain stability?
Upstream is a moving target and we (at Suse at least) always target an
upstream-first policy so most effort is spent there.
>
> What would need to happen for it to be okay to use vendor kernels? Is
> there a minimum LTS version where you would say it would be okay?
>
> I am challenging this standard recommendation here, cause I am not sure
> whether for recent distribution releases it would still be accurate or
> helpful. At some point BTRFS got to be as stable as XFS or Ext4 I would
> think. Again, for me it is.
>
> I have no idea why Ubuntu opted to use a non LTS kernel – especially as
> 4.15 is pretty old and so does not sound to come from a supported Ubuntu
> release unless it is some Ubuntu LTS release, but then I'd expect a LTS
> kernel to be used –, but "-111" indicates they added a lot of patches by
> now. So maybe they provide some kind of LTS support themselves.
All those are valid assumptions - however without direct experience with
the Ubuntu kernel it's not entirely clear how accurate they are. Hence
my recommendation to address Ubuntu kernel people because they should
know best.
>
> Best,
>
next prev parent reply other threads:[~2020-08-10 7:51 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-10 7:03 raid10 corruption while removing failing disk Agustín DallʼAlba
2020-08-10 7:22 ` Nikolay Borisov
2020-08-10 7:38 ` Martin Steigerwald
2020-08-10 7:51 ` Nikolay Borisov [this message]
2020-08-10 8:57 ` Martin Steigerwald
2020-08-11 1:30 ` Chris Murphy
2020-08-10 7:59 ` Agustín DallʼAlba
2020-08-10 8:21 ` Nikolay Borisov
2020-08-10 22:24 ` Zygo Blaxell
2020-08-11 1:18 ` Agustín DallʼAlba
2020-08-11 1:48 ` Chris Murphy
2020-08-11 2:34 ` Chris Murphy
2020-08-11 5:06 ` Agustín DallʼAlba
2020-08-11 19:17 ` Chris Murphy
2020-08-11 20:40 ` Agustín DallʼAlba
2020-08-12 3:03 ` Chris Murphy
2020-08-31 20:05 ` Agustín DallʼAlba
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=376677b3-0e2f-3c53-c706-4362738e6d3f@suse.com \
--to=nborisov@suse.com \
--cc=agustin@dallalba.com.ar \
--cc=linux-btrfs@vger.kernel.org \
--cc=martin@lichtvoll.de \
/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.