gfs2.lists.linux.dev archive mirror
 help / color / mirror / Atom feed
From: Alexander Aring <aahringo@redhat.com>
To: "Valentin Vidić" <vvidic@valentin-vidic.from.hr>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>,
	ocfs2-devel@oss.oracle.com,  David Teigland <teigland@redhat.com>,
	gfs2@lists.linux.dev, ocfs2-devel@lists.linux.dev
Subject: Re: ocfs2 mount error
Date: Tue, 12 Mar 2024 15:55:46 -0400	[thread overview]
Message-ID: <CAK-6q+j8Ag5yj6pn0P4wC2kRSNMVaHGmVY+G6+H5aYD1-ygzSw@mail.gmail.com> (raw)
In-Reply-To: <ZfCg6o3Kmkq78SIG@valentin-vidic.from.hr>

Hi,

On Tue, Mar 12, 2024 at 2:37 PM Valentin Vidić
<vvidic@valentin-vidic.from.hr> wrote:
>
> On Tue, Mar 12, 2024 at 01:55:54PM -0400, Alexander Aring wrote:
> > I was able to reproduce the issue and sent a fix out.
>
> Yay, thanks a lot for testing, I can confirm the patches fix the problem
> for me!
>

Then you can reply to the mail with:

Tested-by: Valentin Vidić <vvidic@valentin-vidic.from.hr>

that would help to bring the fix upstream.

> > If you want to extend your debian testing using different approach
> > (not libdlm) this can be done by:
> >
> > mount -t ocfs2_dlmfs none /dlm
>
> I tried that but fsck than fails to make a lock, so I think ocfs2_dlmfs
> is only relevant when used with o2cb cluster type instead of corosync.
>

Yes, I think you are right. It is a replacement for
corosync/dlm_controld handling. It seems I had both setup working at
the same time, but it still uses kernel DLM locking.

> > I am curious about your testing at debian cluster, do you also test gfs2?
>
> Yes, a similar test for gfs2 would be this:
>
> https://salsa.debian.org/ha-team/gfs2-utils/-/blob/master/debian/tests/corosync?ref_type=heads
>
> It is not really a cluster test since it only runs on one KVM
> machine. But as a package test it works well because even for
> 1-node cluster all the code (userspace and kernel) still needs
> to work correctly.
>

I see, I currently have plans/ideas to add namespace support into
kernel DLM that might help you by extending your testing. The general
idea would be that you can do something like:

...
ip netns exec node1 mount /dev/fake_shared_block /mnt/node1
ip netns exec node2 mount /dev/fake_shared_block /mnt/node2
ip netns exec node3 mount /dev/fake_shared_block /mnt/node3
...

I hope that somehow explains the idea behind it. Setup your cluster
networking with namespaces, e.g. selftests of Linux kernel is doing it
[0] and the DLM lockspaces are separated by net namespaces*. It
"should" run like a cluster with several machines, it is only to do
more testing and we don't need to synchronize tests over the network.
However it will _not_ replace real cluster testing but it is a
starting point to make testing "simpler". Somewhere is even a big fat
"software" switch you can run wireshark on it to see all cluster
traffic going on.

* In theory I am only abusing net namespaces for separate DLM
lockspaces, but I don't want to introduce another kind of namespace
now. However net namespaces fit here because it is necessary for a per
node networking setup anyway.

- Alex

[0] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/testing/selftests/net/icmp.sh


  reply	other threads:[~2024-03-12 19:56 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-03-10 21:46 ocfs2 mount error Valentin Vidić
2024-03-11  8:37 ` Joseph Qi
2024-03-11  9:02   ` Valentin Vidić
2024-03-11  9:27     ` Joseph Qi
2024-03-11 12:28       ` Heming Zhao
2024-03-11 15:20     ` Alexander Aring
2024-03-11 19:27       ` Valentin Vidić
2024-03-11 23:43         ` Alexander Aring
2024-03-12  8:09           ` Valentin Vidić
2024-03-12 17:55             ` Alexander Aring
2024-03-12 18:37               ` Valentin Vidić
2024-03-12 19:55                 ` Alexander Aring [this message]
2024-03-12 20:03                   ` Alexander Aring
2024-03-12 20:42                   ` Valentin Vidić

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=CAK-6q+j8Ag5yj6pn0P4wC2kRSNMVaHGmVY+G6+H5aYD1-ygzSw@mail.gmail.com \
    --to=aahringo@redhat.com \
    --cc=gfs2@lists.linux.dev \
    --cc=joseph.qi@linux.alibaba.com \
    --cc=ocfs2-devel@lists.linux.dev \
    --cc=ocfs2-devel@oss.oracle.com \
    --cc=teigland@redhat.com \
    --cc=vvidic@valentin-vidic.from.hr \
    /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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).