• [gentoo-user] NFS mounting - SOLVED

    From Peter Humphrey@21:1/5 to All on Tue Nov 26 17:20:02 2024
    Greetings,

    [New readers start here... :) ]

    I've spent several days-worth of my time over the last few weeks in trying to get my i5 box to export its portage tree and packages directory to a chroot on my M9 machine. I read all the docs, I thought about the help that was offered here, I changed file systems and partitions around - everything I could think of.

    The answer was simple, and I stumbled over it in a post on Stack Exchange: the behaviour of NFS mount calls changed in NFS v4. I didn't need to change /etc/ exports on the i5, but the NFS-mount call on the M9 did need to change.

    # cat /etc/exports # on the i5
    /mnt/nfs \
    192.168.178.7(rw,sync,no_subtree_check,no_root_squash,crossmnt,fsid=0) /mnt/nfs/portage \
    192.168.178.7(rw,sync,insecure,no_subtree_check,no_root_squash) /mnt/nfs/packages \
    192.168.178.7(rw,sync,insecure,no_subtree_check,no_root_squash)

    $ grep nfs /etc/init.d/nuci # on the M9. Nuci is the I5
    mount -t nfs 192.168.178.4:portage /mnt/nuci/var/db/repos/gentoo
    mount -t nfs 192.168.178.4:packages /mnt/nuci/var/cache/packages

    Notice the absence of '/mnt/nfs' from the M9 mount commands. That's what was tripping me up all that time.

    Someone needs to have a look at the nfs-utils wiki page. I'd do something myself, but how? I raised a bug against a document once, only to be rebuked.

    --
    Regards,
    Peter.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Michael@21:1/5 to All on Mon Dec 2 17:56:38 2024
    On Tuesday 26 November 2024 16:13:01 GMT Peter Humphrey wrote:
    Greetings,

    [New readers start here... :) ]

    I've spent several days-worth of my time over the last few weeks in trying
    to get my i5 box to export its portage tree and packages directory to a chroot on my M9 machine. I read all the docs, I thought about the help that was offered here, I changed file systems and partitions around - everything
    I could think of.

    The answer was simple, and I stumbled over it in a post on Stack Exchange: the behaviour of NFS mount calls changed in NFS v4. I didn't need to change /etc/ exports on the i5, but the NFS-mount call on the M9 did need to
    change.

    Can you please share the link?

    # cat /etc/exports # on the i5
    /mnt/nfs \
    192.168.178.7(rw,sync,no_subtree_check,no_root_squash,crossmnt,fsid=0) /mnt/nfs/portage \
    192.168.178.7(rw,sync,insecure,no_subtree_check,no_root_squash) /mnt/nfs/packages \
    192.168.178.7(rw,sync,insecure,no_subtree_check,no_root_squash)

    $ grep nfs /etc/init.d/nuci # on the M9. Nuci is the I5
    mount -t nfs 192.168.178.4:portage /mnt/nuci/var/db/repos/gentoo
    mount -t nfs 192.168.178.4:packages /mnt/nuci/var/cache/packages

    Notice the absence of '/mnt/nfs' from the M9 mount commands. That's what was tripping me up all that time.

    Someone needs to have a look at the nfs-utils wiki page. I'd do something myself, but how? I raised a bug against a document once, only to be rebuked.

    I had (another) look at the wiki. You're right, it seems to describe NFSv3 only. I don't have NFSv3 here to compare. With NFSv4 you export the global root directory to allow its subdirectories to be exported too - at least this is how I understand it works on my systems.
    -----BEGIN PGP SIGNATURE-----

    iQIzBAABCAAdFiEEXqhvaVh2ERicA8Ceseqq9sKVZxkFAmdN9NYACgkQseqq9sKV ZxlRgA/+O/kSc2Aj1oikVnb2fqXjaDjKEj8NfPAxMdKVtWYb0SCNYd6MCfo3mByd XXkQwecos1ZVk8XsJm4gLYM3M8jV3Ime9M+Gg6FmQEO1Qhfl69S40yvUw1qeJ08p 566K4okp4xBslpVVHgK0BW8yQvwchzEPiiBjhUm8uN7ofuQufadz9rSWlQZ/oiaM xECfG3y0u+EDdEkcqkxfnXNfBVrTXJH/+ze7lGkmbBXHqevxmplEDHCrd6ej3zyH 1aSR9IVbKDTfPi/w2ONpTtp1eumSQUuCa8UblckyhGqJhX2/xLSus0I2T+t4KAtq /KtMGeEaqKi4n9DUgO7IiIRTB5OTvr3002/zAhgF3+k6NXn77knND2ew+49E0HkM qVy5dQCIXz6JXiT5TIFzmpjsGxwmeKjt/lx+e4oXtaaOjwcgb4+47jqKwaM/O5AB s96x40vgCi4SAHD7J713VLUEx0OGcT/OzMezqGRhfs7JGzMwxWrJdnlV+s+LUCb7 gmqb4lqKS42aw0o2TCG3xKc+uTJIZ+DCCOTSCSU78BgzcMEtWkPVHVW51fHL/gRP UZg6/luuUpwwA+Gt4wrh9gBunldHdHOHW8IVq23Gg8Y94Es894PIQkJMpzkBRhyn N16Lbd527Fw4opMwvjhUAe84CzzvXUYHwipw+TWjmkdW0EpNfFY=
    =mAZa
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Peter Humphrey@21:1/5 to All on Tue Dec 3 12:50:01 2024
    On Tuesday 3 December 2024 11:44:50 Greenwich Mean Time Michael wrote:

    ... I think there should be clearer disambiguation with separate examples between v3 and v4. However, isn't NFS v3 considered legacy by now?

    Perhaps, but that wiki page was apparently last changed on 2 August this year.

    --
    Regards,
    Peter.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Peter Humphrey@21:1/5 to All on Tue Dec 3 12:30:01 2024
    On Monday 2 December 2024 17:56:38 Greenwich Mean Time Michael wrote:
    On Tuesday 26 November 2024 16:13:01 GMT Peter Humphrey wrote:
    I've spent several days-worth of my time over the last few weeks in trying to get my i5 box to export its portage tree and packages directory to a chroot on my M9 machine. I read all the docs, I thought about the help
    that was offered here, I changed file systems and partitions around - everything I could think of.

    The answer was simple, and I stumbled over it in a post on Stack Exchange: the behaviour of NFS mount calls changed in NFS v4. I didn't need to
    change /etc/ exports on the i5, but the NFS-mount call on the M9 did need to change.

    Can you please share the link?

    Sorry Michael - it was at the end of infinitely many searches and I didn't make a note of it.

    <snip> I should have said: "in NFSv4 the client specifies the server path *relative* to the virtual root."

    I had (another) look at the wiki. You're right, it seems to describe NFSv3 only. I don't have NFSv3 here to compare. With NFSv4 you export the global root directory to allow its subdirectories to be exported too - at least
    this is how I understand it works on my systems.

    ...but today I see, buried in what you might call the small print: "If the NFS server or client support NFSv3 only, the full path to the exported directory (e.g. /export/home or /export/data) needs to be specified when mounting." [1]

    We could debate whether that's guidance enough.

    1. https://wiki.gentoo.org/wiki/Nfs-utils#Mounting_exports , about a page down.

    --
    Regards,
    Peter.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Michael@21:1/5 to All on Tue Dec 3 11:44:50 2024
    On Tuesday 3 December 2024 11:29:15 GMT Peter Humphrey wrote:
    On Monday 2 December 2024 17:56:38 Greenwich Mean Time Michael wrote:

    I had (another) look at the wiki. You're right, it seems to describe
    NFSv3
    only. I don't have NFSv3 here to compare. With NFSv4 you export the global root directory to allow its subdirectories to be exported too - at least this is how I understand it works on my systems.

    ...but today I see, buried in what you might call the small print: "If the NFS server or client support NFSv3 only, the full path to the exported directory (e.g. /export/home or /export/data) needs to be specified when mounting." [1]

    We could debate whether that's guidance enough.

    1. https://wiki.gentoo.org/wiki/Nfs-utils#Mounting_exports , about a page down.

    Yes, but further up all the IPv4 configuration syntax still shows v3 style examples with '/export/foo' paths. I think there should be clearer disambiguation with separate examples between v3 and v4. However, isn't NFS
    v3 considered legacy by now?
    -----BEGIN PGP SIGNATURE-----

    iQIzBAABCAAdFiEEXqhvaVh2ERicA8Ceseqq9sKVZxkFAmdO7zIACgkQseqq9sKV ZxnahxAApLLnuqAj3Xq+v5JOEZuorJU7Pl1oooAO1KtBliYCXzkqzEVBKFioT6UO X8sazU3v4U9+wBcTvsd2/m8rSBD2tBChoyHrMFl6mqAJguR0t4vrVdHMGyZ3h3xV n5FgSfqoxhS7J1akiOeeYGizIzZYF81GTXyg0F8atRTc4gnz+6fiy0394nTr8pBY pjNBTAi7blFknFvIhhBOxtBenxcB1sQEL3NW52S9Qzwi0LHMxnnRNpinvYEg4Wuz TSaX/oey7qfWkE5WmciYfeURhY8r1mlHJCycrwLrdB0fDW20qLUakjBV5hAFrR3U 27xfUg0T8HdDrj8hmyxd7fJJd+qa5TZVwkCpFbMyaQGAJ9YAFTPCWsrJiDL2X9Ab u4UjIBg7kgMdzJ0V5jrZSMtsLYsKw1upST6ksOkQTpGr5jMTc+xKCl2Rp9iEh0Ys XQg16hSvq+zuLyF6DP2rkqY++BTLWPoO/GNEfhOHI1PQ8gbA1RuAt0WqbdvglB/y 4VtnBtPfEyrLTMfH9FJnNiyHNM8ehOgB3GxzLB1QQCt950ETkQi3JxW6GQyZdMC2 Xp2B0uyFnoQ7vcpBZfKRMioXNQY4wkqcY0bXWPHnFsk/CMR1A3ftXu92ZjjDMV/d uYsnV+3ENFBQh+p7tBHWNWbHO09HfA8ns1kyXu7dC8SbryGG77U=
    =Vops
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Matt Jolly@21:1/5 to Peter Humphrey on Tue Dec 3 14:10:01 2024
    Hi Peter,

    On 27 November 2024 2:13:01 am AEST, Peter Humphrey <peter@prh.myzen.co.uk> wrote:
    Someone needs to have a look at the nfs-utils wiki page. I'd do something >myself, but how? I raised a bug against a document once, only to be rebuked.

    You can raise issues on the "Talk" page for a given article, e.g. https://wiki.gentoo.org/wiki/Talk:Nfs-utils

    Ideally, since it's a wiki, if you know how to fix it you can edit the page directly. Don't be afraid, other editors will help polish your contribution if it's a little rough around the edges as long as it's complete.

    Trying this from Thunderbird mobile. Hopefully it doesn't mangle the reply!

    Cheers,

    Matt

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Peter Humphrey@21:1/5 to All on Tue Dec 3 14:30:01 2024
    On Tuesday 3 December 2024 13:08:51 Greenwich Mean Time Matt Jolly wrote:
    Hi Peter,

    On 27 November 2024 2:13:01 am AEST, Peter Humphrey <peter@prh.myzen.co.uk>
    wrote:
    Someone needs to have a look at the nfs-utils wiki page. I'd do something >myself, but how? I raised a bug against a document once, only to be >rebuked.
    You can raise issues on the "Talk" page for a given article, e.g. https://wiki.gentoo.org/wiki/Talk:Nfs-utils

    Ideally, since it's a wiki, if you know how to fix it you can edit the page directly. Don't be afraid, other editors will help polish your contribution if it's a little rough around the edges as long as it's complete.

    Trying this from Thunderbird mobile. Hopefully it doesn't mangle the reply!

    That's a real help; thank you Matt.

    --
    Regards,
    Peter.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Peter Humphrey@21:1/5 to All on Wed Dec 4 03:30:01 2024
    On Tuesday 3 December 2024 13:28:44 Greenwich Mean Time I wrote:
    On Tuesday 3 December 2024 13:08:51 Greenwich Mean Time Matt Jolly wrote:
    Hi Peter,

    On 27 November 2024 2:13:01 am AEST, Peter Humphrey <peter@prh.myzen.co.uk>
    wrote:
    Someone needs to have a look at the nfs-utils wiki page. I'd do something >myself, but how? I raised a bug against a document once, only to be >rebuked.

    You can raise issues on the "Talk" page for a given article, e.g. https://wiki.gentoo.org/wiki/Talk:Nfs-utils

    Ideally, since it's a wiki, if you know how to fix it you can edit the
    page
    directly. Don't be afraid, other editors will help polish your
    contribution
    if it's a little rough around the edges as long as it's complete.

    Trying this from Thunderbird mobile. Hopefully it doesn't mangle the
    reply!

    That's a real help; thank you Matt.

    I've made a suggestion about the nfs-utils page, but I've also been thinking about Gentoo wiki documents generally, because I find them unsatisfactory: not their content, but the style of presentation. I was documentation manager on a 200-man-year software project years ago (supplier side), so I think I know what I'm talking about [1].

    One problem is the apparent absence of structure in the body of the document. A table of contents appears at the top, complete with numbered sections and subsections in a clear hierarchy, but those numbers appear nowhere else. The complexity of most of these documents is such that, once deep into the text, the relationship with the rest of it is invisible. This sounds academic, but it's real; reliance on font sizes to distinguish headings is not enough on its own. An immediate improvement would result from including the numbers from the contents list. I imagine it would be comparatively easy to do, as well.

    Secondly, the big advantage of viewing on a screen has largely been discarded - colour. There are pale coloured backgrounds in some highlights, but I wonder whether more could be achieved. I dimly remember the adoption of the present style (when was that?), following a 'paper' scheme. (I think that name's right.) A consistent style is essential, of course, but why go backwards to an earlier technology?

    Thirdly, showing terminal commands on a huge black background is much too disruptive, visually. They're so overpowering that things like headings disappear. I find, especially now my vision is deteriorating, that the interleaving of main and secondary topics is often baffling: for instance, the choice between openrc and systemd with other alternative streams.

    Not enough thought has been given to the combined effect, which is to make the documents hard to read and understand.

    What does the team think can be done about it?

    1. The fact that the project was cancelled when it became clear that the first
    50% of the work had taken the first 80% of the time, and the second 50% would take the other 80% of the time, had nothing to do with the documents. :_)

    --
    Regards,
    Peter.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Peter Humphrey@21:1/5 to All on Thu Dec 5 03:00:01 2024
    On Monday 2 December 2024 17:56:38 GMT Michael wrote:
    On Tuesday 26 November 2024 16:13:01 GMT Peter Humphrey wrote:
    Greetings,

    [New readers start here... :) ]

    I've spent several days-worth of my time over the last few weeks in trying to get my i5 box to export its portage tree and packages directory to a chroot on my M9 machine. I read all the docs, I thought about the help
    that
    was offered here, I changed file systems and partitions around -
    everything
    I could think of.

    The answer was simple, and I stumbled over it in a post on Stack Exchange: the behaviour of NFS mount calls changed in NFS v4. I didn't need to
    change
    /etc/ exports on the i5, but the NFS-mount call on the M9 did need to change.

    Can you please share the link?

    https://unix.stackexchange.com/questions/388410/nfs-mount-failed-reason-given-by-server-no-such-file-or-directory

    It's comment 15.

    I found this after I gave up running ~amd64 and reverted to a backup from 12 days ago.

    --
    Regards,
    Peter.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)