NFS is behaving weirdly on a bunch of my systems.unruhhome/
it mounts but the
contents are bizarre. On server, the directory contents look like
ls /local
./ .autofsck http/ lost+found/ .mozilla/ opt/
../ crypt1 jc mageia/ oldinfo/Notebook1/ usershome/ .xdg_menu_cache/
archive/ encrypt/ .kde/ mandriva/ CameraPicturesJul30-2019/persuasion/ usrlocal/
But when mounted on the remote machine ( well on each of 4 machines) the contents of the mounted directory are
ls /local
./ ../ /unruh
Not only is /unruh NOT a directory in /local/ on server, nothing that IS
in /local on server is not on the mounted version.
What could be going on here?
NFS is behaving weirdly on a bunch of my systems.
I have an NFS server (running MGA5) which is supposed to export a
number of directories and partitions to other machines. One of them
is called /local, which is the mounting from a local partition on the
server (let me call it server) to remote machines. exportfs has it
listed. But when I mount that directory on a remote machine, it
mounts but the contents are bizarre. On server, the directory
contents look like ls /local
./ .autofsck http/ lost+found/ .mozilla/
opt/ unruhhome/ ../ crypt1 jc mageia/
oldinfo/ Notebook1/ usershome/ .xdg_menu_cache/ archive/ encrypt/ .kde/ mandriva/ CameraPicturesJul30-2019/ persuasion/ usrlocal/
But when mounted on the remote machine ( well on each of 4 machines)
the contents of the mounted directory are
ls /local
./ ../ /unruh
Not only is /unruh NOT a directory in /local/ on server, nothing that
IS in /local on server is not on the mounted version.
What could be going on here?
NFS is behaving weirdly on a bunch of my systems.unruhhome/
I have an NFS server (running MGA5) which is supposed to export a number
of directories and partitions to other machines. One of them is called /local, which is the mounting from a local partition on the server (let
me call it server) to remote machines. exportfs has it listed.
But when I mount that directory on a remote machine, it mounts but the contents are bizarre. On server, the directory contents look like
ls /local
./ .autofsck http/ lost+found/ .mozilla/ opt/
../ crypt1 jc mageia/ oldinfo/Notebook1/ usershome/ .xdg_menu_cache/
archive/ encrypt/ .kde/ mandriva/ CameraPicturesJul30-2019/persuasion/ usrlocal/
But when mounted on the remote machine ( well on each of 4 machines) the contents of the mounted directory are
ls /local
./ ../ /unruh
Not only is /unruh NOT a directory in /local/ on server, nothing that IS
in /local on server is not on the mounted version.
What could be going on here?
NFS is behaving weirdly on a bunch of my systems.unruhhome/
I have an NFS server (running MGA5) which is supposed to export a number
of directories and partitions to other machines. One of them is called /local, which is the mounting from a local partition on the server (let
me call it server) to remote machines. exportfs has it listed.
But when I mount that directory on a remote machine, it mounts but the contents are bizarre. On server, the directory contents look like
ls /local
./ .autofsck http/ lost+found/ .mozilla/ opt/
../ crypt1 jc mageia/ oldinfo/Notebook1/ usershome/ .xdg_menu_cache/
archive/ encrypt/ .kde/ mandriva/ CameraPicturesJul30-2019/persuasion/ usrlocal/
But when mounted on the remote machine ( well on each of 4 machines) the contents of the mounted directory are
ls /local
./ ../ /unruh
Not only is /unruh NOT a directory in /local/ on server, nothing that IS
in /local on server is not on the mounted version.
What could be going on here?
On 2020-02-20, William Unruh <unruh@invalid.ca> wrote:unruhhome/
NFS is behaving weirdly on a bunch of my systems.
I have an NFS server (running MGA5) which is supposed to export a number
of directories and partitions to other machines. One of them is called
/local, which is the mounting from a local partition on the server (let
me call it server) to remote machines. exportfs has it listed.
But when I mount that directory on a remote machine, it mounts but the
contents are bizarre. On server, the directory contents look like
ls /local
./ .autofsck http/ lost+found/ .mozilla/ opt/
../ crypt1 jc mageia/ oldinfo/ Notebook1/ usershome/ .xdg_menu_cache/
archive/ encrypt/ .kde/ mandriva/ CameraPicturesJul30-2019/ persuasion/ usrlocal/
But when mounted on the remote machine ( well on each of 4 machines) the
contents of the mounted directory are
ls /local
./ ../ /unruh
Not only is /unruh NOT a directory in /local/ on server, nothing that IS
in /local on server is not on the mounted version.
What could be going on here?
hard to say...
does "exportfs -s" on the server give sensible results ?
does "ip addr show" ?
what about "findmnt" on the client? and on the server?
what about "readlink -f /local/."On both server and client
which NFS server software is it?
rpm -qa|grep nfsnfs-utils-1.3.4-4.mga6
rpm -qa |grep nfsnfs-utils-1.3.0-6.mga5
On 2020-02-20, Jasen Betts <jasen@xnet.co.nz> wrote:unruhhome/
On 2020-02-20, William Unruh <unruh@invalid.ca> wrote:
NFS is behaving weirdly on a bunch of my systems.
I have an NFS server (running MGA5) which is supposed to export a number >>> of directories and partitions to other machines. One of them is called
/local, which is the mounting from a local partition on the server (let
me call it server) to remote machines. exportfs has it listed.
But when I mount that directory on a remote machine, it mounts but the
contents are bizarre. On server, the directory contents look like
ls /local
./ .autofsck http/ lost+found/ .mozilla/ opt/
../ crypt1 jc mageia/ oldinfo/ Notebook1/ usershome/ .xdg_menu_cache/
archive/ encrypt/ .kde/ mandriva/ CameraPicturesJul30-2019/ persuasion/ usrlocal/
But when mounted on the remote machine ( well on each of 4 machines) the >>> contents of the mounted directory are
ls /local
./ ../ /unruh
Not only is /unruh NOT a directory in /local/ on server, nothing that IS >>> in /local on server is not on the mounted version.
What could be going on here?
(rw,relatime,vers=4.2,rsize=1048576,wsize=1048576,namlen=255,hard,proto=tcp,tim eo=600,retrans=2,sec=sys,clientaddr=142.123.234.78,local_lock=none,addr=142.123 .234.56)hard to say...
does "exportfs -s" on the server give sensible results ?
Yes.
mount on the client shows
server:/local on /local type nfs4
and those are all the right ip addresses for the server and the client.
does "ip addr show" ?
They show the right addresses on both the server and the client
rw,noatime,data=ordered
what about "findmnt" on the client? and on the server?
On the client it shows the same info as the mount above.
on the server
├─/local /dev/sdb1 ext4
And there is no /local/unruh or /local/unruh/mail mounted from anywhere.
what about "readlink -f /local/."On both server and client
/local
which NFS server software is it?
Not sure what you mean. It is the standard nfs software from mga5
(server) and mga6 (on some clients) and mga7 (on others)
On the client
rpm -qa|grep nfsnfs-utils-1.3.4-4.mga6
lib64nfs8-1.11.0-1.mga6
lib64nfsidmap0-0.27-3.mga6
and on the server
rpm -qa |grep nfsnfs-utils-1.3.0-6.mga5
libnfsidmap-doc-0.25-8.mga5
lib64nfsidmap0-0.25-8.mga5
Note this all worked properly until a couple of weeks ago when
everything had to be shutdown and rebooted due to electrical work.
On Thu, 20 Feb 2020 16:14:50 -0000 (UTC), William Unruh wrote:unruhhome/
On 2020-02-20, Jasen Betts <jasen@xnet.co.nz> wrote:
On 2020-02-20, William Unruh <unruh@invalid.ca> wrote:
NFS is behaving weirdly on a bunch of my systems.
I have an NFS server (running MGA5) which is supposed to export a number >>>> of directories and partitions to other machines. One of them is called >>>> /local, which is the mounting from a local partition on the server (let >>>> me call it server) to remote machines. exportfs has it listed.
But when I mount that directory on a remote machine, it mounts but the >>>> contents are bizarre. On server, the directory contents look like
ls /local
./ .autofsck http/ lost+found/ .mozilla/ opt/
../ crypt1 jc mageia/ oldinfo/ Notebook1/ usershome/ .xdg_menu_cache/
archive/ encrypt/ .kde/ mandriva/ CameraPicturesJul30-2019/ persuasion/ usrlocal/
But when mounted on the remote machine ( well on each of 4 machines) the >>>> contents of the mounted directory are
ls /local
./ ../ /unruh
Not only is /unruh NOT a directory in /local/ on server, nothing that IS >>>> in /local on server is not on the mounted version.
What could be going on here?
I have seen something like that before.
There were files in the mount point that should not have been there.
Once the mount point was actually mounted, then the correct files were available.
Solution was to umount it, delete the files to regain the space and
reduce future confusion, then mount it correctly.
file?hard to say...
does "exportfs -s" on the server give sensible results ?
Yes.
mount on the client shows
server:/local on /local type nfs4 (rw,relatime,vers=4.2,rsize=1048576,wsize=1048576,namlen=255,hard,proto=tcp,tim eo=600,retrans=2,sec=sys,clientaddr=142.123.234.78,local_lock=none,addr=142.123 .234.56)
and those are all the right ip addresses for the server and the client.
What is the client stanza in the server's /etc/exports or /etc/exports.d/*
In my server (wb) I have(rw,nosuid,nodev,noexec,relatime,vers=4.2,rsize=32768,wsize=32768,\
$ grep -v ^# /etc/exports.d/* | grep tb
/spare tb(no_root_squash,sync,no_subtree_check,rw)
$ showmount -e
Export list for wb.home.test:
/spare tb.home.test
And on the client (tb) I have
$ showmount -e wb
Export list for wb:
/spare tb.home.test
And yet my client mount has ,soft,
$ mount | grep :
wb:/spare on /wb_spare type nfs4
namlen=255,soft,proto=tcp,timeo=14,retrans=2,sec=sys,\ clientaddr=192.168.11.100,local_lock=none,addr=192.168.11.132)rw,noatime,data=ordered
whereas yours has ,hard,
does "ip addr show" ?
They show the right addresses on both the server and the client
what about "findmnt" on the client? and on the server?
On the client it shows the same info as the mount above.
on the server
├─/local /dev/sdb1 ext4
And there is no /local/unruh or /local/unruh/mail mounted from anywhere.
what about "readlink -f /local/."On both server and client
/local
which NFS server software is it?
Not sure what you mean. It is the standard nfs software from mga5
(server) and mga6 (on some clients) and mga7 (on others)
On the client
rpm -qa|grep nfsnfs-utils-1.3.4-4.mga6
lib64nfs8-1.11.0-1.mga6
lib64nfsidmap0-0.27-3.mga6
and on the server
rpm -qa |grep nfsnfs-utils-1.3.0-6.mga5
libnfsidmap-doc-0.25-8.mga5
lib64nfsidmap0-0.25-8.mga5
Note this all worked properly until a couple of weeks ago when
everything had to be shutdown and rebooted due to electrical work.
That would lead me to be checking that ip addresses match node names.
On 2020-02-20, Bit Twister <BitTwister@mouse-potato.com> wrote:
On Thu, 20 Feb 2020 16:14:50 -0000 (UTC), William Unruh wrote:
On 2020-02-20, Jasen Betts <jasen@xnet.co.nz> wrote:
On 2020-02-20, William Unruh <unruh@invalid.ca> wrote:
NFS is behaving weirdly on a bunch of my systems.
I have an NFS server (running MGA5) which is supposed to export a number >>>>> of directories and partitions to other machines. One of them is called >>>>> /local, which is the mounting from a local partition on the server (let >>>>> me call it server) to remote machines. exportfs has it listed.
But when I mount that directory on a remote machine, it mounts but the >>>>> contents are bizarre. On server, the directory contents look like
ls /local
./ .autofsck http/ lost+found/ .mozilla/ opt/ unruhhome/
../ crypt1 jc mageia/ oldinfo/ Notebook1/ usershome/ .xdg_menu_cache/
archive/ encrypt/ .kde/ mandriva/ CameraPicturesJul30-2019/ persuasion/ usrlocal/
But when mounted on the remote machine ( well on each of 4 machines) the >>>>> contents of the mounted directory are
ls /local
./ ../ /unruh
Not only is /unruh NOT a directory in /local/ on server, nothing that IS >>>>> in /local on server is not on the mounted version.
What could be going on here?
I have seen something like that before.
There were files in the mount point that should not have been there.
Once the mount point was actually mounted, then the correct files were
available.
No. As I mentioned before mounting, /local was completely empty. After mounting there was one directory thread in /local, namely unruh/mail
with 400 files in there. That subdirectory WAS a subdirectory which was mounted on server, but way inside /local, not at the top of the /local directory.
Solution was to umount it, delete the files to regain the space and
reduce future confusion, then mount it correctly.
I have unmounted and remounted it about 10 times on various machines by
now. No change in the behaviour. Somehow the server is completely
confused.
hard to say...
does "exportfs -s" on the server give sensible results ?
Yes.
mount on the client shows
server:/local on /local type nfs4 (rw,relatime,vers=4.2,rsize=1048576,wsize=1048576,namlen=255,hard,proto=tcp,tim eo=600,retrans=2,sec=sys,clientaddr=142.123.234.78,local_lock=none,addr=142.123 .234.56)
and those are all the right ip addresses for the server and the client.
What is the client stanza in the server's /etc/exports or /etc/exports.d/* file?
/local 142.123.234.0/26
In my server (wb) I have
$ grep -v ^# /etc/exports.d/* | grep tb
/spare tb(no_root_squash,sync,no_subtree_check,rw)
$ showmount -e
Export list for wb.home.test:
/spare tb.home.test
And on the client (tb) I have
$ showmount -e wb
Export list for wb:
/spare tb.home.test
And yet my client mount has ,soft,
$ mount | grep :
wb:/spare on /wb_spare type nfs4 (rw,nosuid,nodev,noexec,relatime,vers=4.2,rsize=32768,wsize=32768,\
namlen=255,soft,proto=tcp,timeo=14,retrans=2,sec=sys,\
clientaddr=192.168.11.100,local_lock=none,addr=192.168.11.132)
whereas yours has ,hard,
does "ip addr show" ?
They show the right addresses on both the server and the client
what about "findmnt" on the client? and on the server?
On the client it shows the same info as the mount above.
on the server
├─/local /dev/sdb1 ext4 rw,noatime,data=ordered
And there is no /local/unruh or /local/unruh/mail mounted from anywhere. >>>
what about "readlink -f /local/."On both server and client
/local
which NFS server software is it?
Not sure what you mean. It is the standard nfs software from mga5
(server) and mga6 (on some clients) and mga7 (on others)
On the client
rpm -qa|grep nfsnfs-utils-1.3.4-4.mga6
lib64nfs8-1.11.0-1.mga6
lib64nfsidmap0-0.27-3.mga6
and on the server
rpm -qa |grep nfsnfs-utils-1.3.0-6.mga5
libnfsidmap-doc-0.25-8.mga5
lib64nfsidmap0-0.25-8.mga5
Note this all worked properly until a couple of weeks ago when
everything had to be shutdown and rebooted due to electrical work.
That would lead me to be checking that ip addresses match node names.
I ssh between them regularly. I ping them regularly. And in all cases i
get the right machine. Now maybe nfs somehow is getting the wrong
address. I will try with the ip address instead of the name. Same
problem.
On 2020-02-20, Bit Twister <BitTwister@mouse-potato.com> wrote:
What is the client stanza in the server's /etc/exports or /etc/exports.d/* file?
/local 142.123.234.0/26
On Thu, 20 Feb 2020 15:11:41 -0500, William Unruh <unruh@invalid.ca> wrote:192.168.10.0/8(no_all_squash,async,secure,no_subtree_check,ro)
On 2020-02-20, Bit Twister <BitTwister@mouse-potato.com> wrote:
What is the client stanza in the server's /etc/exports or /etc/exports.d/* file?
/local 142.123.234.0/26
On my Mageia 7 nfs server, in a configuration setup using mcc, I have ...
# cat /etc/exports
# generated by drakhosts.pl
/etc/profile.d/
Shouldn't matter, but the trailing slash on the directory name may make adifference
(haven't tested to see if it does or not).have
I don't have any Mageia 5 installs still around to check to see if there
been changes in the options used by mcc, or the man page.
Yes, the trouble is that it is the main server, and is the main system I
use, and I have found that reinstalling costs me about a week of time.
Since the almost universal recommendation is to reinstall rather than
upgrade on a major version change, trying to get the new system back to
where the old one was, in terms of software, and configuration is a real
real longterm pain. I might make a copy of my / for 5 and try upgrading
that instead of installing 7 on that partition.
Yes, the trouble is that it is the main server, and is the main system I
use, and I have found that reinstalling costs me about a week of time.
Since the almost universal recommendation is to reinstall rather than
upgrade on a major version change, trying to get the new system back to
where the old one was, in terms of software, and configuration is a real
real longterm pain. I might make a copy of my / for 5 and try upgrading
that instead of installing 7 on that partition.
On Thu, 20 Feb 2020 21:37:10 -0000 (UTC), William Unruh wrote:
Yes, the trouble is that it is the main server, and is the main system I
use, and I have found that reinstalling costs me about a week of time.
Since the almost universal recommendation is to reinstall rather than
upgrade on a major version change, trying to get the new system back to
where the old one was, in terms of software, and configuration is a real
real longterm pain. I might make a copy of my / for 5 and try upgrading
that instead of installing 7 on that partition.
Hmm, seems to me that there were problems for people trying upgrades
and skipping releases like going from 5 to 7 instead of 5 to 6 to 7.
Running unsupported releases could get someone in deep legal dodo.
Saw a cisco security update a week or so ago. Did not get around to installing new firmware on my ~$100 VOIP cisco 7811 phone at home.
Tuesday, I noticed phone did not have dial tone. Phone no longer works
even after reset to factory defaults.
On 2020-02-20, Bit Twister <BitTwister@mouse-potato.com> wrote:
I tried to upgrade the 5 to 6 and then was going to go from 6 to 7. Unfortunately it refused to work. It complained bitterly about not being
able to get rid of libraries which something else depended on.
I did not keep a record.
I finally gave up and reinstalled 7 and have the complete
mess I feared. For example shorewall has been redefined, so that the structure of the files has been changed, and options have been dropped
or redefined (like DROP). Ie, it is a complete mess as I feared. So far
one complete day wasted.
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first
subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
For example, if two exports have the same number, they get confused and things like you describe happen. So ensure the different entries have a explicit different number.
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first
subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
For example, if two exports have the same number, they get confused and
things like you describe happen. So ensure the different entries have a
explicit different number.
And how do I ensure they have different numbers?
I want to (and do) export subdirectories of a mouted filesystem. For
example /local has a subdiretory /local/usrlocal. I want to mount server:/local/usrlocal onto /usr/local of the client. I also want to
mount /local on the client
The weird thing is that when I ran Mga5 I never had any problems with
the nfs. It is only with Mga7 that everything seems to be going to hell.
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first
subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
For example, if two exports have the same number, they get confused and
things like you describe happen. So ensure the different entries have a
explicit different number.
And how do I ensure they have different numbers?
I want to (and do) export subdirectories of a mouted filesystem. For
example /local has a subdiretory /local/usrlocal. I want to mount server:/local/usrlocal onto /usr/local of the client. I also want to
mount /local on the client
Is this liable to confuse nfs? Is it legal to do so? Or can I only
export mountpoints (ie only /local which is mounted from /dev/sdb1 on
the server)?
The problems I seem to be having seem to be on an SSID drive on the
server.
The weird thing is that when I ran Mga5 I never had any problems with
the nfs. It is only with Mga7 that everything seems to be going to hell.
On 22/02/2020 07.03, William Unruh wrote:
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first
subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused >>>> on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/ 192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure)
For example, if two exports have the same number, they get confused and
things like you describe happen. So ensure the different entries have a
explicit different number.
And how do I ensure they have different numbers?
You write them.
I want to (and do) export subdirectories of a mouted filesystem. For
example /local has a subdiretory /local/usrlocal. I want to mount
server:/local/usrlocal onto /usr/local of the client. I also want to
mount /local on the client
Is this liable to confuse nfs? Is it legal to do so? Or can I only
export mountpoints (ie only /local which is mounted from /dev/sdb1 on
the server)?
Better post the exports lines of the server and the fstab of the client,
so that I can see better.
The problems I seem to be having seem to be on an SSID drive on the
server.
The weird thing is that when I ran Mga5 I never had any problems with
the nfs. It is only with Mga7 that everything seems to be going to hell.
nfs version 4 on both?
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure)
Is there some way of checking what numbers are assigned if I do not have
the fsid= as an option. Ie, finding out what it is assigning?
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first
subdirectory, instead of about 20 files or directories as /local has. >>>>>
So it is definitely nfs that has gotten completely and utterly confused >>>>> on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure)
Is there some way of checking what numbers are assigned if I do not have
the fsid= as an option. Ie, finding out what it is assigning?
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure)
Is there some way of checking what numbers are assigned if I do not have
the fsid= as an option. Ie, finding out what it is assigning?
If you haven't configured them, then the file system uuid will be used,
if the file system supports uuid and has one set, if not then it may be
set to 0 which means you may have multiple export roots. Also exporting
multiple mount points on the file system will also generate that you
have same fsid, why it's recommended to just export the common root and
then on the client side mount the right directory.
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
It is an option on the exports file.Is there some way of checking what numbers are assigned if I do not have >>> the fsid= as an option. Ie, finding out what it is assigning?
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>
If you haven't configured them, then the file system uuid will be used,
if the file system supports uuid and has one set, if not then it may be
set to 0 which means you may have multiple export roots. Also exporting
OK, I do have UUIDs on those. So it is probable that it is using the
same fsid for each of those subdirectories that I am exporting from the
main partition.
multiple mount points on the file system will also generate that you
have same fsid, why it's recommended to just export the common root and
then on the client side mount the right directory.
This is really really insane. The exportfs program should assign
different fsid s to each distinct export.
And exporting the whole partition could be very dangerous if there are sensitive things inside those partition. For example I have an encrypted
file on /local. While the encryption should protect the contents, it is insane to given people on all of the clients access to it.
Also I do not know what the purpose is of exporting the root filesystem. There are lots of other partitions mounted on that root, and it is those partitions which I want to use on the other machines, not /.
Or I presume what you meant was that I should export each partition's mountpoint location.
This all strikes me as a real kludge and a very buggy one. Especially as
this is NOT spelled out in the documentation on nfs.
Again, is there any way of finding out what the fsid is that is assigend
to the various exported stuff by exportfs?
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
It is an option on the exports file.Is there some way of checking what numbers are assigned if I do not have >>> the fsid= as an option. Ie, finding out what it is assigning?
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>
If you haven't configured them, then the file system uuid will be used,
if the file system supports uuid and has one set, if not then it may be
set to 0 which means you may have multiple export roots. Also exporting
OK, I do have UUIDs on those. So it is probable that it is using the
same fsid for each of those subdirectories that I am exporting from the
main partition.
multiple mount points on the file system will also generate that you
have same fsid, why it's recommended to just export the common root and
then on the client side mount the right directory.
This is really really insane. The exportfs program should assign
different fsid s to each distinct export.
And exporting the whole partition could be very dangerous if there are sensitive things inside those partition.
For example I have an encrypted
file on /local. While the encryption should protect the contents, it is insane to given people on all of the clients access to it.
Also I do not know what the purpose is of exporting the root filesystem. There are lots of other partitions mounted on that root, and it is those partitions which I want to use on the other machines, not /.
Or I presume what you meant was that I should export each partition's mountpoint location.
This all strikes me as a real kludge and a very buggy one. Especially as
this is NOT spelled out in the documentation on nfs.
Again, is there any way of finding out what the fsid is that is assigend
to the various exported stuff by exportfs?
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first
subdirectory, instead of about 20 files or directories as /local has. >>>>>>
So it is definitely nfs that has gotten completely and utterly
confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>
Is there some way of checking what numbers are assigned if I do not have
the fsid= as an option. Ie, finding out what it is assigning?
Then edit the file and write the option now. Different number for each exported directory.
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first
subdirectory, instead of about 20 files or directories as /local has. >>>>>>
So it is definitely nfs that has gotten completely and utterly confused >>>>>> on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure)
Is there some way of checking what numbers are assigned if I do not have
the fsid= as an option. Ie, finding out what it is assigning?
Then edit the file and write the option now. Different number for each exported directory.
On 22/02/2020 21.47, Carlos E.R. wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first >>>>>>> subdirectory, instead of about 20 files or directories as /local has. >>>>>>>
So it is definitely nfs that has gotten completely and utterly
confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>
Is there some way of checking what numbers are assigned if I do not have >>> the fsid= as an option. Ie, finding out what it is assigning?
Then edit the file and write the option now. Different number for each
exported directory.
fsid=0 for the export root and the rest +1 for each export, this is mentioned in the nfs documentation.
On 22/02/2020 22.11, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
It is an option on the exports file.Is there some way of checking what numbers are assigned if I do not have >>>> the fsid= as an option. Ie, finding out what it is assigning?
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>
If you haven't configured them, then the file system uuid will be used,
if the file system supports uuid and has one set, if not then it may be
set to 0 which means you may have multiple export roots. Also exporting
OK, I do have UUIDs on those. So it is probable that it is using the
same fsid for each of those subdirectories that I am exporting from the
main partition.
Most likely and I guess you don't specify a root (fsid 0) in which you
find the other exported folders.
multiple mount points on the file system will also generate that you
have same fsid, why it's recommended to just export the common root and
then on the client side mount the right directory.
This is really really insane. The exportfs program should assign
different fsid s to each distinct export.
And exporting the whole partition could be very dangerous if there are
sensitive things inside those partition.
No one was saying you should export the whole partition.
For example I have an encrypted
file on /local. While the encryption should protect the contents, it is
insane to given people on all of the clients access to it.
Encryption is only to protect data in rest, nothing else.
Also I do not know what the purpose is of exporting the root filesystem.
There are lots of other partitions mounted on that root, and it is those
partitions which I want to use on the other machines, not /.
The export them properly.
Or I presume what you meant was that I should export each partition's
mountpoint location.
It's always better to export partitions separately and also keep in mind that if you export a mount point before you have mounted the partition,
then the directory will be empty for the client.
The recommendation is to have a nfs export root (path to this can be
what ever you want), bind mount the exports here and set fsid to each
export if you want to be sure that nfs will be completely sure about the exports.
This all strikes me as a real kludge and a very buggy one. Especially as
this is NOT spelled out in the documentation on nfs.
Strange, there is this line for fsid:
For NFSv4, there is a distinguished filesystem which is the root of all
exported filesystem. This is specified with fsid=root or fsid=0 both
of which mean exactly the same thing.
This do also give you a vital hint how you should setup your nfs server regarding to exports. This is a big difference between nfs4 and nfs3 and earlier.
Again, is there any way of finding out what the fsid is that is assigend
to the various exported stuff by exportfs?
You need to do two things, assuming you are using a newer kernel than 2.6.20, check what you configure in your exports file and the use of
blkid for the partitions you export.
On 2020-02-22, J.O. Aho <user@example.net> wrote:/This/Is/my/NFSroot/
On 22/02/2020 22.11, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 19.45, William Unruh wrote:OK, I do have UUIDs on those. So it is probable that it is using the
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
It is an option on the exports file.Is there some way of checking what numbers are assigned if I do not have >>>>> the fsid= as an option. Ie, finding out what it is assigning?
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>>
If you haven't configured them, then the file system uuid will be used, >>>> if the file system supports uuid and has one set, if not then it may be >>>> set to 0 which means you may have multiple export roots. Also exporting >>>
same fsid for each of those subdirectories that I am exporting from the
main partition.
Most likely and I guess you don't specify a root (fsid 0) in which you
find the other exported folders.
multiple mount points on the file system will also generate that you
have same fsid, why it's recommended to just export the common root and >>>> then on the client side mount the right directory.
This is really really insane. The exportfs program should assign
different fsid s to each distinct export.
And exporting the whole partition could be very dangerous if there are
sensitive things inside those partition.
No one was saying you should export the whole partition.
And if I do not, then exporting the subdirectories in that partition
gives me the mess I had to wade through.
For example I have an encrypted
file on /local. While the encryption should protect the contents, it is
insane to given people on all of the clients access to it.
Encryption is only to protect data in rest, nothing else.
I agree. But not making it available in the first place is one more
barrier against it being comprimised.
Also I do not know what the purpose is of exporting the root filesystem. >>> There are lots of other partitions mounted on that root, and it is those >>> partitions which I want to use on the other machines, not /.
The export them properly.
Yes. except that it is not clear what "export them properly" means. I
thought I had exported them properly. It worked for 20 years. Then it
failed to work and landed me in a swamp.
Or I presume what you meant was that I should export each partition's
mountpoint location.
It's always better to export partitions separately and also keep in mind
that if you export a mount point before you have mounted the partition,
then the directory will be empty for the client.
The recommendation is to have a nfs export root (path to this can be
what ever you want), bind mount the exports here and set fsid to each
export if you want to be sure that nfs will be completely sure about the
exports.
This all strikes me as a real kludge and a very buggy one. Especially as >>> this is NOT spelled out in the documentation on nfs.
Strange, there is this line for fsid:
For NFSv4, there is a distinguished filesystem which is the root of all
exported filesystem. This is specified with fsid=root or fsid=0 both
of which mean exactly the same thing.
That I am afraid is pretty opaque to anyone who does not already know
it. For me, root is the / partition. It is not some arbitrary partition.
The root of all the exported filesystems IS / on my system.
But you are saying I believe that I can set up some arbitrary filesystem somewhere, say /This/Is/my/NFSroot/
create directories in this directory which are each of the filesystems I might want to export, bind mount the directories into there, define
as fsid=0 in /exports, and in exports define each of the directories I
want to export relative to this root. And that is not a kludge?
This do also give you a vital hint how you should setup your nfs server
regarding to exports. This is a big difference between nfs4 and nfs3 and
earlier.
Yes, and "earlier" also no longer works.
Again, is there any way of finding out what the fsid is that is assigend >>> to the various exported stuff by exportfs?
You need to do two things, assuming you are using a newer kernel than
2.6.20, check what you configure in your exports file and the use of
blkid for the partitions you export.
kernel 5.5
??? What "use of blkid"?
Again, it appears to be true, from the absense of anyone being able to
tell me how, that it is impossible to know what fsid exportfs assigns to
an exported directory if I do not explicitly enter a fsid= option in /etc/exports for each directory I export.
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 22.11, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
Also I do not know what the purpose is of exporting the root filesystem. >>> There are lots of other partitions mounted on that root, and it is those >>> partitions which I want to use on the other machines, not /.
The export them properly.
Yes. except that it is not clear what "export them properly" means. I
thought I had exported them properly. It worked for 20 years. Then it
failed to work and landed me in a swamp.
/This/Is/my/NFSroot/Or I presume what you meant was that I should export each partition's
mountpoint location.
It's always better to export partitions separately and also keep in mind
that if you export a mount point before you have mounted the partition,
then the directory will be empty for the client.
The recommendation is to have a nfs export root (path to this can be
what ever you want), bind mount the exports here and set fsid to each
export if you want to be sure that nfs will be completely sure about the
exports.
This all strikes me as a real kludge and a very buggy one. Especially as >>> this is NOT spelled out in the documentation on nfs.
Strange, there is this line for fsid:
For NFSv4, there is a distinguished filesystem which is the root of all >> exported filesystem. This is specified with fsid=root or fsid=0 both
of which mean exactly the same thing.
That I am afraid is pretty opaque to anyone who does not already know
it. For me, root is the / partition. It is not some arbitrary partition.
The root of all the exported filesystems IS / on my system.
But you are saying I believe that I can set up some arbitrary filesystem somewhere, say /This/Is/my/NFSroot/
create directories in this directory which are each of the filesystems I might want to export, bind mount the directories into there, define
as fsid=0 in /exports, and in exports define each of the directories I
want to export relative to this root. And that is not a kludge?
This do also give you a vital hint how you should setup your nfs server
regarding to exports. This is a big difference between nfs4 and nfs3 and
earlier.
Yes, and "earlier" also no longer works.
Again, is there any way of finding out what the fsid is that is assigend >>> to the various exported stuff by exportfs?
You need to do two things, assuming you are using a newer kernel than
2.6.20, check what you configure in your exports file and the use of
blkid for the partitions you export.
kernel 5.5
??? What "use of blkid"?
Again, it appears to be true, from the absense of anyone being able to
tell me how, that it is impossible to know what fsid exportfs assigns to
an exported directory if I do not explicitly enter a fsid= option in /etc/exports for each directory I export.
To clarify things for me:
I have a partiton from and ssdisk mounted on /fastlocal. What will the default fsid be assigned if I do not explicitly assign it an fsid in the exports file?
This directory has a subdirectory /fastlocal/usrlocal which I also
export. What will the default fsid be assigned to this subdirectory by exports if I do not assign any fsid= for it in /etc/fstab.
There is another subdirectory directory, /fastlocal/unruh/mail which is mounted from an encryped file package via cryptmount. If I place /fastlocal/unruh/mail/ into /etc/exports with no fsid= line, what fsid
will exportfs assign to this directory. Are any of these liable to have
the same fsid assigned by default by exportfs?
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 21.47, Carlos E.R. wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first >>>>>>>> subdirectory, instead of about 20 files or directories as /local has. >>>>>>>>
So it is definitely nfs that has gotten completely and utterly >>>>>>>> confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>>
Is there some way of checking what numbers are assigned if I do not have >>>> the fsid= as an option. Ie, finding out what it is assigning?
Then edit the file and write the option now. Different number for each
exported directory.
fsid=0 for the export root and the rest +1 for each export, this is
mentioned in the nfs documentation.
As an option, with completely unknown usefullness. Not that it is
absolutely crucial if nfs is to work. It is important enough that it
should be in the sample exports file that is shipped with nfs
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:Is there some way of checking what numbers are assigned if I do not have >>> the fsid= as an option. Ie, finding out what it is assigning?
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first >>>>>>> subdirectory, instead of about 20 files or directories as /local has. >>>>>>>
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>
Then edit the file and write the option now. Different number for each
exported directory.
I have. But I would still like to know what was happening before. It
wasted hours of my time to no purpose.
On 23/02/2020 04.32, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:Is there some way of checking what numbers are assigned if I do not have >>>> the fsid= as an option. Ie, finding out what it is assigning?
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first >>>>>>>> subdirectory, instead of about 20 files or directories as /local has. >>>>>>>>
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>
Then edit the file and write the option now. Different number for each
exported directory.
I have. But I would still like to know what was happening before. It
wasted hours of my time to no purpose.
Well, I told you what was happening :-)
Because I had the same problem once.
Simply give the parent of all your exports number zero, and different numbers to each one of the rest. No need to really export the zero. I don't.
On 23/02/2020 05.06, William Unruh wrote:
To clarify things for me:
I have a partiton from and ssdisk mounted on /fastlocal. What will the
default fsid be assigned if I do not explicitly assign it an fsid in the
exports file?
use blkid to see the UUID of the partition that you mounted to /fastlocal.
This directory has a subdirectory /fastlocal/usrlocal which I also
export. What will the default fsid be assigned to this subdirectory by
exports if I do not assign any fsid= for it in /etc/fstab.
use blkid to see the UUID for the partition, I would hint with it would
be the same.
There is another subdirectory directory, /fastlocal/unruh/mail which is
mounted from an encryped file package via cryptmount. If I place
/fastlocal/unruh/mail/ into /etc/exports with no fsid= line, what fsid
will exportfs assign to this directory. Are any of these liable to have
the same fsid assigned by default by exportfs?
Use blkid and see the UUID of the encrypted partition, in theory it
could the same as for the above, but then your system is quite crap on random values or you changed.
keep in mind that if you do a lot of dd and don't care of changing UUID,
you will end up with many file systems with the same UUID.
On 23/02/2020 04.47, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 22.11, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
Also I do not know what the purpose is of exporting the root filesystem. >>>> There are lots of other partitions mounted on that root, and it is those >>>> partitions which I want to use on the other machines, not /.
The export them properly.
Yes. except that it is not clear what "export them properly" means. I
thought I had exported them properly. It worked for 20 years. Then it
failed to work and landed me in a swamp.
Yes, but you used another older version of nfs, I did my switch like 20
years ago. Back then there was many good online documentations how to configure nfs4, I guess nowadays you may find some not that well written
documentations by people from a country that spits out more than 2
million software developers each year.
Or I presume what you meant was that I should export each partition's
mountpoint location.
It's always better to export partitions separately and also keep in mind >>> that if you export a mount point before you have mounted the partition,
then the directory will be empty for the client.
The recommendation is to have a nfs export root (path to this can be
what ever you want), bind mount the exports here and set fsid to each
export if you want to be sure that nfs will be completely sure about the >>> exports.
This all strikes me as a real kludge and a very buggy one. Especially as >>>> this is NOT spelled out in the documentation on nfs.
Strange, there is this line for fsid:
For NFSv4, there is a distinguished filesystem which is the root of all >>> exported filesystem. This is specified with fsid=root or fsid=0 both >>> of which mean exactly the same thing.
That I am afraid is pretty opaque to anyone who does not already know
it. For me, root is the / partition. It is not some arbitrary partition.
The root of all the exported filesystems IS / on my system.
But you are saying I believe that I can set up some arbitrary filesystem
somewhere, say /This/Is/my/NFSroot/
create directories in this directory which are each of the filesystems I
might want to export, bind mount the directories into there, define /This/Is/my/NFSroot/
as fsid=0 in /exports, and in exports define each of the directories I
want to export relative to this root. And that is not a kludge?
For me it felt quite clear and I did point this out in earlier posts.
This do also give you a vital hint how you should setup your nfs server
regarding to exports. This is a big difference between nfs4 and nfs3 and >>> earlier.
Yes, and "earlier" also no longer works.
You can't expect configurations to just work when switching major
versions of an application, it's like expecting that a modern car has to
be started with a hand crank like the early gas engine cars.
Again, is there any way of finding out what the fsid is that is assigend >>>> to the various exported stuff by exportfs?
You need to do two things, assuming you are using a newer kernel than
2.6.20, check what you configure in your exports file and the use of
blkid for the partitions you export.
kernel 5.5
??? What "use of blkid"?
It's one of the commands that come with linux-utils, it will tell a
block device UUID, the man page will tell you more.
Again, it appears to be true, from the absense of anyone being able to
tell me how, that it is impossible to know what fsid exportfs assigns to
an exported directory if I do not explicitly enter a fsid= option in
/etc/exports for each directory I export.
This is something nfs handles internally, it don't expose it, that's why
you there ain't an easy way to do it and see what nfs has picked and
that is why you need to look at your exports and use blkid.
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 04.32, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:Is there some way of checking what numbers are assigned if I do not have >>>>> the fsid= as an option. Ie, finding out what it is assigning?
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first >>>>>>>>> subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>>
Then edit the file and write the option now. Different number for each >>>> exported directory.
I have. But I would still like to know what was happening before. It
wasted hours of my time to no purpose.
Well, I told you what was happening :-)
Because I had the same problem once.
Simply give the parent of all your exports number zero, and different
numbers to each one of the rest. No need to really export the zero. I don't.
I have no idea what "the parent" is or means. My exports are scattered amongst the subdirectories on the server. So the only "parent" is /
which I sure do not want to export or even have in /etc/exports.
On 2020-02-23, J.O. Aho <user@example.net> wrote:
On 23/02/2020 04.47, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 22.11, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
Also I do not know what the purpose is of exporting the root filesystem. >>>>> There are lots of other partitions mounted on that root, and it is those >>>>> partitions which I want to use on the other machines, not /.
The export them properly.
Yes. except that it is not clear what "export them properly" means. I
thought I had exported them properly. It worked for 20 years. Then it
failed to work and landed me in a swamp.
Yes, but you used another older version of nfs, I did my switch like 20
I used the default nfs of Mandrake/Mandriva/Mageia. I have no idea what
they used.
years ago. Back then there was many good online documentations how to
configure nfs4, I guess nowadays you may find some not that well written
IF fsid are important for the running of nfs, and if all holy hell can
brek out if they are not assigned properly, then it is incumbent on the writers of nfs to a) make sure the defaults are proper and do not allow
that hell to break out, b) are clearly stated in for example the man
pages, and c) have a clear sentence in the example file. The current man
page is entirely opaque and ambiguous.
documentations by people from a country that spits out more than 2
million software developers each year.
Or I presume what you meant was that I should export each partition's >>>>> mountpoint location.
It's always better to export partitions separately and also keep in mind >>>> that if you export a mount point before you have mounted the partition, >>>> then the directory will be empty for the client.
The recommendation is to have a nfs export root (path to this can be
what ever you want), bind mount the exports here and set fsid to each
export if you want to be sure that nfs will be completely sure about the >>>> exports.
This all strikes me as a real kludge and a very buggy one. Especially as >>>>> this is NOT spelled out in the documentation on nfs.
Strange, there is this line for fsid:
For NFSv4, there is a distinguished filesystem which is the root of all
exported filesystem. This is specified with fsid=root or fsid=0 both >>>> of which mean exactly the same thing.
That I am afraid is pretty opaque to anyone who does not already know
it. For me, root is the / partition. It is not some arbitrary partition. >>> The root of all the exported filesystems IS / on my system.
But you are saying I believe that I can set up some arbitrary filesystem >>> somewhere, say /This/Is/my/NFSroot/
create directories in this directory which are each of the filesystems I >>> might want to export, bind mount the directories into there, define /This/Is/my/NFSroot/
as fsid=0 in /exports, and in exports define each of the directories I
want to export relative to this root. And that is not a kludge?
For me it felt quite clear and I did point this out in earlier posts.
This do also give you a vital hint how you should setup your nfs server >>>> regarding to exports. This is a big difference between nfs4 and nfs3 and >>>> earlier.
Yes, and "earlier" also no longer works.
You can't expect configurations to just work when switching major
versions of an application, it's like expecting that a modern car has to
be started with a hand crank like the early gas engine cars.
Yes, I can expect that the defaults "just work".
If I do not assign
fsids in the exports file, then the system assigns them by default.
Those defaults should "just work" whether I come to them from 30 years
of using Linux, or I am a newbe just coming to them today.
I had nothing in my exports file which said anything about fsid. That is
the default, and the default should work properly. I did NOT have any non-unique UUIDs as I have just checked.
Again, is there any way of finding out what the fsid is that is assigend >>>>> to the various exported stuff by exportfs?
You need to do two things, assuming you are using a newer kernel than
2.6.20, check what you configure in your exports file and the use of
blkid for the partitions you export.
kernel 5.5
??? What "use of blkid"?
It's one of the commands that come with linux-utils, it will tell a
block device UUID, the man page will tell you more.
Yes, I do and did know "blkid". I am wondering what your phrase "use of
blkid for the partitions you export" meant. And as I have stated, I do
NOT just export partitions. I export subdirectories of those
partitions. What is the fsid of one of those subdirectories?
Again, it appears to be true, from the absense of anyone being able to
tell me how, that it is impossible to know what fsid exportfs assigns to >>> an exported directory if I do not explicitly enter a fsid= option in
/etc/exports for each directory I export.
This is something nfs handles internally, it don't expose it, that's why
you there ain't an easy way to do it and see what nfs has picked and
that is why you need to look at your exports and use blkid.
Apparently it does it badly. Of course I am assuming it is because of non-unique fsids that I had the trouble I had, and apparently there is
no way of finding out.
Anyway, I do NOT want to infer what the fsid assigned by default were. Iwant to
know what they were. Again, you are skirting the question. HOw do I
check what the default fsids are. Not "How do I read a badly written documentation and infer what they are" And again, using inference, what
are the fsids of subdirectories of a mounted partitions? Are they the
fsids of that mounted partition, or is there some other default
assignment?
On 23/02/2020 17.08, William Unruh wrote:
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 04.32, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:Is there some way of checking what numbers are assigned if I do not have
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote:
On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first >>>>>>>>>> subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>>>
the fsid= as an option. Ie, finding out what it is assigning?
Then edit the file and write the option now. Different number for each >>>>> exported directory.
I have. But I would still like to know what was happening before. It
wasted hours of my time to no purpose.
Well, I told you what was happening :-)
Because I had the same problem once.
Simply give the parent of all your exports number zero, and different
numbers to each one of the rest. No need to really export the zero. I don't.
I have no idea what "the parent" is or means. My exports are scattered
amongst the subdirectories on the server. So the only "parent" is /
which I sure do not want to export or even have in /etc/exports.
Then don't export it, but it figuratively is number zero.
On 23/02/2020 17.37, William Unruh wrote:
On 2020-02-23, J.O. Aho <user@example.net> wrote:
On 23/02/2020 04.47, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 22.11, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
Also I do not know what the purpose is of exporting the root filesystem.
There are lots of other partitions mounted on that root, and it is those
partitions which I want to use on the other machines, not /.
The export them properly.
Yes. except that it is not clear what "export them properly" means. I
thought I had exported them properly. It worked for 20 years. Then it
failed to work and landed me in a swamp.
Yes, but you used another older version of nfs, I did my switch like 20
I used the default nfs of Mandrake/Mandriva/Mageia. I have no idea what
they used.
years ago. Back then there was many good online documentations how to
configure nfs4, I guess nowadays you may find some not that well written
IF fsid are important for the running of nfs, and if all holy hell can
brek out if they are not assigned properly, then it is incumbent on the
writers of nfs to a) make sure the defaults are proper and do not allow
that hell to break out, b) are clearly stated in for example the man
pages, and c) have a clear sentence in the example file. The current man
page is entirely opaque and ambiguous.
documentations by people from a country that spits out more than 2
million software developers each year.
Or I presume what you meant was that I should export each partition's >>>>>> mountpoint location.
It's always better to export partitions separately and also keep in mind >>>>> that if you export a mount point before you have mounted the partition, >>>>> then the directory will be empty for the client.
The recommendation is to have a nfs export root (path to this can be >>>>> what ever you want), bind mount the exports here and set fsid to each >>>>> export if you want to be sure that nfs will be completely sure about the >>>>> exports.
This all strikes me as a real kludge and a very buggy one. Especially as
this is NOT spelled out in the documentation on nfs.
Strange, there is this line for fsid:
For NFSv4, there is a distinguished filesystem which is the root of all
exported filesystem. This is specified with fsid=root or fsid=0 both
of which mean exactly the same thing.
That I am afraid is pretty opaque to anyone who does not already know
it. For me, root is the / partition. It is not some arbitrary partition. >>>> The root of all the exported filesystems IS / on my system.
But you are saying I believe that I can set up some arbitrary filesystem >>>> somewhere, say /This/Is/my/NFSroot/
create directories in this directory which are each of the filesystems I >>>> might want to export, bind mount the directories into there, define /This/Is/my/NFSroot/
as fsid=0 in /exports, and in exports define each of the directories I >>>> want to export relative to this root. And that is not a kludge?
For me it felt quite clear and I did point this out in earlier posts.
This do also give you a vital hint how you should setup your nfs server >>>>> regarding to exports. This is a big difference between nfs4 and nfs3 and >>>>> earlier.
Yes, and "earlier" also no longer works.
You can't expect configurations to just work when switching major
versions of an application, it's like expecting that a modern car has to >>> be started with a hand crank like the early gas engine cars.
Yes, I can expect that the defaults "just work".
Just that your default ain't default for the tool you are using, it's a "default" for a older version of a tool that require rpc to work.
If I do not assign
fsids in the exports file, then the system assigns them by default.
Yes, it does, but it expect you to have a common root for the nfs export.
Those defaults should "just work" whether I come to them from 30 years
of using Linux, or I am a newbe just coming to them today.
nfs4 do work with default configuration, if you make a nfs4
configuration. Don't forget those who switched from nfs2 to nfs3 had
also strange behavior, they also had to change configuration if they
wanted things to keep on working kind of as before.
I had nothing in my exports file which said anything about fsid. That is
the default, and the default should work properly. I did NOT have any
non-unique UUIDs as I have just checked.
But your exports didn't follow the nfs4 standard, so it didn't manage to understand what you tried to do, so it did as well it could.
Again, is there any way of finding out what the fsid is that is assigend
to the various exported stuff by exportfs?
You need to do two things, assuming you are using a newer kernel than >>>>> 2.6.20, check what you configure in your exports file and the use of >>>>> blkid for the partitions you export.
kernel 5.5
??? What "use of blkid"?
It's one of the commands that come with linux-utils, it will tell a
block device UUID, the man page will tell you more.
Yes, I do and did know "blkid". I am wondering what your phrase "use of
blkid for the partitions you export" meant. And as I have stated, I do
NOT just export partitions. I export subdirectories of those
partitions. What is the fsid of one of those subdirectories?
Yes, and as they are on the same file system, they will have the same
UUID. I can say nfs4 handles duplicates as long as you do have a common
root for your exports (fsid=0), then just bind mount all the directories
you want to export there, sure this differs from nfs3, but they are
similar but not the same and those the recommended config is also a bit different.
Again, it appears to be true, from the absense of anyone being able to >>>> tell me how, that it is impossible to know what fsid exportfs assigns to >>>> an exported directory if I do not explicitly enter a fsid= option in
/etc/exports for each directory I export.
This is something nfs handles internally, it don't expose it, that's why >>> you there ain't an easy way to do it and see what nfs has picked and
that is why you need to look at your exports and use blkid.
Apparently it does it badly. Of course I am assuming it is because of
non-unique fsids that I had the trouble I had, and apparently there is
no way of finding out.
Things works badly for you use a nfs3 export, which ain't what you are supposed to be using with nfs4. just do the transformation you should
do, which some of us did already 20 years ago.
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:/etc/exports
On 23/02/2020 17.08, William Unruh wrote:
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 04.32, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:Is there some way of checking what numbers are assigned if I do not have
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote: >>>>>>>>>> On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first >>>>>>>>>>> subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>>>>
the fsid= as an option. Ie, finding out what it is assigning?
Then edit the file and write the option now. Different number for each >>>>>> exported directory.
I have. But I would still like to know what was happening before. It >>>>> wasted hours of my time to no purpose.
Well, I told you what was happening :-)
Because I had the same problem once.
Simply give the parent of all your exports number zero, and different
numbers to each one of the rest. No need to really export the zero. I don't.
I have no idea what "the parent" is or means. My exports are scattered
amongst the subdirectories on the server. So the only "parent" is /
which I sure do not want to export or even have in /etc/exports.
Then don't export it, but it figuratively is number zero.
But if / is in /etc/exports (where I assume you would assign fsid=0 to
it) then you are exporting it. How do you include a directory in
and then not export it, but still assign an fsid to it?
On 23/02/2020 17.23, William Unruh wrote:
Anyway, I do NOT want to infer what the fsid assigned by default were. I want to
know what they were. Again, you are skirting the question. HOw do I
check what the default fsids are. Not "How do I read a badly written
documentation and infer what they are" And again, using inference, what
are the fsids of subdirectories of a mounted partitions? Are they the
fsids of that mounted partition, or is there some other default
assignment?
It's what you get for the partition with blkid unless you configured a
fsid for the export. The important is that you assing the fsid 0/root to
the nfs export root directory, things are generally handled well, just
see to that you have mounted all file systems you want to export before
you start your nfs daemon, or else you may export empty directories.
On 23/02/2020 21.56, William Unruh wrote:
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 17.08, William Unruh wrote:
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 04.32, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 07.03, William Unruh wrote:Is there some way of checking what numbers are assigned if I do not have
On 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote: >>>>>>>>>>> On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first >>>>>>>>>>>> subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>>>>>
the fsid= as an option. Ie, finding out what it is assigning?
Then edit the file and write the option now. Different number for each >>>>>>> exported directory.
I have. But I would still like to know what was happening before. It >>>>>> wasted hours of my time to no purpose.
Well, I told you what was happening :-)
Because I had the same problem once.
Simply give the parent of all your exports number zero, and different >>>>> numbers to each one of the rest. No need to really export the zero. I don't.
I have no idea what "the parent" is or means. My exports are scattered >>>> amongst the subdirectories on the server. So the only "parent" is /
which I sure do not want to export or even have in /etc/exports.
Then don't export it, but it figuratively is number zero.
But if / is in /etc/exports (where I assume you would assign fsid=0 to
it) then you are exporting it. How do you include a directory in /etc/exports
and then not export it, but still assign an fsid to it?
Sigh...
Just imagine you wrote it. But don't write it.
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 21.56, William Unruh wrote:
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 17.08, William Unruh wrote:
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 04.32, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote: >>>>>>>>>> On 22/02/2020 07.03, William Unruh wrote:Then edit the file and write the option now. Different number for each
Is there some way of checking what numbers are assigned if I do not haveOn 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote: >>>>>>>>>>>> On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first
subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>>>>>>
the fsid= as an option. Ie, finding out what it is assigning? >>>>>>>>
exported directory.
I have. But I would still like to know what was happening before. It >>>>>>> wasted hours of my time to no purpose.
Well, I told you what was happening :-)
Because I had the same problem once.
Simply give the parent of all your exports number zero, and different >>>>>> numbers to each one of the rest. No need to really export the zero. I don't.
I have no idea what "the parent" is or means. My exports are scattered >>>>> amongst the subdirectories on the server. So the only "parent" is /
which I sure do not want to export or even have in /etc/exports.
Then don't export it, but it figuratively is number zero.
But if / is in /etc/exports (where I assume you would assign fsid=0 to
it) then you are exporting it. How do you include a directory in /etc/exports
and then not export it, but still assign an fsid to it?
Sigh...
Just imagine you wrote it. But don't write it.
Ah just as I did when it stopped working. and created a mess.
OK, I have always been doing what you suggest, if I now understand it. Althought maybe you mean that is my imagining I did it, that makes it
work:-)
On 2020-02-23, J.O. Aho <user@example.net> wrote:
On 23/02/2020 17.37, William Unruh wrote:
On 2020-02-23, J.O. Aho <user@example.net> wrote:
On 23/02/2020 04.47, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
On 22/02/2020 22.11, William Unruh wrote:
On 2020-02-22, J.O. Aho <user@example.net> wrote:
??? What "use of blkid"?
It's one of the commands that come with linux-utils, it will tell a
block device UUID, the man page will tell you more.
Yes, I do and did know "blkid". I am wondering what your phrase "use of
blkid for the partitions you export" meant. And as I have stated, I do
NOT just export partitions. I export subdirectories of those
partitions. What is the fsid of one of those subdirectories?
Yes, and as they are on the same file system, they will have the same
UUID. I can say nfs4 handles duplicates as long as you do have a common
root for your exports (fsid=0), then just bind mount all the directories
you want to export there, sure this differs from nfs3, but they are
similar but not the same and those the recommended config is also a bit
different.
Sorry, nowhere in the docs does it say that you MUST have a root
filesystem. Nowhere does it say that it will assign the same fsid to two different exported exports. Nowhere does it say what you say it does
(which is stupid to assign the same fsid to two different exports). It
is bug, which you seem intent on making into a feature (features usually
On 23/02/2020 23.11, William Unruh wrote:
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 21.56, William Unruh wrote:
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 17.08, William Unruh wrote:
On 2020-02-23, Carlos E.R. <robin_listas@es.invalid> wrote:
On 23/02/2020 04.32, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote:
On 22/02/2020 19.45, William Unruh wrote:
On 2020-02-22, Carlos E.R. <robin_listas@es.invalid> wrote: >>>>>>>>>>> On 22/02/2020 07.03, William Unruh wrote:Then edit the file and write the option now. Different number for each
Is there some way of checking what numbers are assigned if I do not haveOn 2020-02-21, Carlos E.R. <robin_listas@es.invalid> wrote: >>>>>>>>>>>>> On 20/02/2020 21.20, William Unruh wrote:
OK, It is getting weirder. On server, if I do
mount server:/local /media
I get the same weird behaviour, /media now has unruh as its first
subdirectory, instead of about 20 files or directories as /local has.
So it is definitely nfs that has gotten completely and utterly confused
on server.
Verify the fsid field on the server.
How do I do that?
It is an option on the exports file.
/data/
192.168.1.0/24(fsid=1235,rw,no_root_squash,nohide,no_subtree_check,insecure) >>>>>>>>>>
the fsid= as an option. Ie, finding out what it is assigning? >>>>>>>>>
exported directory.
I have. But I would still like to know what was happening before. It >>>>>>>> wasted hours of my time to no purpose.
Well, I told you what was happening :-)
Because I had the same problem once.
Simply give the parent of all your exports number zero, and different >>>>>>> numbers to each one of the rest. No need to really export the zero. I don't.
I have no idea what "the parent" is or means. My exports are scattered >>>>>> amongst the subdirectories on the server. So the only "parent" is / >>>>>> which I sure do not want to export or even have in /etc/exports.
Then don't export it, but it figuratively is number zero.
But if / is in /etc/exports (where I assume you would assign fsid=0 to >>>> it) then you are exporting it. How do you include a directory in /etc/exports
and then not export it, but still assign an fsid to it?
Sigh...
Just imagine you wrote it. But don't write it.
Ah just as I did when it stopped working. and created a mess.
OK, I have always been doing what you suggest, if I now understand it.
Althought maybe you mean that is my imagining I did it, that makes it
work:-)
But you do have to write proper entries for nfs version 4, remember. And
you need write fsid for every entry in there. Like 1, 2, 3, 4. Just
number them. And reserve the 0 for the root, even if you don't export it.
Just stop complaining and write them numbers. You do not need to wonder
about what would be the automatic number. Just write them.
| Sysop: | Keyop |
|---|---|
| Location: | Huddersfield, West Yorkshire, UK |
| Users: | 483 |
| Nodes: | 16 (2 / 14) |
| Uptime: | 231:02:02 |
| Calls: | 9,612 |
| Files: | 13,686 |
| Messages: | 6,155,340 |