[lxc-users] LXC not starting with already mounted LV as read-only mount
Claudio Kuenzler
ck at claudiokuenzler.com
Thu Nov 27 08:14:58 UTC 2014
Fajar, thanks for your response and you're right.
I did some testing on the physical host before I wrote to the mailing list
but mounted twice the same LV as rw.
Now I just did as you said and tried to do a rw mount followed by a ro
mount. Indeed the same behavior:
# mount /dev/mapper/vglxc-apps /mnt/1
# mount -o ro /dev/mapper/vglxc-apps /mnt/2
mount: /dev/mapper/vglxc-apps already mounted or /mnt/2 busy
mount: according to mtab, /dev/mapper/vglxc-apps is mounted on /mnt/1
So its not LXC's fault (which I'm glad for!). Very interesting though that
this happens in general.
On Thu, Nov 27, 2014 at 4:42 AM, Fajar A. Nugraha <list at fajar.net> wrote:
> I'm pretty sure it's NOT lxc's fault.
>
> For starters, have you tried testing the same (i.e. mount rw, and then
> mount again ro) in normal, non-lxc setup? That does not work, and does
> indeed fail with a message like this
>
> mount: /dev/zd208 already mounted or /mnt/tmp2 busy
> mount: according to mtab, /dev/zd208 is mounted on /mnt/tmp
>
> So lxc does the right thing: it says "Device or resource busy - failed to
> mount".
>
>
> A more relevant question would be "can you mount something in a path rw,
> while it's ro in another path at the same time". The answer is, AFAIK, no.
> Not directly, or with bind-mount. Either both will be rw, or both will be
> ro.
>
> A workaround would probably be to use an additional fs layer on top of it,
> like nfs or aufs, and then mount that additional layer ro.
>
> --
> Fajar
>
>
> On Wed, Nov 26, 2014 at 7:29 PM, Claudio Kuenzler <ck at claudiokuenzler.com>
> wrote:
>
>> I made a couple of mistakes in my modified patch - so please ignore
>> that.But the basic idea of it was to handle special mount flags (in this
>> case RDONLY) before it hits
>>
>> if (mount(fsname, target, fstype, mountflags & ~MS_REMOUNT, data))
>>
>> Will be glad to test additional patches for you.
>>
>> On Wed, Nov 26, 2014 at 11:19 AM, Claudio Kuenzler <
>> ck at claudiokuenzler.com> wrote:
>>
>>> I applied the patch but it does not work for the read only mount.
>>>
>>> lxc-start 1416993491.172 INFO lxc_conf -
>>> conf.c:check_autodev:3821 - Autodev not required.
>>> lxc-start 1416993491.172 ERROR lxc_conf -
>>> conf.c:mount_entry:1916 - Device or resource busy - failed to mount
>>> '/dev/vglxc/apps' on '/usr/local/lib/lxc/rootfs/mnt/apps'
>>>
>>> I have modified the patch order a bit. Please note that I'm not a
>>> developer so this might be plain wrong:
>>>
>>> 1901 static int mount_entry(const char *fsname, const char *target,
>>> 1902 const char *fstype, unsigned long mountflags,
>>> 1903 const char *data, int optional)
>>> 1904 {
>>> 1905 #ifdef HAVE_STATVFS
>>> 1906 struct statvfs sb;
>>> 1907 #endif
>>> 1908
>>> 1909 if ((mountflags & MS_REMOUNT) || (mountflags & MS_BIND)) {
>>> 1910 DEBUG("remounting %s on %s to respect bind or
>>> remount options",
>>> 1911 fsname ? fsname : "(none)", target ? target :
>>> "(none)");
>>> 1912 unsigned long rqd_flags = 0;
>>> 1913 if (mountflags & MS_RDONLY)
>>> 1914 rqd_flags |= MS_RDONLY;
>>> 1915
>>> 1916
>>> 1917 else if (mount(fsname, target, fstype, mountflags &
>>> ~MS_REMOUNT, data)) {
>>> 1918 if (optional) {
>>> 1919 INFO("failed to mount '%s' on '%s'
>>> (optional): %s", fsname,
>>> 1920 target, strerror(errno));
>>> 1921 return 0;
>>> 1922 }
>>> 1923 else {
>>> 1924 SYSERROR("failed to mount '%s' on '%s'",
>>> fsname, target);
>>> 1925 return -1;
>>> 1926 }
>>> 1927 }
>>> 1928
>>> 1929 #ifdef HAVE_STATVFS
>>>
>>> This time the LV was mounted:
>>>
>>> lxc-start 1416996121.708 INFO lxc_conf -
>>> conf.c:check_autodev:3823 - Autodev not required.
>>> lxc-start 1416996121.708 DEBUG lxc_conf -
>>> conf.c:mount_entry:1975 - mounted '/dev/vglxc/apps' on
>>> '/usr/local/lib/lxc/rootfs/mnt/apps', type 'ext4'
>>> lxc-start 1416996121.708 INFO lxc_conf -
>>> conf.c:mount_file_entries:2224 - mount points have been setup
>>>
>>> But I got different errors - which could be because of the modified code
>>> or because of something else:
>>>
>>> lxc-start 1416996121.730 DEBUG lxc_conf -
>>> conf.c:setup_rootfs_pivot_root:1082 - pivot_root syscall to
>>> '/usr/local/lib/lxc/rootfs' successful
>>> lxc-start 1416996121.731 INFO lxc_conf - conf.c:setup_pts:1641
>>> - created new pts instance
>>> lxc-start 1416996121.731 INFO lxc_conf -
>>> conf.c:setup_personality:1658 - set personality to '0x0'
>>> lxc-start 1416996121.731 ERROR lxc_conf -
>>> conf.c:setup_caps:2339 - unknown capability mac_admin
>>> lxc-start 1416996121.731 ERROR lxc_conf - conf.c:lxc_setup:4174
>>> - failed to drop capabilities
>>> lxc-start 1416996121.731 ERROR lxc_start - start.c:do_start:688
>>> - failed to setup the container
>>> lxc-start 1416996121.731 ERROR lxc_sync - sync.c:__sync_wait:51
>>> - invalid sequence number 1. expected 2
>>> lxc-start 1416996121.731 WARN lxc_conf -
>>> conf.c:lxc_delete_network:3175 - failed to remove interface '(null)'
>>> lxc-start 1416996121.758 ERROR lxc_start -
>>> start.c:__lxc_start:1088 - failed to spawn 'mycontainer'
>>> lxc-start 1416996121.759 WARN lxc_commands -
>>> commands.c:lxc_cmd_rsp_recv:172 - command get_init_pid failed to receive
>>> response
>>> lxc-start 1416996121.759 WARN lxc_cgmanager -
>>> cgmanager.c:cgm_get:963 - do_cgm_get exited with error
>>> lxc-start 1416996126.764 ERROR lxc_start_ui -
>>> lxc_start.c:main:342 - The container failed to start.
>>> lxc-start 1416996126.765 ERROR lxc_start_ui -
>>> lxc_start.c:main:344 - To get more details, run the container in foreground
>>> mode.
>>> lxc-start 1416996126.765 ERROR lxc_start_ui -
>>> lxc_start.c:main:346 - Additional information can be obtained by setting
>>> the --logfile and --logpriority options.
>>>
>>> Will try to redo this in a clean installation.
>>>
>>> On Wed, Nov 26, 2014 at 7:31 AM, Claudio Kuenzler <
>>> ck at claudiokuenzler.com> wrote:
>>>
>>>> I've read that mailing list post yesterday, too and was actually hoping
>>>> this could solve the mount issue. When I looked at it, it seems that this
>>>> patch hasnt been merged into the github repo yet, correct?
>>>>
>>>> Will try it and let you know.
>>>> On Nov 26, 2014 6:35 AM, "Serge Hallyn" <serge.hallyn at ubuntu.com>
>>>> wrote:
>>>>
>>>>> Quoting Claudio Kuenzler (ck at claudiokuenzler.com):
>>>>> > Hello all,
>>>>> >
>>>>> > OS: Ubuntu 14.04.1 LTS
>>>>> > LXC Version: 1.0.6-0ubuntu0.1
>>>>> >
>>>>> > Experienced a weird issue when I tried to launch a container which
>>>>> mounts
>>>>> > an additional logical volume (ext4 formatted) through the
>>>>> container's fstab
>>>>> > file:
>>>>> >
>>>>> > # cat /var/lib/lxc/mycontainer/fstab
>>>>> > /dev/vglxc/apps mnt/apps ext4 ro 0 0
>>>>> >
>>>>> > This LV is already mounted on the physical host:
>>>>> >
>>>>> > # mount /dev/vglxc/apps /mnt
>>>>> >
>>>>> > Now when I try to launch the container, I get the following error:
>>>>> >
>>>>> > # lxc-start -n mycontainer -l trace -o mycontainer.out
>>>>> > lxc-start: Device or resource busy - failed to mount
>>>>> '/dev/vglxc/apps' on
>>>>> > '/usr/lib/x86_64-linux-gnu/lxc/mnt/apps'
>>>>> > lxc-start: failed to setup the mounts for 'mycontainer'
>>>>> > lxc-start: failed to setup the container
>>>>> > lxc-start: invalid sequence number 1. expected 2
>>>>> > lxc-start: failed to spawn 'mycontainer'
>>>>> > lxc-start: The container failed to start.
>>>>> > lxc-start: Additional information can be obtained by setting the
>>>>> --logfile
>>>>> > and --logpriority options.
>>>>> >
>>>>> > In the output file the following information is shown:
>>>>> >
>>>>> > lxc-start 1416921918.699 INFO lxc_conf - Autodev not
>>>>> required.
>>>>> > lxc-start 1416921918.699 ERROR lxc_conf - Device or
>>>>> resource busy
>>>>> > - failed to mount '/dev/vglxc/apps' on
>>>>> > '/usr/lib/x86_64-linux-gnu/lxc/mnt/apps'
>>>>> > lxc-start 1416921918.699 ERROR lxc_conf - failed to setup
>>>>> the
>>>>> > mounts for 'mycontainer'
>>>>> >
>>>>> > When I change the mount option in the container's fstab file from
>>>>> "ro" to
>>>>> > "defaults" (or basically "rw"), the container starts up:
>>>>> >
>>>>> > # cat /var/lib/lxc/mycontainer/fstab
>>>>> > /dev/vglxc/apps mnt/apps ext4 defaults 0 0
>>>>> >
>>>>> > # lxc-start -n mycontainer -l trace -o mycontainer.out
>>>>> >
>>>>> > lxc-start 1416921799.912 INFO lxc_conf - Autodev not
>>>>> required.
>>>>> > lxc-start 1416921799.912 DEBUG lxc_conf - mounted
>>>>> > '/dev/vglxc/apps' on '/usr/lib/x86_64-linux-gnu/lxc/mnt/apps', type
>>>>> 'ext4'
>>>>> >
>>>>> > If the LV /dev/vglxc/apps is nowhere mounted yet, the container
>>>>> starts up -
>>>>> > with the read only mount option, too.
>>>>> >
>>>>> > Is this a (known) bug I ran into?
>>>>>
>>>>> I think it is. Can you please try with this patch:
>>>>>
>>>>>
>>>>> https://lists.linuxcontainers.org/pipermail/lxc-devel/2014-November/010876.html
>>>>>
>>>>> applied?
>>>>> _______________________________________________
>>>>> lxc-users mailing list
>>>>> lxc-users at lists.linuxcontainers.org
>>>>> http://lists.linuxcontainers.org/listinfo/lxc-users
>>>>
>>>>
>>>
>>
>> _______________________________________________
>> lxc-users mailing list
>> lxc-users at lists.linuxcontainers.org
>> http://lists.linuxcontainers.org/listinfo/lxc-users
>>
>
>
> _______________________________________________
> lxc-users mailing list
> lxc-users at lists.linuxcontainers.org
> http://lists.linuxcontainers.org/listinfo/lxc-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxcontainers.org/pipermail/lxc-users/attachments/20141127/2e573a76/attachment.html>
More information about the lxc-users
mailing list