Re: [vserver] Problem with delete vserver with SYS_ADMIN set

From: Herbert Poetzl <herbert_at_13thfloor.at>
Date: Fri 26 Aug 2011 - 12:44:58 BST
Message-ID: <20110826114458.GB26071@MAIL.13thfloor.at>

On Fri, Aug 26, 2011 at 08:25:10PM +1000, Steve Kieu wrote:
> By the way the kernel is:

> # uname -a
> Linux seaspray.m5.local 2.6.32.43-vs2.3.0.36.29.7 #1 SMP Wed Aug 24 16:06:37
> EST 2011 sparc64 GNU/Linux

> vserver util is standard debian.

please update to something recent and try again, it
might be that the debian util-vserver doesn't even
create a proper guest environment

TIA,
Herbert

> On Fri, Aug 26, 2011 at 8:24 PM, Steve Kieu <msh.computing@gmail.com> wrote:

>> OK I tried to umount the volue, mkfs.ext3 on it, mount it again with option
>> tag removed. crate a test vserver with these MOUNT flags in ccapabilities
>> and strat it, stop and delete, got error. I ran strace rm -f a file inside
>> the vserver dir and capture its output attached.

>> No speacial erroe at dmesg and message at all. The error message when
>> running vserver delete is just thousand of lines like this

>> remove `/var/lib/vservers/test/sbin/udevadm': Read-only file system

>> I found that if I unmount the /var/lib/vservers and then mount it again,
>> and I am able to remove the directory. so do not need to reboot the whole
>> server to do that

>> cheers

>> On Thu, Aug 25, 2011 at 10:30 PM, Herbert Poetzl <herbert@13thfloor.at>wrote:

>>> On Thu, Aug 25, 2011 at 10:15:24PM +1000, Steve Kieu wrote:

>>>>> an what happens when you do not give those ccapabilities?

>>>> without these, it behaves normally - delete fine, no problem so
>>>> far. It only happens in sparc , not in x86 and x86_64.

>>> that would suggest that something inside the guest
>>> happens which requires those ccapabilities

>>> any idea what that could be on a 'normal' guest
>>> start?

>>>>> could you upload the entire log for that and keep the
>>>>> resulting filesystem for further investigation?

>>>> I will try tomorrow with ext3. Have migrated all vs into
>>>> another host so it can be free to experiment.

>>>> What log are you talking about? dmesg log or kernel.log.

>>> the log of your guest start/stop and the errors you
>>> get (basically a copy of the shell input/output)
>>> but of course, if you get some output in dmesg,
>>> please add that as well

>>>> Should I strace it when
>>>> I do a rm -rf /var/lib/vservers/NAME?
>>>> (after vserver NAME delete failed, I tried to manually
>>>> rm -rf it and same error.

>>> yes, please do so, from the information you provided
>>> so far, it sounds like something inside the guest is
>>> either modified or mounted (in some strange way) so
>>> this information could be quite helpful

>>>> I can work around this limitation after all we I hit the brick
>>>> wall - will sshfs mount from the host through vserver fstab
>>>> which will work fine I guess.It might even be better solution
>>>> in terms of security; vserver is still only virtualization for
>>>> sparc for now (lxc is still far more unstable in my test). so
>>>> it is good if this can be fixed.

>>> I'm pretty sure we can identify the issue and if it is
>>> actually Linux-VServer related we will fix it :)

>>> best,
>>> Herbert

>>> PS: if you have some time for interactive testing, you
>>> could pay a visit to the IRC channel and look for Bertl

>>>> Thanks,

>>>>> Repeat with mkfs.ext4 - still the same

>>>>> let's stick with ext3 for now, that's easier to debug :)

>>>> -- Steve Kieu

>> --
>> Steve Kieu

> --
> Steve Kieu
Received on Fri Aug 26 12:45:11 2011

[Next/Previous Months] [Main vserver Project Homepage] [Howto Subscribe/Unsubscribe] [Paul Sladen's vserver stuff]
Generated on Fri 26 Aug 2011 - 12:45:12 BST by hypermail 2.1.8