[DRBD-user] Re: [Vserver] failed to setup a guest on a DRBD-device

Herbert Poetzl herbert at 13thfloor.at
Thu Jun 16 19:26:30 CEST 2005

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On Thu, Jun 16, 2005 at 01:10:13PM +0200, Helmut Wollmersdorfer wrote:
> Herbert Poetzl wrote:
> >On Thu, Jun 16, 2005 at 03:04:10AM +0200, Helmut Wollmersdorfer wrote:
> 
> >>Debian Sarge
> >>kernel-patch-vserver 1.9.5.3
> >>util-vserver   0.30.204-5sarge
> >>kernel-image-2.6.8+vserver_custom01_i386
> >>drbd0.7-module-2.6.8+vserver_0.7.10-3+custom01_i386
> 
> >I'd advise to update to 2.0-rc4 and 0.30.207
> >(IIRC, the latter one should be in broken^Wunstable :)
> 
> With utils upgraded, it is the same.
> Hopefully I have time tomorrow to build 2.0-rc4.
> Or are there working .debs somewhere?
> 
> >>But I would like to have everything belonging to the guest on a 
> >>DRBD-device, to restart the guest in case of failover on the other node.
> 
> >a single guest, or the /vservers (/var/lib/vservers 
> >on debian) path?
> 
> Obviously only one guest per DRBD-device makes sense in conjunction with 
> heartbeat. But for testing^Wtrial /vservers would be enough.
> 
> >should work fine with /dev/drbd0 mounted at /var/lib/vservers
> 
> Thnx, some solutions are so simple;-)
> 
> >but might give 'unexpected' issues on /var/lib/vservers/ha-vser42
> >(the common path for vservers is configured at build time, and
> >for debian this is /var/lib/vservers)
> 
> Is there a special reason for hardcoding?

well, yes and no .. you have to ask Enrico for details here ...

> As a naive user I would expect to configure the /path/to/my/vserver 
> _per_ vserver.
> And newvserver(1) of debian-vservertools says
> 
> | --vsroot          location of "/vserver/" directory Required:
> 
> whatever this means ... it's not the only error in this manpage.

the newvserver is a debian specific tool, which is not
maintained by linux-vserver, so you have to check with the
author (Ola) but I guess it's a 'future' option ...

> >>After some trial and error I did 'cp -a /etc/vservers' to /ha-vser42,
> >>mkdir /ha-vser42/var/<bla>, exchanged /etc/vservers against a symbolic 
> >>link, changed the symbolic links in .defaults
> 
> >why the heck /etc/vservers? those are the config files,
> >and I doubt you want to put those into the drbd?
> 
> Why not. It's a recommended method to keep the config in sync on both 
> nodes of a cluster. But I can live with scp, rsync etc., if 
> /etc/vservers does not change often, AND if I know that it changed.

well, the symlink probably works fine .. but it sounded
more like focusing on the config than on the guests ...

> >>cel600:/ha-vser42/etc/vservers/.defaults# vserver -v vs24 build --force 
> >>-m debootstrap --hostname vs24.nack.at --netdev eth0 --interface 
> >>192.168.0.27/24 --context 24 -- -d sarge -m 
> >>http://ftp.at.debian.org/debian/
> >>chattr: Function not implemented while setting flags on 
> >>/etc/vservers/.defaults/vdirbase/vs24
> >>I: Retrieving debootstrap.invalid_dists_sarge_Release
> >>[...]
> >>I: Extracting libstdc++5...
> >>chroot: cannot run command `mount': Permission denied
> 
> >also make sure that your debootstrap is not
> >buggy or some 'special' security is keeping
> >you from installing a 'working' debian system
> >with debootstrap ...
> 
> I cannot see a difference via 'ls -all' and 'lsattr' between the default 
> /var/lib/vservers and /dev/drbd0 mounted on it.
> Here are the fstab entries:
> 
> /dev/hda7 / ext3    defaults,user_xattr,errors=remount-ro 0      1
> /dev/hda5 none      swap    sw              0       0
> /dev/hda8 /save     ext3    defaults,errors=remount-ro 0       1
> #/dev/drbd0 /var/lib/vservers  ext3    rw,user,noauto,user_xattr,errors=rem$
> /dev/drbd0 /var/lib/vservers  ext3    rw,user,noauto,errors=remount-ro 0 
>    1
> 
> 'user_xattr' or not does not make a difference.
> Maybe I do not understand some basics.
> Hints welcome.
> 
> >if those 'unusual' permission denied messages
> >(please check the kernel logs for more info)
> >remain after updating, please let us know as
> >it might be an issue which results from the
> >combination of drbd and linux-vserver ...
> 
> Nothing found in the logs.
> The only abnormal messages are
> 
> Jun 15 23:16:54 cel600 kernel:  limit=19272832
> Jun 15 23:16:54 cel600 kernel: attempt to access beyond end of device
> Jun 15 23:16:54 cel600 kernel: drbd0: rw=1, want=19319512, limit=19272832
> [last 2 lines repeat very often]

okay, please make sure that you compile the kernel
yourself and enable the VSERVER_DEBUG* options
in the .config, this will allow us to narrow down
the issue by some simple tests (2.0-rc4)

when you have a debug setup running, just pay a 
visit to #vserver and we'll do some testing with you

best,
Herbert

> Maybe 'tar' or 'dump' caused this, as I tried to move a vserver, but I 
> cannot find out the time of these actions. It does not happen during 
> 'vserver build'. To be sure I redid a 'mke2fs -j /dev/drbd0', which does 
> not help.
> 
> >(i.e. we will look into it)
> 
> I cc this to drbd-user, maybe someone has a hint.
> 
> Helmut Wollmersdorfer
> ___________________________________________
> >Vserver mailing list
> >Vserver at list.linux-vserver.org
> >http://list.linux-vserver.org/mailman/listinfo/vserver
> 
> _______________________________________________
> Vserver mailing list
> Vserver at list.linux-vserver.org
> http://list.linux-vserver.org/mailman/listinfo/vserver



More information about the drbd-user mailing list