[DRBD-user] GFS2 freezes

Zohair Raza engineerzuhairraza at gmail.com
Wed Oct 31 11:29:08 CET 2012

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


Hi Lars,

I will have a look on your suggestions.

thanks
On Wed, Oct 31, 2012 at 1:43 PM, Lars Ellenberg
<lars.ellenberg at linbit.com>wrote:

> On Wed, Oct 31, 2012 at 09:33:02AM +0000, Maurits van de Lande wrote:
> > Hello Zohair,
> >
> > >So If I don't have real fencing device, I can't get a cluster?
> >
> > It depends, I have some clusters without fencing, they are mainly for
> > virtualization. The cluster is configured not to relocate the services
> > in case of  a failure. I do this manually. (all services are
> > redundant) The Virtual Machine configuration files are stored on a
> > GFS2 file system. I configured this file system to be always available
> > (also when quorum is lost). Because I do all the recovering manually
> > this does not matter, also no data is written to the GFS2 file system.
> >
> > So, you can have a cluster without fencing if a quorum loss does not
> > corrupt your data. You should also set the cluster timeout's in such a
> > way that even with a WAN connection the cluster does not loose quorum
> > during normal operation. (I have not tried this)
> >
> > For SAMBA you might need "clustered SAMBA"
> http://ctdb.samba.org/samba.html
>
> clustered samba on GFS2 via WAN to remote locations with bandwidth
> constraints,
> not to speak of latency and possible flakyness of the link.
>
> very very VERY BAD idea.
>
> cluster file system accros a WAN: already a bad idea.
>
> cluster file system without fencing is a sure subscription to data loss.
>
> Cluster file systems are very sensitive to latency,
> both storage latency and network latency (for the DLM component).
> So even if you would manage to get it working,
> the performance of it would sucksucksuck big time
> (on a WAN with bandwidth and latency constraints ...).
>
> In a word: Don't.
>
> Maybe look again why you think rsync does not work for you,
> use csync2, couple it with inotify maybe,
> look into lsyncd (which does the inotify part, and can by coupled with
> both rsync or csync2 or any other similar tool), ...
>
> > Did you have a look at glusterfs? http://www.gluster.org/ It supports
> > synchronization (I do not know if it also works over a WAN connection)
> > But it has  nothing to do with drbd. The drbd 8.3 branch is very
> > mature I do not know if this is the same for glusterfs.
> >
> > Best regards,
> >
> > Maurits
> >
> >
> > Van: drbd-user-bounces at lists.linbit.com [mailto:
> drbd-user-bounces at lists.linbit.com] Namens Zohair Raza
> > Verzonden: woensdag 31 oktober 2012 9:51
> > Aan: Felix Frank
> > CC: drbd-user at lists.linbit.com
> > Onderwerp: Re: [DRBD-user] GFS2 freezes
> >
> > So If I don't have real fencing device, I can't get a cluster?
> >
> > My requirement is to synchronized two Samba boxes between remote
> > locations, I can't use rsync because of bandwidth consumption and
> > system processing each time it will run it will go through each file
> > and see if it is synced or not.
> >
> > While GFS seemed to be the right option, but as two servers are
> > distant from each other I can not have fencing device as it may
> > experience power outage or network failures quite often.
> >
> > What do you guys suggest in such scenario?
> >
> > Regards,
> > Zohair Raza
> >
> >
> > On Wed, Oct 31, 2012 at 12:30 PM, Felix Frank <ff at mpexnet.de<mailto:
> ff at mpexnet.de>> wrote:
> > On 10/31/2012 12:02 AM, Lars Ellenberg wrote:
> > >>> Manual fencing is not in any way supported. You must be able to call
> > >>> > > 'fence_node <peer>' and have the remote node reset. If this
> doesn't
> > >>> > > happen, your fencing is not sufficient.
> > >> > fence_node <peer> doesn't work for me
> > >> >
> > >> > fence_node node2 says
> > >> >
> > >> > fence node2 failed
> > > Which is why you need a *real* fencing device
> > > for automatic fencing.
> > ...which is bound to sound more than a little cryptic to the
> > uninitiated, I assume.
> >
> > An example for a "classical" fencing method is a power distribution unit
> > with network access. The surviving node accesses the PDU and cuts the
> > power to its peer.
> > This is just one example. Similar results can be achieved using
> > IPMI/ILOM technologies etc.
> >
> > HTH,
> > Felix
> > _______________________________________________
> > drbd-user mailing list
> > drbd-user at lists.linbit.com<mailto:drbd-user at lists.linbit.com>
> > http://lists.linbit.com/mailman/listinfo/drbd-user
> >
>
> > _______________________________________________
> > drbd-user mailing list
> > drbd-user at lists.linbit.com
> > http://lists.linbit.com/mailman/listinfo/drbd-user
>
>
> --
> : Lars Ellenberg
> : LINBIT | Your Way to High Availability
> : DRBD/HA support and consulting http://www.linbit.com
>
> DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.
> __
> please don't Cc me, but send to list   --   I'm subscribed
> _______________________________________________
> drbd-user mailing list
> drbd-user at lists.linbit.com
> http://lists.linbit.com/mailman/listinfo/drbd-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20121031/1b7837b5/attachment.htm>


More information about the drbd-user mailing list