[DRBD-user] can't mount on secondary node

Lawrence Strydom qholloi at gmail.com
Wed Feb 1 13:42:41 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.


Ah sorry for the cc oversite.

Network policies pretty much block me from any funky moves so it seems I
will be going with the dual primary, OCFS2, Fencing and Pacemaker solution.
How does dual primary impact performance? Obviously there will be slower IO
due to locks etc but how significant is this?

Thanks

Lawrence


On 1 February 2012 14:19, Felix Frank <ff at mpexnet.de> wrote:

> Hi Lawrence,
>
> please be careful to CC the list (everyone's hearing half a conversation
> otherwise ;-)
>
> On 02/01/2012 12:38 PM, Lawrence Strydom wrote:
> > Hi Felix.
> >
> > My setup is, supposed to be, fairly straightforward.  Two web servers,
> > one in production and the second as a hot standby.
> > The problem comes in with availabillity of IP addresses though. These
> > are hosted servers and the hosting company won't provide more than one
> > Live IP per machine. This means I can't use a floating virtual IP which
> > fails over to the second node as that would require a third live IP.
>
> You could send node1's IP floating and remove node2's public IP. Add
> internal IPs to both machines, so the passive node is reachable by
> hopping through the one with the public IP.
>
> Unless, of course, there are network policies in place that forbid such
> shenanigans.
>
> > Instead the hosting company provides fail over of the real IP of the
> > primary server to the secondary server through their control panel.
> > Weird I know but I didn't choose them, just got given the job.  Ok so
> > now I need data to be synchronous over both servers in case of primary
> > node failure so dual primary would make sense as the IP fail over is
> > already taken care of and no further action will be required to make the
> > data available on the secondary node.
>
> Fair enough. So the answer would probably still be the same: Do use
> pacemaker. The easier way will be to just manage DRBD+Apache, or even
> DRBD only. However, the production node is selected independently from
> pacemaker, which makes for an odd (and error-prone) design.
>
> Otherwise, you can set up Pacemaker + Fencing + OCFS2 + Dual-Primary and
> get away without promotions. But it's more challenging.
>
> HTH,
> Felix
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20120201/344c5037/attachment.htm>


More information about the drbd-user mailing list