Re: Voting disk TIE
- From: amonte <ax.mount@xxxxxxxxx>
- To: "K Gopalakrishnan" <kaygopal@xxxxxxxxx>
- Date: Thu, 10 May 2007 22:30:57 +0200
Thanks all for the replies.
The possible situations you guys state, have you seen in practice or they
are paper?
It's quite tricky for me to test this since I would need a three node RAC
and currently dont have one :-(
Two nodes is not really useful to check this since the lower node rule
applies, which isnt quite good IMHO, losing interconnect equal RAC total
failure
Alex
On 5/10/07, K Gopalakrishnan <kaygopal@xxxxxxxxx> wrote:
Jeremy,
Majorityof votes /Lowest node number rule is simple. You need majority
of votes to survive. Groups having majority of the votes will win and
evict the other cluster groups/nodes from the cluster. That is why we
say if you want to survive N failures you need to have to 2N+1 voting
disks. Here even after N failures you will have N+1 voting disks
,which is majority of votes for the cluster to survive. This is
'majority of votes rule'
There is a special case, If you have a two node cluster with both the
nodes can access the voting disk(s) and an interconnect interconnect
failure.Here the node with lowest number will survive. This is 'lowest
node number rule'
In short both meant the same. I was trying to address both the
situations in one sentence , and that caused the mis undertanding..
Sorry for the confusion !!
-Gopal
On 5/10/07, Jeremy Paul Schneider <jeremy.schneider@xxxxxxxxxxxxxx> wrote:
> That's not quite accurate and it's a commonly misunderstood.
>
> See:
>
http://download-west.oracle.com/docs/cd/B19306_01/install.102/b14203/storage.htm#sthref677
>
> "An absolute majority of voting disks configured (more than half) must
be
> available and responsive at all times for Oracle Clusterware to
operate."
>
> In other words: if you have six voting disks ...
> and then there's a network failure that causes a split brain scenario
where:
> node 1 -> three disks
> node 2 -> two disks
>
> Then both nodes will shutdown. at least four disks - more than half the
> configured disks - are required for the cluster to operate (at least
> according to the documentation).
>
> If this same scenario happens but you've only configured five voting
disks
> then node 1 will survive.
>
> -Jeremy
>
>
> On 5/10/07, K Gopalakrishnan <kaygopal@xxxxxxxxx> wrote:
> >
> > Tricky situation. Cluster group which is hosting the majority of
> > votes lowest node number will survive. Other groups will be kicked
> > out.
> >
> >
> > --
> > Best Regards,
> > K Gopalakrishnan
> > Co-Author: Oracle Wait Interface, Oracle Press 2004
> > http://www.amazon.com/exec/obidos/tg/detail/-/007222729X/
> >
> > Author: Oracle Database 10g RAC Handbook, Oracle Press 2006
> > http://www.amazon.com/gp/product/007146509X/
> > --
> > //www.freelists.org/webpage/oracle-l
> >
> >
> >
>
>
>
Other related posts:
