[Linux-cluster] what is it "the same network path"?

Harri.Paivaniemi at tietoenator.com Harri.Paivaniemi at tietoenator.com
Wed Nov 19 14:17:54 UTC 2008


Yes,

If you "network stop" for exampele on node A, it won't see another node anymore and node B won't see node A anymore. When you don't have qdisk nor network tiebraker, both nodes have a right to believe that another node has failed - so even node A that has network stopped can believe node B has failed.

So you end up to fence loop or having both nodes down because both nodes still have a working fence-connection to use to shoot the partner...

If ILO-address lives in the same subnet (same network path) that cluster communication, "network stop" would also drop node A's fence-ILO and it can't fence node ..

By using qdiskd and it's heuristics you can totally avoid this..

-hjp





-----Original Message-----
From: linux-cluster-bounces at redhat.com on behalf of Stepan Kadlec
Sent: Wed 11/19/2008 16:03
To: linux clustering
Subject: [Linux-cluster] what is it "the same network path"?
 
hello,
what is it "the same network path" - mentioned in 
http://sources.redhat.com/cluster/faq.html#two_node_correct ?

I have two node cluster using HP iLO fencing. the iLO interfaces are 
connected in separate VLAN and the interfaces used for CMAN 
communication are connected to another VLAN (but both provided by single 
switch).

do I need quorum disc for such configuration?

currently I don't use quorum disc and the nodes easily end in fence loop 
(if not booted in exactly same time) - is this because I am breaking 
"the same network path" rule? why?

bye stepan.

--
Linux-cluster mailing list
Linux-cluster at redhat.com
https://www.redhat.com/mailman/listinfo/linux-cluster

-------------- next part --------------
A non-text attachment was scrubbed...
Name: winmail.dat
Type: application/ms-tnef
Size: 3396 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20081119/0991ffeb/attachment.bin>


More information about the Linux-cluster mailing list