Red Hat Bugzilla – Bug 177418
Doc: Need an explanation of logic for split-brain syndrome with RHCS4
Last modified: 2015-04-19 20:46:59 EDT
Description of problem:
Extracted from from email to cluster-list from Shinya Takagi, 1/10/06:
Need an explanation of logic for split-brain syndrome with RHCS4
"...As all we know, RHCS4 doesn't have network tiebreaker anymore especially
4 or more nodes.
Then, how application (in this case. cman is...I guess) will detect to
be a master node and what kind of logic are following to be a master
node ? And I wonder which node will become the master node if 2 and 2
split brain syndrome was occurred in 4 node cluster system. Is that same
logic as 2-node cluster, which is basically faster node to be able to
contact to fencing device (very primitive way :-) ) ?
In RHEL3 CS, each node has a "map" who is a master and always being
So that if split brain was occurred, voting to other nodes and decided
which one will be fenced.
Does it RHCS3 and RHCS4 are used all same logic for fencing ?..."
Version-Release number of selected component (if applicable):
rh-cs(EN)-4-HTML-RHI (2005-11-30T16:52) and
Steps to Reproduce: N/A
Additional info: none
NEEDINFO_ENG has been deprecated in favor of NEEDINFO or ASSIGNED. Changing
status to ASSIGNED for ENG review.
Unless there is a defect in the cluster software, there is no possiblity of a
split-brain cluster. However, the concept of split-brain is covered in RHCS4