Implement proper handling of network partition cases that cause one or more
nodes to disconnect from the group
--------------------------------------------------------------------------------------------------------------
Key: SEQUOIA-1061
URL: https://forge.continuent.org/jira/browse/SEQUOIA-1061
Project: Sequoia
Type: New Feature
Versions: sequoia 2.10.11
Environment: All environments
Reporter: Robert Hodges
Priority: Critical
Sequoia needs to have a robust mechanism for handling network partitions. The
commercial version uni/cluster uses specialized mechanisms adapted to two-node
cases, but Sequoia still needs a more general mechanism that will work for
multi-node situations. This JIRA is a roll-up of requirements and proposed
implementation for solving the problem.
First the requirements:
1.) A controller must be able to recognize situations in which it has left the
active group and cannot continue processing safely. When this happens the
controller must immediately shut down its VDBs.
2.) In the event of a network partition between controllers there must be a
procedure that ensures as many controllers as possible continue to operate. In
other words, we should not shut down all controllers in the event of a
partition.
One way to implement this requirements would be to add proper support for
quorum semantics, which would require controllers to run in odd numbers. There
are a number of issues to consider when taking this approach. Please add
comments to this JIRA if you have thoughts either on the requirements or on the
implementation.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
https://forge.continuent.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
