I've never experience this myself, but I do know that running a two-node
cluster with ethernet cross-over cables for the interconnect network can
cause the entire cluster to fail when one node crashes. You must use a
hub/switch for the interconnect networks (in production anyway). And of
course, to avoid single-points-of-failure, you should use a separate
switch for each network.
As for "single points of failure in Linux", I'm not sure what you mean.
Pretty much any computer/operating system (except maybe Tandem) has
single points of failure. This is what RAC is *suposed* to avoid...
By the way -- as for the Dell question -- it sounds like one could
summarise the reported problems with something like: "Avoid implementing
your RAID using the PERC controllers or PowerVault." Is that a fair
assessment? It sounds like a diskless Dell box configured to boot from
a relaible SAN device should be okay...
David wrote:
Have any of you run into the issue whereby you lost an entire RAC db(crash) due to a an instance loss?
Single point of failure does exist for Linux...anyone know where that is in terms of RAC?
-- //www.freelists.org/webpage/oracle-l