Quote:
To make it simple, if you were to share disk space lets say in LVM vgXX/lvol01 between nodes (some form of special clusters...) You would make them all see the same vgXX/lvol01. Only UNIX in principle caches it I/O and more the writes (that is why you should always shutdown properly a such system...), see you now what can happen when many systems access concurrently the same logical volume?
One workaround is having a federating bit of software that is aware of what all nodes are doing and therefore rules the writing process instead of the different systems OS.
A way to not have OS caching is by using raw devices... (Hope you understand something... there are bad day where I don't understand myself, ending up in not being clear in french or english...)
thank you vbe
this is for RAC cluster with four nodes ( i am Trying to understand this )
what load is shared between the nodes (is it only (I/O) of share storage), on which node this software process runs.