Roadmap

• Fix the known bugs. There are two bugs that concern me:
  • It seems that there are still some problems with highmem support hidden in the code. Hint: Set highmem to "off" when compiling your kernel for use with DRBD.
  • While DRBD worked fine on top of software raid under Linux 2.2.x it seems that there are some problems under Linux 2.4.x.
• Move to a new major number
• Use zero copy network IO
• Currently the decission about a resynchronisation is found by code in kernel space. Since there are some situations without obvious solutione (E.g. Both parts of the cluster run in split-brain mode...), this decission should be found by some code in user spce. This would make it possible to implement different policies...
• more documentation
• real shared disk semantics for use with GFS.