The reassignment tool outputs the original assignment before executing the next one. If you have that saved, you can initiate another assignment to make it go to its initial state. That is probably a safer way to fix the reassignment.
On Wed, Dec 17, 2014 at 3:12 PM, Salman Ahmed <ahmed.sal...@gmail.com> wrote: > > I had an issue where one kafka node was filling up on disk space. I used > the reassignment script in an incorrect way, overloading large number of > topics/partition on two target machines, which caused kafka to stop on > those machines. > > I would like to cancel the reassignment process, and restore it to original > or no further reassignment. Has anyone had to edit znodes and clean up > reassignment so replicas and ISR are as they were prior to reassignment. If > so please share the steps. > > Thank you > Sal > -- Thanks, Neha