On Wed, Oct 14, 2015 at 10:03 PM, Saladi Naidu <naidusp2...@yahoo.com>
wrote:

> Thanks for the reply. yes this indeed due to range tombstones with MAP
> data, even after tombstone past the gc_grace_poeriod and compactions ran in
> cluster, still no change with tombstone data in SSTables. Do you or nayone
> in the group know how to delete this bad data from the cluster?
>

Other than applying the patch from the upthread tickets, you could :

1) stop node

2) for each sstable affected:
a) dump to JSON with sstable2json
b) remove all duplicate range tombstones from JSON
c) recreate sstable with same name via json2sstable

3) start node

However based on my understanding of this issue, they'll just build up
again until you are running a version with the upthread patches.

=Rob

Reply via email to