Try dropping the min compaction threshold via node tool to 2, that should force every compaction bucket to get processed.

Use nodetool flush to trigger minor compaction for all cf's.

Clean will only remove load if you have moved tokens.


Aaron Morton
Freelance Cassandra Developer

On 3/08/2011, at 2:13 AM, Adi <> wrote:

Cassandra 0.7.6 , 4 node cluster 

I ran a nodetool repair for a column family a week ago which caused the load on that node to double. The diagnosis was that I had hit
and that minor compactions will get rid of the unnecessary data and reduce the load. That node has finished doing a round of Minor compactions over the weekend and still the load is 2x of what it is supposed to be. 

Is nodetool cleanup an option to remove the extra data. I am running a cleanup and it has gone through 50% of the data and retained all of it.

Any suggestions?