we have an CF which is about 2 gb in size, it has a seondary index on one field (UUID).

the index has a size on disk of about 10 gb. it only shrinks a little when forcing a compaction through jmx.

if i use sstabledump i see a lot of these:

    "partition" : {
      "key" : [ "123c50d1-1ceb-489d-8427-2f34065325f8" ],
      "position" : 306166973
    "rows" : [
        "type" : "row",
        "position" : 306167031,
        "clustering" : [ "f28f46930805495aa7d6cba291d92e87" ],
        "liveness_info" : { "tstamp" : "2017-10-30T16:49:37.160361Z" },
        "cells" : [ ]


normally i can find the key as an indexed field, but most of the keys in the dump do no longer exist in the parent CF.

these keys are sometimes months old. (we have gc_grace_seconds set to 30 mins)

if i use nodetool rebuild_index it does not help, but if i drop the index und recreate it size goes down  two several hundred mb!

what is the reason the cleanup does not work automatically and how can i fix this?


To unsubscribe, e-mail: user-unsubscr...@cassandra.apache.org
For additional commands, e-mail: user-h...@cassandra.apache.org

Reply via email to