memmory shows negative number during rebalance in window

8 nodes with 30GB ebs disk for data only, 7.5GB RAM.
Create a 7 node cluster with 1.8.1 build
Create default bucket and load 23 million items to cluster with size from 128 to 512 bytes
Mutate keys with new value from 512 to 1024 bytes
Do online upgrade (swap rebalance, add a new node with 2.0.0-1947 into cluster and remove a 1.8.1 node out of cluster)
Online upgrade passed. Couchbase server in all nodes in cluster swapped to 2.0.0-1947 build.
Create a design doc with 2 view in cluster.
Load more items to cluster.
Let initial index finished
Restart load and do query view on cluster.
Add a new 2.0.0-1947 to cluster and rebalance. During rebalance, one node shows memory size in negative number.

Link to manifest file of build 2.0.0-1947 http://builds.hq.northscale.net/latestbuilds/couchbase-server-enterprise_x86_64_2.0.0-1947-rel.rpm.manifest.xml