[system test] disk and memory show negative number in windows

Environment:
8 windows server 2008 R2 64 bit in ec2. Each server has 7.5 GB RAM and 45GB EBS disk to store data.
Create a 6 nodes windows cluster in ec2
Create default bucket with one replica and index enable. Erlang set to +S 16:16

10.96.107.176
10.110.222.56
10.214.194.95
10.111.66.215
10.2.63.249
10.110.206.19

Loading phase:

• Create one default bucket and load 24 million items to this bucket.. Each item has size from 128 to 512 bytes.
• Right after start loading items to cluster, create a doc with 2 views
  curl -v -X PUT -H 'Content-Type: application/json' 'http://ip:8092/default/_design/d1' -d '{"views": { "v1":
  Unknown macro: {"map"}

  , "v2":

  Unknown macro: {"map"}

  }}'

• Wait until indexing reach more than 80%, do query 2 views with command

curl -X GET 'http://IP:8092/default/_design/d1/_view/v1?stale=ok&limit=500'
curl -X GET 'http://IP:8092/default/_design/d1/_view/v2?stale=ok&limit=500'

Then I saw a node in cluster unstable, then down due to control memcached timeout.
At the time one node down, I tried to got information using cbstats but the node did not response.
Then the node came up in UI with N/A in RAM Usage.
Then other node in cluster, one by one became up and down and got the correct value for disk and memory. 4 of 6 nodes shows negative number in both memory and disk storage.
In cluster overview, it shows only total of 2 nodes out of 6 nodes (90 GB instead of 270 GB)

Link to manifest file of build http://builds.hq.northscale.net/latestbuilds/couchbase-server-enterprise_x86_64_2.0.0-1952-rel.setup.exe.manifest.xml