disk fetch failed in add in node

Install couchbase server 1.8.0r-703 with hotfix on 7 nodes.
Create a cluster with 4 nodes.
Load 6 million items to cluster to 2x dgm
Add 2 nodes 54 and 55 one by one to cluster and rebalance.
Remove 2 nodes 55 and 54 one by one out of cluster.
Add a new node 56 (not 2 nodes removed before) to cluster.
Check disk fetch on all node. See disk fetch on new added node is zero.
Check stats

[root@win32-2003-3 ~]# /opt/couchbase/bin/cbstats 10.1.3.50:11210 all | grep fetch
ep_bg_fetched: 9973
ep_tap_bg_fetch_requeued: 0
ep_tap_bg_fetched: 827004

[root@win32-2003-3 ~]# /opt/couchbase/bin/cbstats 10.1.3.51:11210 all | grep fetch
ep_bg_fetched: 8939
ep_tap_bg_fetch_requeued: 0
ep_tap_bg_fetched: 1072767
[root@win32-2003-3 ~]# /opt/couchbase/bin/cbstats 10.1.3.52:11210 all | grep fetch
ep_bg_fetched: 8605
ep_tap_bg_fetch_requeued: 0
ep_tap_bg_fetched: 1385824
[root@win32-2003-3 ~]# /opt/couchbase/bin/cbstats 10.1.3.53:11210 all | grep fetch
ep_bg_fetched: 9075
ep_tap_bg_fetch_requeued: 0
ep_tap_bg_fetched: 991581

[root@win32-2003-3 ~]# /opt/couchbase/bin/cbstats 10.1.3.56:11210 all | grep fetch
ep_bg_fetched: 0
ep_tap_bg_fetch_requeued: 0
ep_tap_bg_fetched: 0
[root@win32-2003-3 ~]#