[system test] memcached hang during access phase

Enviroment:
8 nodes centos 6.2 64-bit (each has 32 GB RAM and SSD )
1:10.6.2.42
2:10.6.2.43
3:10.6.2.44
4:10.6.2.45
5:10.6.2.66
6:10.6.2.68
7:10.6.2.69

8:10.6.2.89

Setup:
Create a 7 nodes cluster with couchbase server 2.5.0-960.
Create 2 buckets:
sasl-1 with 1 replica, 1 doc with 1 view
sasl-2 with 2 replica, 1 doc with 1 view

Load items with size from 128 bytes to 512 bytes into cluster.
Load dgm until active resident got 90% at each bucket.
At this load dgm phase, memcached on node 10.6.2.69 hang

I check cbstats at other nodes, it runs ok
cbstats in node 69 is hang due to memcached hang

[root@pine-11804 ~]# /opt/couchbase/bin/cbstats 10.6.2.68:11210 raw warmup -b sasl-2 -p password
ep_warmup: enabled
ep_warmup_dups: 0
ep_warmup_estimate_time: 46
ep_warmup_estimated_key_count: 0
ep_warmup_estimated_value_count: 0
ep_warmup_item_expired: 0
ep_warmup_key_count: 0
ep_warmup_keys_time: 337
ep_warmup_min_item_threshold: 100
ep_warmup_min_memory_threshold: 100
ep_warmup_oom: 0
ep_warmup_state: done
ep_warmup_thread: complete
ep_warmup_time: 539
ep_warmup_value_count: 0
root@pine-11804 ~]# /opt/couchbase/bin/cbstats 10.6.2.69:11210 raw warmup -b sasl-1 -p password
^Z
[3]+ Stopped /opt/couchbase/bin/cbstats 10.6.2.69:11210 raw warmup -b sasl-1 -p password

Cluster is in hang state