with simple scenario get: "All memory allocated to this bucket is used for metadata"

http://qa.hq.northscale.net/job/centos_x64--03_01--failover_tests-P0/8/console

./testrunner -i resources/jenkins/centos-64-5node-failover.ini GROUP=P0,get-cbcollect-info=True, -t failover.failovertests.FailoverTests.test_failover_firewall,replicas=1,items=100000,dgm_run=True,GROUP=P0

there are no any specific in the scenario, except bucket was created with enough small ram quota

bucketType=membase&evictionPolicy=valueOnly&threadsNumber=3&ramQuotaMB=170&proxyPort=11211&authType=sasl&name=default&flushEnabled=1&replicaNumber=1&replicaIndex=1&saslPassword=None

then 1+rebalance 4 nodes

start data load
after 3 minutes we get error on server:

Hard Out Of Memory Error. Bucket "default" on node 10.1.3.141 is full. All memory allocated to this bucket is used for metadata. (repeated 15 times)

from test:
2014-03-11 02:40:31 | ERROR | MainProcess | load_gen_task | [data_helper._setMulti_seq] ["Memcached error #130 'No Memory': Out of memory for vbucket :820 to mc 10.1.3.141:11210"]
Traceback (most recent call last):
File "lib/memcached/helper/data_helper.py", line 1056, in _setMulti_rec
rec_caller_fn(exp, flags, keyval, pause, timeout - pause) # Start all over again for these key vals.
File "lib/memcached/helper/data_helper.py", line 1021, in _setMulti_seq
raise errors[0]
MemcachedError: Memcached error #130 'No Memory': Out of memory for vbucket :820 to mc 10.1.3.141:11210

nodes:
seems like tests suite also is broken because we try to reach dgm(dgm_run=True) but only 4017 items were loaded since the beginning of the test but they are constantly updated