Timeout on message publish acknowledge

[gloffreda]

Environment

• VerneMQ Version:1.10.1

• OS:ubuntu 16

• VerneMQ configuration (vernemq.conf) or the changes from the default
  DOCKER_VERNEMQ_DISCOVERY_NODE=10.42.1.1
  DOCKER_VERNEMQ_ACCEPT_EULA=yes
  DOCKER_VERNEMQ_LISTENER__TCP__DEFAULT=0.0.0.0:1883
  DOCKER_VERNEMQ_LISTENER__SSL__DEFAULT=0.0.0.0:8883
  DOCKER_VERNEMQ_LISTENER__HTTP__DEFAULT=0.0.0.0:8888
  DOCKER_VERNEMQ_PLUGINS__VMQ_PASSWD=off
  DOCKER_VERNEMQ_PLUGINS__VMQ_ACL=off
  DOCKER_VERNEMQ_PLUGINS__VMQ_DIVERSITY=on
  DOCKER_VERNEMQ_VMQ_DIVERSITY__SCRIPT_DIR=/vernemq/share/lua
  DOCKER_VERNEMQ_VMQ_DIVERSITY__AUTH_MONGODB__ENABLED=on
  DOCKER_VERNEMQ_VMQ_DIVERSITY__MONGODB__HOST=xxxxxxxxx
  DOCKER_VERNEMQ_VMQ_DIVERSITY__MONGODB__PORT=27017
  DOCKER_VERNEMQ_VMQ_DIVERSITY__MONGODB__LOGIN=xxxxxxxxx
  DOCKER_VERNEMQ_VMQ_DIVERSITY__MONGODB__PASSWORD=xxxxxxxxxxx
  DOCKER_VERNEMQ_VMQ_DIVERSITY__MONGODB__DATABASE=xxxxxxxxx
  DOCKER_VERNEMQ_VMQ_DIVERSITY__MONGODB__SSL=off
  DOCKER_VERNEMQ_PERSISTENT_CLIENT_EXPIRATION=1d
  DOCKER_VERNEMQ_MAX_MESSAGE_SIZE=10000
  DOCKER_VERNEMQ_MAX_ONLINE_MESSAGES=10000
  DOCKER_VERNEMQ_MAX_OFFLINE_MESSAGES=10000
  DOCKER_VERNEMQ_LEVELDB__MAXIMUM_MEMORY__PERCENT=10
  DOCKER_VERNEMQ_SHARED_SUBSCRIPTION_POLICY=prefer_local
  DOCKER_VERNEMQ_LEVELDB__WRITE_BUFFER_SIZE_MIN=1048576
  DOCKER_VERNEMQ_LEVELDB__WRITE_BUFFER_SIZE_MAX=2097152

• Cluster size/standalone:
  2 nodes

Expected behaviour

keep memory stable and clients receiving QoS 1 ack normally

Actual behaviour

I have one major publisher with an average of 20 messages per second. Some peaks with 60 to 70 messages per second during 30 to 50 minutes per day. QoS 1.
It works fine for about one week, then the publisher stops receiving ACK and drops the connection.
Message: Timeout on message publish acknowledge
To reconnect I have to restart the client connection cleaning the message queue to not flood the broker and get a connection closed again.

If I restart VerneMQ, it works fine for about 1 week again.

I noticed the memory keeps around 120MB for 3.. 4 days and then starts to increase. Usually when I need to restart the broker it is around 350MB each node.

Is there any config I can change to stop this behaviour ?

Thanks