connecting state in JChannel

[dpy1123]

I want to monitor if every node has connected to jgroup cluster. I'm using JMX's state attrubite.
I setup a 3 node cluster, and I drop tcp packeges for 1 node, and check the state is still connected.

config file:

<config
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xmlns="urn:org:jgroups"
        xsi:schemaLocation="urn:org:jgroups http://www.jgroups.org/schema/jgroups.xsd"
        version="5.2.14.Final">
    <TCP bind_addr="10.100.147.225"
         bind_port="7800"
         recv_buf_size="${tcp.recv_buf_size:130k}"
         send_buf_size="${tcp.send_buf_size:130k}"
         max_length="64K"
         sock_conn_timeout="300"
         thread_pool.min_threads="0"
         thread_pool.max_threads="20"
         thread_pool.keep_alive_time="30000"
         />
    <TCPPING async_discovery="true"
             initial_hosts="10.100.161.247[7800],10.100.153.203[7800],10.100.147.225[7800]"
             port_range="2"/>
    <MERGE3  min_interval="10000"
             max_interval="30000"/>
    <FD_SOCK/>
    <FD_ALL timeout="9000" interval="3000" />
    <VERIFY_SUSPECT timeout="1500"  />
    <BARRIER />
    <pbcast.NAKACK2 use_mcast_xmit="false"
                   discard_delivered_msgs="true"/>
    <UNICAST3 />
    <pbcast.STABLE desired_avg_gossip="50000"
                   max_bytes="4M"/>
    <pbcast.GMS print_local_addr="true"/>
    <UFC max_credits="2M"
         min_threshold="0.4"/>
    <MFC max_credits="2M"
         min_threshold="0.4"/>
    <FRAG2 frag_size="60K"  />
    <pbcast.STATE_TRANSFER/>
    <raft.ELECTION/>
    <raft.RAFT members="10.100.161.247[7800],10.100.153.203[7800],10.100.147.225[7800]" raft_id="10.100.147.225[7800]"/>
    <raft.REDIRECT/>
</config>

7800 port is banned.

[rhusar]

I setup a 3 node cluster, and I drop tcp packeges for 1 node, and check the state is still connected.

What did you use to perform the dropping?

I just remember people trying to simulate network partitions using iptables and configuring DROP, but that's just not going to immediately break the established TCP connections...

[dpy1123]

using this in a k8s cluster.

apiVersion: chaosblade.io/v1alpha1
kind: ChaosBlade
metadata: {name: loss-pod-network-by-names}
spec:
  experiments:
  - action: loss
    desc: loss pod network by names
    matchers:
    - name: names
      value: [test-app-67948b99bf-84qvz]
    - name: namespace
      value: [backend]
    - name: local-port
      value: ['7800']
    - name: remote-port
      value: ['7800']
    - name: interface
      value: [eth0]
    - name: percent
      value: ['100']
    # - name: destination-ip
    #   value: []
    scope: pod
    target: network

[belaban]

If you do this to a connected cluster, FD_ALL will suspect and exclude nodes and TCP connections will be torn down. Re-establishment won't work, so the cluster will stay in disconnected mode.