Re: Which ports does ignite cluster need to run normally?

[Juan Carlos Franzoy <jf...@gmail.com>]

I am facing a similar problem, trying to use apache-ignite 2.6.
May be, you can help me.

I set up my discovering using TcpDiscoverySpi and
TcpDiscoveryMulticastIpFinder:
    ipFinder.setMulticastGroup(228.10.10.157)
    ipFinder.setMulticastPort(10501)
    discoverySpi.setLocalPort(10502)
    discoverySpi.setLocalPortRange(0)
    commSpi.setLocalPort(10003)
    commSpi.setLocalPortRange(0)
    cfg.setTimeServerPortBase(10004)  // actually not using
    cfg.setTimeServerPortRange(0)

All work in development environment where there is no firewall. But inside
OpenStack, where only 22 and 10000 to 11000 ports are open (both TCP and
UDP) it does not work.

I use tcpdump to figure out what was happening when e new IgniteServer is
launched.
- ServerA the ignite server that is already running (IP 192.168.6.40)
- When ServerB (IP 192.168.6.41) starts, it sends an UDP datagram to
McastAddr:McastPort using a source ephemeral port
      (192.168.6.41:40415) -> (228.10.10.157:10501)  data=00 00 47 47
  I assume this is a hello everybody message.
- Then ServerA try to send a reply
      (192.168.6.40:10501) -> (192.168.6.41:40415) dataLen=584
  This datagram *never arrives* because it is sent to a port outside allowed
range (10000/11000)

Is there a way to make Ignite to bind to a localPort / localPortRange in
client TCP and UDP sockets?

Thanks



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/