Could someone help me build a PBS complex on VMs that have multiple network interfaces?
I built several virtual machines running CentOS 7.4 with Vagrant and VirtualBox so that they can communicate each other via VirtualBox’s “internal” network (192.168.0.*).
In addition, VirtualBox automatically creates a mandatory “host-only” network (10.0.2.15) on each VM as eth0, which connects the host (physical) machine and the VM.
The host-only network cannot be used for VM-to-VM communication.
I thought PBS communicates via the internal network (192.168.0.*).
[node00] eth0: 10.0.2.15 = host-only network eth1: ***.***.2.90 = bridged network (global IP not shown) eth2: 192.168.0.100 = internal network [node01] eth0: 10.0.2.15 = host-only network eth1: 192.168.0.101 = internal network
I installed PBS Pro 17.1.0 on these two VMs and started it.
node00:~$ sudo service pbs status pbs_server is pid 23916 pbs_mom is pid 23598 pbs_sched is pid 23613 pbs_comm is 23568 node01:~$ sudo service pbs status pbs_mom is pid 20422
Then I’ve got the following logs in server_logs.
01/15/2018 16:23:06;0001;Server@node00;Svr;Server@node00;is_request, bad attempt to connect from 10.0.2.15:15003
I guess the MoM on node01 reports her IP as 10.0.2.15, which confuses the server on node00 since the same IP exists on node00.
I’ve read ‘PBS with Multihomed Systems’ (IG-62) and ‘Name Resolution and Network Configuration’ (IG-8), but they are kind of too abstract for me. X-(
For now I’ve confirmed that ‘hostname -i’ returns the internal network address (192.168.0.*).
node01:~$ hostname -i 192.168.0.101
What can I do in this situation?
Any comments or suggestions would be appreciated. Thank you,