[TS-1537] Traffic Server Cluster is sending multicast to the wrong interface even when being set on proxy.config.cluster.ethernet_interface

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.3.0
• Fix Version/s: Docs
• Component/s: Clustering, Documentation

Environment

CentOS 6.3 64bit

• Assignee: Jon Sime
• Reporter: Edwin Lim
• Created: 17/Oct/12 11:46
• Updated: 15/Aug/16 20:28

Description

With the following network setup of 8 servers

server 1-3
eth1: 10.0.0.x (WAN ISP A) Intel NIC
eth2: 192.168.10.x (LAN) Onboard BNX

server 4-8
eth1: 172.16.0.x (WAN ISP B) Intel NIC
eth2: 192.168.10.x (LAN) Onboard BNX

LAN has no gateway defined and sharing the same switch.
WAN is split by Vlan on a same switch for 2 different ISP

traffic_line -s proxy.config.proxy_name -v cluster01
traffic_line -s proxy.local.cluster.type -v 1
traffic_line -s proxy.config.cluster.ethernet_interface -v eth2

even with proxy.config.cluster.ethernet_interface set to eth2, multicast seems to be sending via eth1.

doing traffic_line -r proxy.process.cluster.nodes shows

3 on ISP A
5 on ISP B

tcpdump -i eth1 -n | grep 224.0.1.37 yields the following

19:38:40.167123 IP 172.16.0.162 > 224.0.1.37: udp
19:38:40.167128 IP 172.16.0.162.47310 > 224.0.1.37.8089: UDP, length 77
19:38:40.233291 IP 172.16.0.161.60059 > 224.0.1.37.8089: UDP, length 438
19:38:40.233311 IP 172.16.0.161.60059 > 224.0.1.37.8089: UDP, length 79
19:38:40.233413 IP 172.16.0.161.60059 > 224.0.1.37.8089: UDP, length 1958

Listening on eth2 do not yield any result.

update: After finding the nodes on the different nic, all data are transmitted normally through eth2. Seems only the multicast is sending via the wrong nic

Rgrds
Edwin Lim

Issue Links

• blocks: TS-4026 Remove clustering feature in 7.0.0 release