0

I compiled trafficserver-4.1.2 on two openvz containers running on Debian Squeeze, located on two different physical root nodes. Everything including caching is working fine, except for the clustering. I added the same name to the two nodes, as

traffic_line -s proxy.config.proxy_name -v fetest

Configured to run on port 80

traffic_line -s proxy.config.http.server_ports -v 80

Followed the doc on http://trafficserver.readthedocs.org/en/4.0.x/admin/cluster-howto.en.html

traffic_line -r proxy.local.cluster.type
1

and 

traffic_line -r proxy.config.cluster.ethernet_interface
venet0:0

But

traffic_line -r proxy.process.cluster.nodes
1

The second node is not showing up. Only a single node is being showed. Whether the other node is up or not, this is all it shows up, and there are no entries added up in cluster.config.

f_puras
  • 2,521
  • 4
  • 33
  • 38
nohup
  • 3,105
  • 3
  • 27
  • 52

2 Answers2

1

This looks like a compatibility issue with openVZ containers network interface. I switched this to physical nodes, and this started to identify and share information.

# traffic_line -r proxy.process.cluster.nodes
2
nohup
  • 3,105
  • 3
  • 27
  • 52
1

This looks more like a multicasting issue. Make sure your root nodes/mother ships are on the same switch/vlan. A tcp dump on the interfaces can show you whether the cluster is able to communicate each other using multicast.

file2cable
  • 564
  • 6
  • 15