Network Troubleshooting - Hitachi Hyper Scale-Out Maintaining And Troubleshooting

Network troubleshooting

Problem/Symptom
Internal cluster IP addresses
conflict with IP addresses on
external systems
Problems with network connections
to the cluster and/or the need to
determine which switch the nodes
are connected to and which nodes
are in which rack
Nodes are not joining the cluster
Cluster or nodes are not getting an
IP address
Nodes are not cabled properly
26
Hyper Scale-Out Platform Supporting and Troubleshooting
Troubleshooting action to take
The HSP internal network uses a supernetwork (255.255.0.0) of
the 192.168.x.x and 192.169.x.x ranges. This internal network
cannot be shared with any other resource and no external
systems or routers in this IP address range can access an HSP
cluster.
If you have conflicts with this reserved range in your network,
contact Customer Support for assistance.
The script cluster_macs reports the MAC address of eth0 for all
nodes in the cluster. This information can be used by network
administrators to determine the switch and port a node is plugged
into.
For example:
admin@Node003:~$ cluster_macs -u admin -p admin
hostname,eth0 mac,eth2 mac,eth3 mac,switch eth2,switch eth3
Node001,f4:52:14:80:7c:30,68:05:ca:32:10:e0,68:05:ca:32:10:e1,cc:4e:24:38:a8:74,cc:4e:24:38:a8:80
Node002,f4:52:14:80:ca:40,68:05:ca:32:10:d8,68:05:ca:32:10:d9,cc:4e:24:38:a8:74,cc:4e:24:38:a8:80
Node003,f4:52:14:80:8b:b0,68:05:ca:32:10:98,68:05:ca:32:10:99,cc:4e:24:38:a8:74,cc:4e:24:38:a8:80
You can also use this script to determine if you plugged eth2 and
eth3 into the wrong switches. The last two columns report the
chassis ID of the brocade switches, which is the MAC address of
it's management port. (Note: The backend switches must be
correctly configured to enable lldp for this script to work.)
In this example, the eth2 and eth3 cables swapped in Node002:
admin@Node003:~$ cluster_macs -u admin -p admin
hostname,eth0 mac,eth2 mac,eth3 mac,switch eth2,switch eth3
Node001,f4:52:14:80:7c:30,68:05:ca:32:10:e0,68:05:ca:32:10:e1,cc:4e:24:38:a8:74,cc:4e:24:38:a8:80
Node002,f4:52:14:80:ca:40,68:05:ca:32:10:d8,68:05:ca:32:10:d9,cc:4e:24:38:a8:80,cc:4e:24:38:a8:74
Node003,f4:52:14:80:8b:b0,68:05:ca:32:10:98,68:05:ca:32:10:99,cc:4e:24:38:a8:74,cc:4e:24:38:a8:80
Nodes in different racks will be attached to different backend
switches if you have multiple racks,
The node_check --nic script does a quick check for link and runs an
internal self test of the NIC.
For example:
admin@Node003:~$ node_check --nic
Testing eth0
Testing eth2
Testing eth3
No Ethernet errors detected
Troubleshooting