Switches Are Disconnected from the NMS Due to an Inconsistent Numbers of Eth-Trunk Member Interfaces on an S3700

Publication Date:  2013-12-31 Views:  272 Downloads:  0
Issue Description
Two S5700SIs are deployed at the aggregation layer, which are configured in a stack. S3728EIs are deployed at the access layer. The S5700SI and S3728EI versions are both V100R005C01SPC100, and the patch version is SPH016. An access switch connects to the two aggregation switches using two physical links that are bundled into an Eth-Trunk.
When a fault occurs, the S3700 access switches are disconnected from the NMS, and the access switches fail to ping the gateway (S5700).
Services on the access switches are normal. Users connected to the access switches can obtain IP addresses, and traffic on access interfaces is normal.
The gateway can ping the S3700 successfully. After the S3700s are recovered, the NMS can manage the S3700s. However, the S3700s are disconnected from the NMS after ARP entries age out.

Alarm Information
No alarm is generated on the access switches.
MAC address flapping occurs on the downlink interfaces of the access switches.
Run the display arp command. The command output shows incomplete ARP learning and incorrect interface information.

display arp
------------------------------------------------------------------------------      0819-a608-7030         I -  Vlanif1      Incomplete      1      D-0  Eth0/0/14
Total:2         Dynamic:1       Static:0    Interface:1 
Handling Process
On the S3700 with an incorrect Eth-Trunk configuration, add two uplinks to Eth-Trunk 1. Other switches can properly ARP entries.
Root Cause
Check the switches. It is found that configurations on the access and aggregation switches are correct.

Three of the 15 access switches are faulty, so check configurations on the faulty access switches one by one. It is found that only one member interface is added to the Eth-Trunk on an S3700. As a result, a loop occurs.
The traffic volume on each interface is lower than 1%, which is normal.
Be cautious about stack configurations, and ensure that links are correctly configured to prevent loops.
During stack deployment, prepare labels and connect network devices to the master and backup devices in a stack according to the labels, preventing misoperations.
If permitted, connect the master and backup devices using network cables of different colors to facilitate maintenance and fault location.