The network topology is shown in the appendix. The packet loss ratio approximates 1.5% when the gateway of the NE80E are pinged from the NC server.
1. Because the NE80E cannot collect information about the ping echo packets that it responds with, the traffic statistics on the NE80E cannot lead to accurate judgment on whether the NE80E correctly responds to the ping request packets. However, according to the statistics, the NE80E receives the ping request packets from the NC server.
2. Attach a PC directly to the NE80E on the site. Ping the IP address of the NE80E from the PC. It is found that packets are still lost. Capture the packets. It is found that the NE80E can receive all ping request packets from the PC, but fails to respond to some other packets. Therefore, it is deduced that packets are lost on the NE80E.
3. Analyze the captured packets (for details, see the appendix). It is found that the NE80E sends an ARP request packet every time it discards a packet. Therefore, packet loss on the NE80E must be caused by the lack of ARP entries.
4. Check ARP entries of the NE80E. It is found that ARP entries are refreshed from time to time. The aging time is 20 minutes. That is, the NE80E refreshes ARP entries at least once a minute.
5. The check of the log buffer finds no exception.
6. The check of hardware logs finds a large number of BPDU packets sent from port 9/0/23 (about once every 30s).
Feb 25 2009 01:23:37 DongZhiMen_NE80E_1 %%01MSTP/6/RECEIVE_CISTTC(l): MSTP received BPDU with TC, instance 0, port name is GigabitEthernet9/0/23.
According to the preceding information, the NE80E frequently refreshes ARP entries because it receives BPDU packets from time to time.
7. Check devices connecting to port 9/0/23 according to this log. It is found that a certain port of the S8016 attached to the NE80E goes Up and then Down from time to time, which generates a large number of BPDU packets.
8. Close this port of the S8016 that goes Up and then Down from time to time. No packets are found lost on the NE80E in the ping test on the NC server. That is, the packet loss ratio is 0.
1. Identify where packets are lost through traffic statistics and packet capture.
2. If packets are not lost on the NE80E, but on the competitor's device, inform the customer of the fact and ask the customer to find out the cause.
3. If packets are lost on the NE80E, find out the cause according to information about the CPU usage and ARP entries.
4. If the cause cannot be identified according to basic information, collect hardware logs and diagnosis information for further analysis.
1. The log level of the NE80E for receiving BPDU packets is quite low. By default, BPDU packets are not recorded in the log buffer.
2. It is strongly recommended that the port of the switch connecting to the server be set to STP edge port and BPDU protection is enabled globally on the switch, thus avoiding unnecessary BPDU packets.
3. The NE80E handles ARP entries as follows when receiving BPDU packets:
a. Fast: Delete all related ARP entries immediately from the ports included in the STP instance, and then relearn ARP entries based on the traffic.
b. Normal: Zero the aging time of ARP entries, and then refreshes ARP entries according to the normal ARP aging mechanism.
In fast mode, the STP refreshing and convergence time is short, but it leads to frequent network oscillation. In normal mode, the STP refreshing and convergence time is comparatively long, but it is quite stable. You can choose either according to network conditions. By default, the NE80E adopts the fast mode, which can be configured using the stp converge command.