We use several Dell PowerConnect 5324 switches in our network infrastructure and for the most part, they've worked fantastically. As part of trouble-shooting a network issue earlier this year, we updated the firmware version to 18.104.22.168. Since then, we noticed an issue where our network monitoring system would become sporadically unreachable for polling the management interface on several, but not all of the upgraded switches.
The checks were done by SNMP or ICMP pings, and the management address would stop responding for periods of up to 10 minutes, several times per day. On two occasions, the management address stopped responding completely.
Packet dumps revealed that the management interfaces weren't responding to ARP who-has requests. Manually adding the MAC address of the management port of the switch to the ARP table on our monitoring server has completely cleared up the problem. The following is information from a login session on one of the effected switches (the switch name, MAC, and IP addresses are just examples).
switch-x# show system System Description: Neyland 24T System Up Time (days,hour:min:sec): 21,04:27:46 System Contact: System Name: switch-x System Location: System MAC Address: aa:00:00:00:00:01 System Object ID: 22.214.171.124.4.1.674.10895.3004 Type: PowerConnect 5324 Main Power Supply Status: OK Fan 1 Status: OK Fan 2 Status: OK switch-x# show version SW version 126.96.36.199 ( date 04-Mar-2008 time 10:55:29 ) Boot version 1.0.2.02 ( date 23-Jul-2006 time 16:45:47 ) HW version 00.00.02
To manually add an entry to the Linux ARP table
arp -s 192.168.24.1 aa:00:00:00:00:01
Our monitoring system is running a version of Red Hat, so we used something like the following script in /etc/sysconfig/network-scripts/ifup-local to add the static ARP entries at boot time.
#!/bin/sh if [ "$1" = "eth0" ] then /sbin/arp -s 192.168.24.1 aa:00:00:00:00:01 /sbin/arp -s 192.168.24.2 aa:00:00:00:00:02 fi
We've gotten no acknowledgement from Dell support that they have even received the problem report we submitted via their web-form. However, with this solid work-around, we don't expect to spend any more time on the issue.comments powered by Disqus