Re: 2950 Poor Performance and Ping Times

jhgoldberg · ‎06-21-2004

I have a new 2950 that I just installed. Each port is configured to be on a different VLAN. Port 24 is set as a trunk. It trunks up to a Netscreen 204 Firewall which then trunks up to a 6509. I run in this configuration all over my network except with 3524 switches and it works fine.

When the 2950 first boots the management IP and the servers connected to it all ping and connect fine. About 2-5 minutes later the pings of the management IP go to over 1000ms and then a minute or two later it fails to respond to ping. The performance of the servers connected to the switch gets very slow. This problem even occurs with no devices connected to the switch except the trunk port.

The switch is running the latest version of EA IOS for the 2950.

Any help you can provide would be appreciated.

rfroom · ‎06-22-2004

Foremost, pinging the management IP of the 2950 is not a good way to measure performance. The respond time is strictly of function on the internal CPU and not relevant to traffic flow. In addition, ping is very low on the CPU priority list.

Nevertheless, when other traffic is slow, check the spanning-tree topology, port counters, and mac address table. Do you notice anything anomolous such as execessive spanning-tree topology changes or excessive broadcasts?

jhgoldberg · ‎06-22-2004

I understand that ping is not a good test but 1000ms ping times for a switch on the local LAN with no stations on it and only an up link port is a little unreasonable. The switch also is unusably slow in telnet once it slows down. The CPU is hovering at 4%.

I see nothing out of the ordinary in the arp cache and spanning tree looks normal. I have even tried disabling spanning tree (as this is point to point with no chance of a loop). The problem still exists.

MUHAMMAD SHAHEEN · ‎06-22-2004

Try setting the port speed to 10 mbps or 100 mbps instead od auto and change the duplex to half. I have seen and fixed this problem many times. Ensure the device connecting to each port has same settings. Some IOS have a bug and does not negotiate this correctly.

I hope this will help to resolve this issue. Please do let us know the outcome.

jhgoldberg · ‎06-22-2004

I have the port on the switch and the port on the firewall set for hard coded for 100Mbps, Full Duplex as I do not trust autonegotiate. Are you saying that I need to run this at half in order to get it to work? I am not sure that running at half duplex is a workable solution for me long term but I will give it a shot. Thanks for the advice. I will post on here tomorrow morning (Eastern time) whether it worked.

jhgoldberg · ‎06-23-2004

I tried changing the hard coded full duplex settings to 100/half on both the switch and the firewall and the same issue occured.

I even tried 10/half and 10/full but still after about 2-5 minutes the switch will slow to a crawl and become unusable (even with no station connected to it except the uplin).

Thanks

glen.grant · ‎06-23-2004

Sounds like it is time to get a sniffer on the link . What does the link say for usuage when you do a show int when it is slow ? is something using all the bandwidth ? You can try doing some debugs , it shouldn't be a problem if the cpu is low , if it's high then I wouldn't go that route as this could burry the switch even faster . Nothing showing up in the logs during the affected period ?