3750G-24TS Stack Losing Management

Unanswered Question
Nov 29th, 2007
User Badges:

We have multiple 3750G switch stacks of 5-6 switches that are getting intermittent packet loss. Our monitoring software reports them as being down once 5 pings to the management interface ellict no responses, yet the switch never goes down and no production impact is reported.

Also when we do have management to these switches, they are very sluggish. An extended ping to the management IP usually results in roughly 5% packet loss during the day even though all successful responses are 1ms.

Bandwidth utilization was the first place I looked and it does not seem to be oversubscribed (total of 6GB uplink on each stack)....But you never know I guess.

CPU utilization is usually 5% and never goes above 10%. There does not seem to be any drops on the interfaces either. No routing is done on the switches (only default gateway configured). No QOS is configured on the switches (but there is some on the core switch but only for routing protocols...).

All switches are the same and all are running version 12.2(20)SE1 and image C3750-I5K91-M.

Please help! I will provide any additional information you request! Thanks!

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Hieu Cao Thu, 11/29/2007 - 13:48
User Badges:

Have you checked port speed/duplex setting where your workstation is connected in the switch? Do a 'show log'..anything obvious that you can tell in the log that can contribute to this problem? From the switch pinging back to your workstation, do you get the same response? how about ping to and from another workstation, do you experience the same problem?


takethesewords Fri, 11/30/2007 - 05:21
User Badges:

my workstation is not connected to the switch, not even in the same data center. servers are connected on the back side (lots of them) so i thought it was possibly being oversubscribed but bandwidth usage never goes above 70% (and that is during peak times).

there is nothing in the "show log" that indicates any sort of problem, on any of the 'problem' stacks. we have some stacks that do not have any issues but they are WS-3750G-24TS-1U and running IOS 12.2(25)SEB2 or 12.2(25)SEB4 (no mismatches on the stack).

i just now did a ping from the switch to my workstation, and had a 70% success rate. at the same time i was doing a ping from my WS to the switch management interface and about 80% were successful. pings to/from other workstations return the same results.

the weird thing is, pings to servers directly connected to the stack are all 100% successful....im trying to find bugs related to the 12.2(20)SE1 IOS version but the cisco.com bug toolkit is kind of flaky at the moment. thanks for your help!

cerp@verizon Wed, 12/05/2007 - 05:19
User Badges:

We are experiencing the same issue with a set of stacked 3750's. We have three 3750-48PS-S running ipbasek9-mz.122025.SEE stacked and delpoyed as an access-layer switch. Switch is linked back to the core via dual gigabit links. The switch is configured with a management vlan, data vlan ,printer vlan, and voice vlan. All users, phones, and printers hanging off the switch are working fine. CPU utilization runs at around 10%. Pings to the management vlan drop about 40% of the packets, telent/ssh to the switch is a painfull affair if you can get in all. Any thoughts would be helpfull.

Hieu Cao Wed, 12/05/2007 - 09:50
User Badges:

I am using C3750-24PS-S stack with c3750-advipservicesk9-mz.122-25.SEE4 image and I've no issues with it. Is upgrading the IOS an option that you can try out?


takethesewords Wed, 12/05/2007 - 09:58
User Badges:

i thought that an IOS upgrade would do the trick but after exploring that idea, its really not an option. too much production traffic to take a down a stack to upgrade the IOS. the stacks have been up for over a year and a half but the issues we are experiencing did not start until about 4 or 5 months ago...

cerp@verizon Mon, 12/10/2007 - 08:55
User Badges:

I've obtained downtime and we plan on upgrading to the Advanced IP services image. We plan on pushing routing to the edge...no more management vlan. Hopefully this will take care of the problem.


This Discussion