I have this problem at one of my sites, where users report slow performance when connected to one of the two core switches whilst accessing a file server connected to the other core switch.
The file servers are in a cluster, with dual NICs (one connected to each switch). I have confirmed this slow performance by performing FTP tests from a laptop connected to each switch to each of the file servers. The performance from the "slow" switch is 3 - 4 times slower than from the other switch.
Does anybody have any ideas on what could be causing this? Have attached a diagram of the network, and have also listed below all the things that I have ruled out.
1. To eliminate the possibility of it being an MSFC issue, we changed the laptop (A and B) to VLAN 2 (the same VLAN as the server). The active HSRP gateway and STP root bridge is switch/msfc 11.
2. Configured Laptop B to VLAN 120 (same VLAN as working devices in Westline)
3. To eliminate the possibility of it being a module issue by patching the laptop into various modules within switch 12 (modules 5, 7 and 9) ? all modules experience the same issues.
4. The utilization of the Etherchannel between switches 11 and 12 is at around 30% (of a 6gbps link).
5. We ruled out the possibility of it being a speed/duplex issue on either the switch or the laptop by performing the FTP of the same file whilst using each combination of speed duplex settings.
6. To rule out an issue on the supervisor module on switch 11 (at this point we still had one of the issues appearing on both switches), we failed over the supervisor modules on switch 11.
7. To rule out an issue on the supervisor module on switch 12, we failed over the supervisor modules on switch 12.
8. To rule out the possibility of it being the NIC connecting the primary server to switch 12, we shut down this link, forcing the traffic from Laptop B to traverse the inter-switch links on its way to the server. Therefore, using the same connection as traffic from switch 11.
9. The server team removed the master server from the cluster so that the secondary server became the master.
10. Checked backplane utilization on both switches, switch 12 runs at approximately 0-5% and its maximum is 15%. Both switches modules are fabric enabled, yet the Supervisors do not support fabric switching. Therefore, all Layer 2 decisions are made by the Supervisor (even if source and destination are on the same module).
11. All CPU?s and memory checks are showing that no over utilization is present on either switch.
Many thanks in advance,