11-11-2017 09:07 AM - edited 03-08-2019 12:42 PM
I've been having some performance issues with one of my Cisco switches for a while now. I originally posted on Reddit seeking some help. Now that the issue continues, I wanted to see if anyone had some ideas over here.
I'm using a Cisco 3560G to handle some basic vlans/routing for a rack of servers. The switch has a gigabit uplink and I've recently only been getting a few MBps from external tests to various local servers as well as internally between systems across vlans.
My original configuration had a lot of secondary addresses on various vlans, after discovering that this wasn't best practice - I've offloaded a lot of the addressing for each vlan to a small ER-X. Unfortunately, that hasn't fully corrected the issue.
Here's my current config: https://pastebin.com/raw/dYaZ6q22
Any thoughts/advice would be appreciated!
11-11-2017 10:55 AM
Hello,
just a few things I can think of:
You don't need the default-gateway, since you already have ip routing enabled and a default route. So remove this line:
--> no ip default-gateway 124.124.124.241
Also, check if CEF is enabled.
Switch1(config)#ip cef
The access ports are typically configured with 'spanning-tree portfast', you might want to add this:
interface GigabitEthernet0/3
switchport access vlan 542
switchport mode access
spanning-tree portfast
11-11-2017 11:18 AM
Regarding STP, beside portfast on access ports, you should configure rapid PVST on all switches in domain.
spanning-tree mode rapid-pvst
According to configuration - summarize all routes since next hop is the same address:
ip route 123.123.123.16 255.255.255.248 124.124.124.244 ip route 123.123.123.24 255.255.255.248 124.124.124.244 ip route 123.123.123.32 255.255.255.224 124.124.124.244 ip route 123.123.123.42 255.255.255.255 Null0 ip route 123.123.123.64 255.255.255.224 124.124.124.244 ip route 123.123.123.96 255.255.255.248 124.124.124.244 ip route 123.123.123.104 255.255.255.248 124.124.124.244 ip route 123.123.123.112 255.255.255.248 124.124.124.244 ip route 123.123.123.120 255.255.255.248 124.124.124.244 ip route 123.123.123.128 255.255.255.248 124.124.124.244 ip route 123.123.123.136 255.255.255.248 124.124.124.244 ip route 123.123.123.144 255.255.255.248 124.124.124.244 ip route 123.123.123.160 255.255.255.224 124.124.124.244 ip route 123.123.123.192 255.255.255.248 124.124.124.244 ip route 123.123.123.200 255.255.255.248 124.124.124.244 ip route 125.125.125.16 255.255.255.240 124.124.124.244 ip route 125.125.125.32 255.255.255.248 124.124.124.244 ip route 125.125.125.40 255.255.255.248 124.124.124.244 ip route 125.125.125.48 255.255.255.240 124.124.124.244
11-12-2017 04:12 PM
I've set rapid PVST on the switch. Hopefully this helps.
With regards to the routes - are you simply referring to combining routes? For example, taking two /29 subnets that are currently routed separately and routing them all together as a single /28?
11-12-2017 09:21 PM
Yes.
11-12-2017 04:05 PM
Thanks for the reply! I've removed the default gateway entry. 'ip cef' returns incomplete command, perhaps it's not enabled. Is this something I should consider enabling?
I still need to look into configuring each access port with spanning tree portfast.
11-13-2017 02:40 PM
Hello,
what are your options at 'ip cef' ?
ip cef ?
11-13-2017 06:57 PM
ip cef ? distributed Distributed Cisco Express Forwarding linecard CEF linecard commands load-sharing Load sharing optimize Optimizations traffic-statistics Enable collection of traffic statistics
11-11-2017 09:08 PM
Also, regarding IPv6, do you need 2 equal cost default routes?
ipv6 route ::/0 1234:F80:0:1000::3:1 ipv6 route ::/0 5678:1900::1
11-12-2017 04:02 PM
No reason for it, I've removed the extra route. Thanks!
11-12-2017 09:25 PM
What is your internet connection type (cable, ADSL)?
11-12-2017 09:59 PM
It's just a gigabit Ethernet uplink to the DC.
11-13-2017 02:12 AM - edited 11-13-2017 02:12 AM
Hello
Can also you post -
sh process cpu sort
show spanning-tree detail | in is|top|tran|Link|BPDU:
sh int | in tx|rx|drops|error
res
Paul
11-13-2017 08:51 AM
11-13-2017 10:17 AM
There is a lot of output drops on some interfaces. QoS or congestion avoidance (WTD) should be implemented. Implementing WTD would be fastest and it would prevent TCP global synchronization issue.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide