Unsolved
This post is more than 5 years old
1 Rookie
•
110 Posts
0
359014
January 14th, 2013 08:00
PowerConnect 6248 RSTP
Hi I have some quick questions regarding RSTP, Iv'e looked around in the forum and the suggestions for my question differ so I would appreciate if someone could provide the correct best practice answers..
I have a Dell PC6248 that I use as a core switch. It has STP priority 0. Recently I started noticing some performance decrease on our network. I get several packet drops when I ping the VLAN interface I'm sitting on. And I get also really poor performance when I'm logged on to the switch via telnet, I even get kicked out as the connection drops. It's really frustrating.
When I issue a #show logging, I get a bunch of this....
<189> JAN 14 16:25:15 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1576 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:28:21 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1577 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:29:31 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1578 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:29:35 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1579 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:29:51 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1580 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:32:57 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1581 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:34:55 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1582 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:38:01 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1583 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:39:14 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1584 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:40:09 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1585 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:43:17 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1586 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:45:09 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1587 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:48:15 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1588 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:48:31 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1589 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:48:35 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1590 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:49:55 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1591 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:53:01 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1592 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:53:21 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1593 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:53:25 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1594 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:55:03 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1595 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:58:11 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1596 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 16:59:25 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1597 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:02:33 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1598 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:03:37 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1599 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:04:14 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1600 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:06:43 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1601 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:08:33 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1602 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:11:39 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1603 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:16:15 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1604 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:18:35 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1605 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:19:23 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1606 %% Spanning Tree Topology Change: 0, Unit: 1
<189> JAN 14 17:23:45 192.168.150.1-1 TRAPMGR[125687056]: traputil.c(611) 1607 %% Spanning Tree Topology Change: 0, Unit: 1
The IP address is one of the VLAN interfaces. Do I have a loop? Where do I start to find it? Do I need to...
Set all ports connected to end devices as portfast?
Set all trunk ports to non-portfast?
Set all LACP channels as portfast? What about if they are trunks?
0 events found
DELL-Willy M
802 Posts
1
January 14th, 2013 10:00
All the systems that you have provided are pointing to a loop. I would start by reading thru this great post from our main page posted as a sticky. It gives a great overview of steps to trouble shoot spanning tree.
en.community.dell.com/.../19465205.aspx
To start I would make sure you have the same Version of Spanning tree on all devices in your network. Some switches default to STP and some default to RSTP.
Portfast is a feature that one typically wants to enable for a port connected to a host. When the link comes up on this port, the first stages of the STP are skipped and the port directly transitions to the forwarding mode. This can obviously be dangerous when not used correctly. You will not want port fast set on a trunk. This is where you would be connecting from switch to switch or router and no hosts are connected. The same for LACP Channels.
Set all ports connected to end devices as portfast? Possibly
Set all trunk ports to non-portfast? No
Set all LACP channels as portfast? What about if they are trunks? No
Hope this helps clear up some things
FamousRuler
1 Rookie
•
110 Posts
0
January 15th, 2013 01:00
ok what about the port where I have an WiFi AP with RADIUS? Or whre I have a small 4-port dektop switch? Should I configure those ports with portfast?
And another thing, all my switches are DELL. PC62xx, PC55xx and PC54xx. All switches are configured with RSTP but I noticed that the PC55xx are configured with RSTP Long and the PC54xx with RSTP short. Does this matter?
DELL-Willy M
802 Posts
1
January 15th, 2013 09:00
You will only want Portfast set on physical ports that are connecting to an end device: Laptop, Desktop, sometimes servers depending on your environment.
Here is what I pulled from the 5548 CLI User Guide on RSTP long vs short:
spanning-tree pathcost method
Use the spanning-tree pathcost method Global Configuration mode
command to set the default path cost method. Use the no form of this
command to return to the default configuration.
Syntax
spanning-tree pathcost method {long | short}
no spanning-tree pathcost method
Parameters
• long—Specifies that the default port path costs are within the range:
1–200,000,000.
• short—Specifies that the default port path costs are within the range:
1–65,535.
Default Configuration: Short path cost method.
Command Mode: Global Configuration mode
User Guidelines
This command applies to all the spanning tree instances on the switch.
• If the short method is chosen, the switch use for the default cost values in
the range 1 through 65,535.
• If the long method is chosen, the switch use for the default cost values in
the range 1 through 200,000,000.
Example
The following example sets the default path cost method to Long.
Console(config)# spanning-tree pathcost method long
I would recommend making the path cost the same on the switches. Since Short is the default you may want to choose that setting unless you have a reason to choose Long.
FamousRuler
1 Rookie
•
110 Posts
0
January 17th, 2013 04:00
DELL-Willy M
802 Posts
1
January 17th, 2013 10:00
Can you provide further information by running these commands and showing the output
console#show spanning-tree blockedports
console#show spanning-tree active
console#show spanning-tree summary
Then we can have a view of what ports are forwarding or blocking. And compare that to your physical layout to see what we are dealing with more specifically.
thanks
FamousRuler
1 Rookie
•
110 Posts
0
January 17th, 2013 14:00
Offcourse, here it goes..
DELL-Willy M
802 Posts
1
January 17th, 2013 15:00
Ok, this is what would normally see on the Root switch where all ports are Designated. How many switches are connected directly to this 6248 core switch? We will want to do a similar process on the connecting switches and map out what ports are blocking/forwarding on a diagram to get a good picture of where traffic is allowed to pass.
We can also look at running this command for all the interfaces and port channels listed in the active output. This can show if there are any abnormally large amounts of traffic on a specific port and any discarded frames.
console #show statistics ethernet 1/g25
FamousRuler
1 Rookie
•
110 Posts
0
January 18th, 2013 09:00
DELL-Willy M
802 Posts
0
January 18th, 2013 15:00
ethernet 1/xg4 is showing
802.3x Pause Frames Received................... 547
Unacceptable Frame Type........................ 42
The rest of the ports are showing 0 for the same.
I do not see any other discards listed from the output that you provided.
ch1 is a LAG of 8ports, 1/g1-1/g8 connecting a PC5424
ch2 is a LAG of 8ports 1/g9-1/g16 connecting a PC5424
1/g45 is a 1Gb fiber connecting a PC5548
1/g46 is a 1Gb fiber connecting a PC5524
1/g47 is a 1Gb fiber connecting a PC5524
1/g48 is a 1Gb fiber connecting a PC5548
1/xg4 is a 10Gb fiber connecting a PC6224
Are all the switches listed above separate switches? I'm trying to understand if there are possibilities for redundant paths or links.
You probably want to look at the PC6224 that is connecting to 1/xg4 with the same show commands and see if anything abnormal sticks out.
FamousRuler
1 Rookie
•
110 Posts
0
January 18th, 2013 16:00
yes they are all different switches, You are correct. The network design is not redundant. No switch is crossconnected. I will check the uplink ports on the PC6224.
What does "Unacceptable Frame Type" mean?
FamousRuler
1 Rookie
•
110 Posts
0
January 30th, 2013 06:00
I think I have isolated the problem originating on two switches, the core 6248 and another 6248 connected to it. Both switches are connected on port 1/xg4
The core 6248 still shows lots of topology changes (I paste only a snippet)
FamousRuler
1 Rookie
•
110 Posts
0
January 30th, 2013 09:00
I got this when showing logging on the core switch at the time the lagging/packet drops occurred..
<190> JAN 30 17:42:10 192.168.2.2-1 UNKN[104098896]: ewtask.c(5397) 446 %% EMWEB TransmitPending : EWOULDBLOCK error sending data
<190> JAN 30 17:57:45 192.168.2.2-1 UNKN[104098896]: ewtask.c(5397) 447 %% EMWEB TransmitPending : EWOULDBLOCK error sending data
<189> JAN 30 17:58:34 192.168.2.2-1 TRAPMGR[104098896]: traputil.c(611) 448 %% Multiple Users: Unit: 0 Slot: 5 Port: 1
DELL-Willy M
802 Posts
1
January 30th, 2013 10:00
I'm looking into what information we can gather from your latest post. It is good to see that you have it isolated to a single connection between the two 6248 switches.
Have you tried using a know good cable between the two switches to help rule out any possible hardware as the problem?
FamousRuler
1 Rookie
•
110 Posts
0
January 31st, 2013 01:00
I don't have other SFP+ or fiber cables to test with.
Should I remove portfast on all ports just to make sure that if a user have plugged in some switch and thus creating a loop?
DELL-Willy M
802 Posts
1
January 31st, 2013 16:00
An “unacceptable frame type” is simply a frame that is somehow malformed and therefore rejected by the switch. In this case, considering the very large number of frames transmitted and received on the port in question, 1 unacceptable frame is inconsequential. The pause frames may be worth looking at if they constitute more than 0.1% of all frames Tx or Rx on that interface. It is possible that one NIC in a server/workstation can go bad and generate enough “noise” that flowcontrol is activated which can be identified by the pause frames found on the interface directly connected to that NIC.
It will not hurt to run cable diagnostics on that connection to rule out hardware. Page 152 of the User Guide found below starts the discussion about the built in cable diagnostics on the switch.
support.dell.com/.../gsg_62.zip
We could also look at changing the root priority of the switch to spanning-tree priority 4096 Using 0 in a mixed vendor environment can indicate the switch is not participating in the STP root election process. Also keep in mind that once a flat network reaches a certain number of nodes it can experience more STP topology changes than one that is segmented with multiple vlans.
You can cut devices that are affected by a broadcast by segmenting your environment with VLANs thus breaking your overall network into smaller broadcast domains.