Clarification on STP, portfast and intermittent connectivity

[ktan]

From what I've read, portfast is usually used in association with Spanning Tree Protocol (STP) and solves issues during the startup process of a leaf-node (i.e., non-switch, non-routing device such as a workstation or printer), right? As such, portfast must be used in conjunction with STP and is unlikely to be the issue for a workstation that is experiencing intermittent but outright drops in connectivity to a server, right?

[Danilo Dy]

You're right. PortFast is recommended in the switch port that is directly connected to a host. It is a no-no to a switchport that is connected to another switch, hub, or router.

However, intermittent in an STP could be cause by an STP loop.

Here are some links on how to troubleshoot or to looup up if you think you have an STP problem

http://www.cisco.com/warp/public/473/16.html

http://tcpmag.com/archives/article.asp?EditorialsID=20

[ktan]

Thanks for the links. I'll check them out.

At the moment, I haven't quite figured out if it's an STP problem or a speed/duplex problem. There is only a host workstation (a Macintosh) connected to that port and it was discovered that the port on the Cisco was set to 1Gbps/full-duplex instead of to autonegotiate while the Macintosh was set to autonegotiate. The Macintosh has now been set to 1Gbps/full-duplex to match the Cisco port in the hopes that the intermittent dropouts in connectivity it has been experiencing will go away. The network administrator has assured me that since it is a Macintosh, STP is disabled on that port (and, I assume, portfast is also disabled since it's connected to STP). I am thinking of suggesting that he re-enable STP but ensure that portfast is also set for those ports which have Macintoshes connected to them.

[avillalva]

Hi,

You will find portfast is a mechanism to stop a port using STP rules. Since STP takes 50 seconds to go into forwarding mode some PCs have problems aquiring IP addresses because at the time they request an IP the port is not yet forwarding.

So when he says STP is disabled on that port, he means he has enabled portfast.

As far as your network connections, ensure that both sides (i.e. the MAC and the switch port) are set the same.

If you still experience problems, have him check the port stats, he will be able to see any errors at Layer2.

HTH,

Andres

[ktan]

Okay, in other words, portfast is only of concern during the bootup phase. However, we're not experiencing issues during bootup but intermittently while connected to a server.

From viewing the logs on the server, even though the server has been hardcoded to 1Gbps/full-duplex to match the hardcoded port on the Cisco Catalyst (i.e., both are NOT in autonegotiation mode), I've identified a log entry on the server which indicates that it was a physical link layer error:

Apr 3 11:55:57 xserve01 kernel [0]: Apple BCM5701Ethernet: 0 4 setupCopperPhy - link is down

and this is what the network admin got from the switch:

MAC4006> (enable) sh port 4/12

* = Configured MAC Address

Port Name Status Vlan Level Duplex Speed Type

----- ------------------ ---------- ---------- ------ ------ ----- ------------

4/12 connected 1 normal full 1000 1000BaseT

Port Security Violation Shutdown-Time Age-Time Max-Addr Trap IfIndex

----- -------- --------- ------------- -------- -------- -------- -------

4/12 disabled shutdown 0 0 1 disabled 84

Port Num-Addr Secure-Src-Addr Age-Left Last-Src-Addr Shutdown/Time-Left

----- -------- ----------------- -------- ----------------- ------------------

4/12 0 - - - - -

Port Flooding on Address Limit

----- -------------------------

4/12 Enabled

Port Send FlowControl Receive FlowControl RxPause TxPause Unsupported

admin oper admin oper opcodes

----- -------- -------- --------- --------- ---------- ---------- -----------

4/12 on disagree desired off 0 0 0

Port Status Channel Admin Ch

Mode Group Id

----- ---------- -------------------- ----- -----

4/12 connected auto silent 125 0

Port Status ErrDisable Reason Port ErrDisableTimeout Action on Timeout

---- ---------- ------------------- ---------------------- -----------------

4/12 connected - Enable No Change

Port Align-Err FCS-Err Xmit-Err Rcv-Err UnderSize

----- ---------- ---------- ---------- ---------- ---------

4/12 - 0 0 0 0

Port Single-Col Multi-Coll Late-Coll Excess-Col Carri-Sen Runts Giants

----- ---------- ---------- ---------- ---------- --------- --------- ---------

4/12 0 0 0 0 0 0 0

Last-Time-Cleared

--------------------------

Tue Apr 3 2007, 10:37:46

Idle Detection

--------------

If anyone could assist in interpreting the above output from the Catalyst 4006, I'd appreciate it.

[sundar.palaniappan]

You might want to try a couple of things;

1. Set the speed/duplex to auto on both switch and the PC NIC.

2. If you still have problems replace the cable.

HTH

Sundar

[surojitmitra]

I have faced similar problem in SUN RSC cards. The problem was resolved by removing port fast from that particular switch port. I was using cisco 4503 switches in STP

[ktan]

I'm curious from the theoretical perspective as to what setting the speed/duplex to auto instead of hard-coding it to "1Gbps/full-duplex, flow control" would do (the settings of both ends matching up, of course)? Would a consistently autonegotiated speed below 1Gbps then indicate hardware issues in the cable (e.g., interference, etc)?

Also, from the above port status command, could anyone point me to where STP status and portfast status would be indicated?

[ktan]

In the following snippet of a return from issuing the "sh spantree" command, it appears that STP is disabled. If STP is disabled, does it make sense to enable PortFast? What happens when both STP and PortFast are disabled? Could this explain the intermittent dropouts/connectivity losses during a session (this is *NOT* during the boot phase of a leaf node device but afterwards, in the midsts of a connection session itself)?

MAC4006> (enable) sh spantree

VLAN 1

Spanning tree disabled

Bridge ID MAC ADDR 00-05-31-64-e8-00

Bridge ID Priority 32768

Bridge Max Age 20 sec Hello Time 2 sec Forward Delay 15 sec

Port Vlan Port-State Cost Prio Portfast Channel_id

------------------------ ---- ------------- --------- ---- -------- ----------

1/1 1 forwarding 20 32 disabled 0

1/2 1 not-connected 4 32 disabled 0

2/1 1 forwarding 15 32 enabled 0

2/2 1 forwarding 15 32 enabled 0

2/3 1 forwarding 15 32 enabled 0

2/4 1 forwarding 15 32 enabled 0

2/5 1 forwarding 15 32 enabled 0

2/6 1 forwarding 15 32 enabled 0

2/7 1 forwarding 15 32 enabled 0

[avillalva]

Hi,

Enabling or disabling portfast with Spanning tree disabled would make no difference. you should note that spanning tree protects the network and disabling it is not recommended.

In your instance, the only problem that you could encounter is a layer 2 loop which would take out VLAN1 and most likely all other vlans with it due to CPU utilisation. But an intermittent fault would not be a spanning tree thing (excluding the possibility that you are adding and removing switches and there is convergence happenning assuming it, STP, was switched on)

Dont forget that intermittent faults can be, and often are, faults at higher layers. For example IP address conflicts with a device that is not necessarily switched on all day.

If you have discarded one layer as the problem , move onto the next. look at arp tables and mac address tables to verify the ip is going to the correct mac. do that when connectivity for your device is down.

HTH,

Andres