Some VLANs not working

[kalmanTRF]

This is a mind boggler to me and after looking at it for a while I am at the point where I am not advancing in any cure or it. Maybe someone knows exactly the problem and I am just spacing it out for some reason.

Here is the problem. We have many vlans and we need to go through a readdressing scheme, so rather than switch everything over at once, I created new vlans with the new addressing scheme so that we can migrate cleanly. All appears to be working correctly until several weeks ago when we started noticing certain Ports on our 3548 switches not working. moving the port into a different vlan worked. Now it is not just one Vlan that is affected and not all the time. I.E. VLAN X may work at one point, but not work later, VLANs Y and Z may both be working and then VLAN Y will not work and VLAN X will now work or VLAN X and Y will not work.

When putting a protocol analyzer on the port where the VLAN is no longer working, All I see is traffic I generate from that port, no other broadcasts or other traffic. If I switch the port to a different VLAN, bam! I see a lot of traffic and everything works great. I switch it back and I see nothing. Tomorrow, all of the VLANs could be working perfectly, but another switch may have this problem or 2 switches. There appears to be no rhyme or reason. Our core is a pair of 6509 with sup720, Trunking is done for all closets for all vlans.

Vlans are active on all of the switches, vtp status looks correct.

What am I missing here? Obviously I must have missed a setting somewhere, or took something for granted, but can't pinpoint it.

Could there be too many VLANs? Too much Trunking going on??

Let me know if you need more info, just don't want to flood this post with too much info and scare help off.

Help....

Thanks,

Ken

 

[vipergg]

The 6509's wouldn't have a restriction them . How many vlans are you talking about ? I know some the older switches like the 2924 and maybe the 3548's were limited to 64 vlans so if you are going over this limit this may be a problem .I'm really not sure on the 3548's but the 2924's had this restriction. You may have to manually prune off the vlans you don't need on these older switches which is good practice anyway . If need to add them later just add them as you need them .

 

[kalmanTRF]

There are about 40 Vlans, about half will be removed when the ip conversion is complete. I have turned on pruning on the 6500s, but have not changed the "allowed vlans" so that only the needed vlans are there. One reason being that I am trying to clean up our network to get only 2 or three vlans per switch. I just need to make sure that there are now problems with the vlans before I migrate the rest of the machine over to the new ones. Either way we have under the 64 vlans that a 3548 can handle.

 

[firestormsp1]

Ok, maybe silly questions, but are they the same ports that are affected each time? (duff switch?)

Also, the ports aren't error disabling are they? (Check config)
Have you assigned anything like MAC security to the edge switch ports and have someone repatching your kit? (Check config)

Hope this helps

Colin

 

[kalmanTRF]

No they are different ports and different switches. IOS is not the factor either, as I tried switches with different versions. It is hard to recreate sometimes and it might be VLAN XXX having a problem on one switch and vlan yyy on another. Doing a show int on the port shows that there are no errors. However a show controllers on the interface say that there are "16809 No dest, broadcast", no other errors. Is this indicative of something.

There is no mac security set up, and no disabling. If a port was error disabled, it wouldn't get a link light would it?

 

[chieftan999]

Have you checked your spanning tree configs?

Sounds like STP is blocking the ports and not forwarding out of them.
Have a look at all your spanning tree information and ensure that the ports that are not communicating are in FORWARDING mode and not BLOCKING - you may see them going to Listening or learning state but ,ultimately, they need to be forwarding.

Reamin positive. The affect on those around you will amaze.

 

[kalmanTRF]

'Show spanning-tree int' for the port that is not working shows that the port is in a forwarding state. It has the correct root and path costs so it is getting that info correctly. I have been reading on a subject of Spanning tree instances though and that a 3500XL series switch can only handle 64 instances. Anyone know how to tell how many instances there are running on a particular switch. Is it just a show spanning-tree summary and the number at the bottom. If so then there is no way a 3548 can handle 250 VLans as stated, unless you don't running spanning tree on them that is.

Just a weird issue.