Interminent DHCP issues

[MartinM]

I once had a similar issue.
I put a Wireshark trace on the phone and found that the phone got it's initial IP address on the native vlan, read the options, switched to the voice vlan and issued another DHCP request.  The DHCP on the voice vlan was issued from the 3300.  This was normal operation but what I saw that wasn't normal was that after the second DHCP request both the DHCP server on the native VLAN and the 3300 responded to it.   Since the server on the native vlan responded faster than the 3300 the phone took that IP address.  So the phone got an IP address for the data vlan but was trying to talk on the voice vlan.

We finally traced the issue to the server on the data vlan was connected to a dataswitch port that had both vlans assigned to it.   Changing the dataswitch port to be in Access Mode for the data vlan resolved the problem.

Ralph

Thanks for that info Ralph.

That is strange behaviour but you might be on the right track. I have set the DHCP delay on the voice scopes to 1000ms and the issue seems to be much better now. Still happens once every 5-6 tries but not as often as before.

The uplink for the servers on our blade centre is tagged as this, there is no Voice VLAN tag on the port. Not sure if I will be able to change the uplink as it supplies traffic to the whole vSphere and all the required data paths.

Code: [Select]


!
interface GigabitEthernet1/0/9
 switchport access vlan 10
 switchport trunk native vlan 10
 switchport mode trunk


Thanks again

M.

[jrg0852]

You can also disable CDP for one phone and see if it makes a difference.

[ralph]

This may be the problem:

switchport trunk native vlan 10
 switchport mode trunk

I'm not a CNA but that doesn't look right to me.
It would be an easy after hours test to check it.

Ralph

[martyn]

This may be the problem:

switchport trunk native vlan 10
 switchport mode trunk

I'm not a CNA but that doesn't look right to me.
It would be an easy after hours test to check it.

Ralph

That part will be ok, but it will nullify the switchport access vlan command.

Are the links to the blade centre a port channel/etherchannel/LAG/LACP? If so, what is the Port Channel config, and does every single port that is a member of that Port Channel have identical config? If there is one that doesn't then that could explain some strangeness.
If you do a show int trunk, they should all match up as well across the Port Channel and the member interfaces.

Is DHCP being done on a physical controller, or on a server on the blade centre (either windows or a vMCD)?

Also btw, your dscp marking should be 46 not 56 on your option 125.

[MartinM]

You can also disable CDP for one phone and see if it makes a difference.

Already tried this and it didnt seem to make any difference

Thanks though

[MartinM]

That part will be ok, but it will nullify the switchport access vlan command.

Are the links to the blade centre a port channel/etherchannel/LAG/LACP? If so, what is the Port Channel config, and does every single port that is a member of that Port Channel have identical config? If there is one that doesn't then that could explain some strangeness.

I am sorry, I am not sure what you mean by this? I am not fully CNA qualified, could you please aleborate?

If you do a show int trunk, they should all match up as well across the Port Channel and the member interfaces.

This is what shows on the 2 switches when the command is run

Code: [Select]

User/Phone switch:
show int trunk

Port        Mode             Encapsulation  Status        Native vlan
Gi3/0/1     on               802.1q         trunking      31

Port        Vlans allowed on trunk
Gi3/0/1     1-4094

Port        Vlans allowed and active in management domain
Gi3/0/1     1,10,31,1001

Port        Vlans in spanning tree forwarding state and not pruned
Gi3/0/1     1,10,31,1001

Server switch:

Server_Switch#show int trunk

Port        Mode             Encapsulation  Status        Native vlan
Gi1/0/1     on               802.1q         trunking      10
Gi1/0/9     on               802.1q         trunking      10
Gi2/0/9     on               802.1q         trunking      10

Port        Vlans allowed on trunk
Gi1/0/1     1-4094
Gi1/0/9     1-4094
Gi2/0/9     1-4094

Port        Vlans allowed and active in management domain
Gi1/0/1     1,10,31,104,1001,1100-1101,1111,2000
Gi1/0/9     1,10,31,104,1001,1100-1101,1111,2000
Gi2/0/9     1,10,31,104,1001,1100-1101,1111,2000

Port        Vlans in spanning tree forwarding state and not pruned
Gi1/0/1     1,10,31,104,1001,1100-1101,1111,2000
Gi1/0/9     1,10,31,104,1001,1100-1101,1111,2000
Gi2/0/9     1,10,31,104,1001,1100-1101,1111,2000


Is DHCP being done on a physical controller, or on a server on the blade centre (either windows or a vMCD)?

It is done via Windows Server 2008 R2 Virtual Machine on VMware 5.5

Also btw, your dscp marking should be 46 not 56 on your option 125.

Thank you, what does the difference mean please?

Thanks again for the help.

M.

[martyn]

DSCP is the QoS marking for the RTP stream during a phone call. The RTP stream should be marked as 46/EF/ToS 5, and the signalling should be set to DSCP 26.

If you have the RTP set to be 56, then this makes it ToS 6, which is usually used for routing protocols, and network control information.

So if for instance you have voice traffic going over a WAN between sites, and your provider is expecting to be seeing RTP marked with a value of 46, and you are marking it as 56, then they will not be giving it the priority over the other forms of traffic it needs as it wont be matching what they are expecting.

Do you have a drawing of how the network looks, as I am guessing from your info above on the trunk information that there is more than one switch involved here.

[127.0.0.1]

For the phone interface you could use either

Code: [Select]

switchport mode access
switchport access vlan 10
switchport voice vlan 1001
or

Code: [Select]

switchport mode trunk
switchport trunk native vlan 10
switchport voice vlan 1001
switchport trunk allowed vlan 10,1001


The latter is a little older method but it's the one I've seen more of.  The former only works on newer IOS releases. 


If this still doesn't help you'll need to dig deeper.  As a test, put a DHCP server in an access port for 1001 on the same switch as the phone.  No dhcp-helper needed, no trunking to other switches.  Break this down to the bare minimum.  You can do all of this w/o affecting production if you restrict the temp local dhcp server with an appropriate reservation to your test phone MAC.  It might still hit your production dhcp server but it sounds like that's the whole issue.

[MartinM]

DSCP is the QoS marking for the RTP stream during a phone call. The RTP stream should be marked as 46/EF/ToS 5, and the signalling should be set to DSCP 26.

If you have the RTP set to be 56, then this makes it ToS 6, which is usually used for routing protocols, and network control information.

So if for instance you have voice traffic going over a WAN between sites, and your provider is expecting to be seeing RTP marked with a value of 46, and you are marking it as 56, then they will not be giving it the priority over the other forms of traffic it needs as it wont be matching what they are expecting.

Do you have a drawing of how the network looks, as I am guessing from your info above on the trunk information that there is more than one switch involved here.

Thanks, I will change the RTP value as suggested.

As for the diag, we have very complex environment, but to summarise, 2 user switches (one for each level) connected to our core firewall. The Core switch that the Mitel and the DHCP are both connected to is also connected to the Core firewall. So the handset goes to the user switch, then firewall, then core switch then Mitel controller

[MartinM]

For the phone interface you could use either

Code: [Select]

switchport mode access
switchport access vlan 10
switchport voice vlan 1001
or

Code: [Select]

switchport mode trunk
switchport trunk native vlan 10
switchport voice vlan 1001
switchport trunk allowed vlan 10,1001


Thanks I will try and see if this makes any difference

If this still doesn't help you'll need to dig deeper.  As a test, put a DHCP server in an access port for 1001 on the same switch as the phone.  No dhcp-helper needed, no trunking to other switches.  Break this down to the bare minimum.  You can do all of this w/o affecting production if you restrict the temp local dhcp server with an appropriate reservation to your test phone MAC.  It might still hit your production dhcp server but it sounds like that's the whole issue.

Yes, I have thought of this, but due to the complexity of the environment, this is not as easy as it seems.
I have however considered moving the Voice VLAN scope to the Cisco switch instead of the WinServer DHCP. Do you think this could potentially help? Anyone else have experience with this? The 2 user switches voice vlans already draw on 2 different DHCP scopes anyway, so this would be an issue?

Thanks again

M.

[127.0.0.1]

Try everything you can.  Usually you need to change to option 43 when serving from Cisco routers, not sure about switches.  Attaching a wireshark filtered on bootp would be good.  Why multiple scopes?


Sent from my iPad using Tapatalk

[MartinM]

Try everything you can.  Usually you need to change to option 43 when serving from Cisco routers, not sure about switches.  Attaching a wireshark filtered on bootp would be good.  Why multiple scopes?


Sent from my iPad using Tapatalk

I can set up port mirror and capture the bootp messages if you can read those. They are pretty confusing to me. I will post those when I get a chance

As for why the different scopes - to accommodate different floors with lot of users.

Thanks for the help

M.

[127.0.0.1]

Different floors in the same voice vlan, right?  So you have one scope with two pools or two separate scopes with one pool each? 


Sent from my iPad using Tapatalk

[MartinM]

2 scopes with one pool each

[ralph]

One thing I guess I forgot to ask.
Are you using the same DHCP server for both vlans or is the 3300 handling the voice vlan?


Ralph