Boot switch from DHCP

Discussion in 'Cisco' started by Nick, Jan 18, 2005.

  1. Nick

    Nick Guest

    We currently have over 2000 2950s on our network. To configure each of
    these, the switch pulls an address from DHCP and downloads a small
    config from TFTP. Our engineers find the IP address by "sh cdp nei" at
    the distribution switch.

    Recently this has stopped working. So far we've seen that the Vlan1
    interface is no longer receiving the IP address, it's a random ethernet
    port. Then when booting another switch, it gets the same IP address,
    thus shutting down the VLAN of the first booted switch. The skinny
    config is never downloaded. It would seem like a DHCP server issue,
    but we can't prove this to be true. Also, loading an older IOS, "wr
    erasing" and reloading will boot from DHCP successfully.

    Any ideas?
    Nick, Jan 18, 2005
    #1
    1. Advertising

  2. In article <>,
    Nick <> wrote:
    :We currently have over 2000 2950s on our network.

    Over 2000? Yow!! And that's just the switches.. At say 75% population
    on a 48 port switch, that would imply .. 70,000-ish hosts??
    Just out of curiosity -- what do you use for your WAN links?

    At a guess, with over 2000 switches, you probably are doing
    self-maintenance rather than SmartNet, at least on the bulk of them.

    On the other hand, SAU (SA = software updates within one major release,
    U = updates between major releases) SmartNet contracts appear to cost
    relatively close to SNT (8x5xNBD software and hardware support)
    contract prices, so it's hard to predict how someone would handle the
    maint issues.


    :To configure each of
    :these, the switch pulls an address from DHCP and downloads a small
    :config from TFTP. Our engineers find the IP address by "sh cdp nei" at
    :the distribution switch.

    :Recently this has stopped working. So far we've seen that the Vlan1
    :interface is no longer receiving the IP address, it's a random ethernet
    :port.

    Could you expand on that point? Is the implication that the
    base configuration of the switches (saved into nvram) has a
    basic VLAN configuration with vlan 1 only on a small subset
    of ports and other ports initialized to other VLANs? I am
    not sure what you mean by "it's a random ethernet port" -- is
    the IP address being attached to a Layer 3 interface on a port
    other than the one you expect? Or do you mean that something
    odd is happening in your topology and the DHCP response packets
    are being received on a port other than what you expect for your
    uplink?

    :Then when booting another switch, it gets the same IP address,
    :thus shutting down the VLAN of the first booted switch. The skinny
    :config is never downloaded.

    The first switch thinks it has completed a DHCP cycle, but
    the DHCP server thinks it hasn't?? The tftp server is not accessible
    from the layer 3 port that is being assigned the IP address??

    :It would seem like a DHCP server issue,
    :but we can't prove this to be true. Also, loading an older IOS, "wr
    :erasing" and reloading will boot from DHCP successfully.

    With 2000 switches, you must have a good way to capture packets on
    arbitrary segments -- RSPAN or an inline packet analyzer or whatever.
    What do the DHCP packets say? If the DHCP cycle is complete and
    the server is assigning the IP address elsewhere then [at least
    some of] the problem is at the DHCP server. But if the switch's
    acceptance of the DHCP offer is not making it back to the DHCP
    server then the DHCP server is allowed to assign the IP elsewhere
    [presumably after some kind of timeout.]

    Sorry, I'm just not quite catching the situation.
    --
    If a troll and a half can hook a reader and a half in a posting and a half,
    how many readers can six trolls hook in six postings?
    Walter Roberson, Jan 18, 2005
    #2
    1. Advertising

  3. Nick

    PES Guest

    Nick wrote:
    > We currently have over 2000 2950s on our network. To configure each of
    > these, the switch pulls an address from DHCP and downloads a small
    > config from TFTP. Our engineers find the IP address by "sh cdp nei" at
    > the distribution switch.
    >
    > Recently this has stopped working. So far we've seen that the Vlan1
    > interface is no longer receiving the IP address, it's a random ethernet
    > port. Then when booting another switch, it gets the same IP address,
    > thus shutting down the VLAN of the first booted switch. The skinny
    > config is never downloaded. It would seem like a DHCP server issue,
    > but we can't prove this to be true. Also, loading an older IOS, "wr
    > erasing" and reloading will boot from DHCP successfully.
    >
    > Any ideas?
    >


    You need to packet sniff this and determine if there is an extra dhcp
    server somewhere (look at the mac addresses). It could also be a bootp
    relay configured (or ip helper address) on a router. I have seen
    situations where one server issues the ip address and the other does a
    dhcpnak at the same time. It creates a lot of issues. The issue would
    be less defined with a client, because they usually store the ip address
    of the dhcp server and don't broadcast unless completely necessary.
    Also, they will store their last retrieved address and try to renew it.
    However, this is not possible with the switch. My money is on a
    second dhcp server somewhere.

    --
    -------------------------
    Paul Stewart
    Lexnet Inc.
    Email address is in ROT13
    PES, Jan 18, 2005
    #3
  4. Nick

    Hansang Bae Guest

    PES wrote:
    > You need to packet sniff this and determine if there is an extra dhcp
    > server somewhere (look at the mac addresses). It could also be a
    > bootp relay configured (or ip helper address) on a router. I have
    > seen situations where one server issues the ip address and the other
    > does a dhcpnak at the same time. It creates a lot of issues. The
    > issue would be less defined with a client, because they usually store
    > the ip address of the dhcp server and don't broadcast unless
    > completely necessary. Also, they will store their last retrieved
    > address and try to renew it. However, this is not possible with the
    > switch. My money is on a second dhcp server somewhere.


    But putting the older IOS makes it work. So I'll go with "it's another
    bug"


    --

    hsb


    "Somehow I imagined this experience would be more rewarding" Calvin
    **************************ROT13 MY ADDRESS*************************
    Due to the volume of email that I receive, I may not not be able to
    reply to emails sent to my account. Please post a followup instead.
    ********************************************************************
    Hansang Bae, Jan 19, 2005
    #4
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. =?Utf-8?B?SGVpbkQ=?=

    Wireless DHCP clients cannot obtain an IP address from the DHCP se

    =?Utf-8?B?SGVpbkQ=?=, Jan 8, 2006, in forum: Wireless Networking
    Replies:
    0
    Views:
    2,811
    =?Utf-8?B?SGVpbkQ=?=
    Jan 8, 2006
  2. Ingo Hauf

    if Active Directory no DHCP? or: Where ist my DHCP

    Ingo Hauf, Oct 17, 2003, in forum: Computer Support
    Replies:
    2
    Views:
    498
    Ralph Wade Phillips
    Oct 18, 2003
  3. Fayza
    Replies:
    3
    Views:
    12,632
    Dan Shea
    May 12, 2004
  4. Vimokh
    Replies:
    3
    Views:
    5,588
    Vimokh
    Sep 6, 2006
  5. sonicgravy
    Replies:
    0
    Views:
    1,413
    sonicgravy
    Nov 13, 2006
Loading...

Share This Page