slow dhcp on 1G backbone.

[Dean Montgomery]

Here is our setup details...

* Debian etch running dhcp3-server V3.0.4

* HP Procurve 2810 core switch with 2610 switches in the classrooms.

* Linux server 2.6.18-6-vserver-amd64 #1 SMP GNU/Linux

* The link settings for the server:
===
ethtool eth0
Settings for eth0:
        Supported ports: [ MII ]
        Supported link modes:   10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
                                1000baseT/Half 1000baseT/Full
        Supports auto-negotiation: Yes
        Advertised link modes:  10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
                                1000baseT/Half 1000baseT/Full
        Advertised auto-negotiation: Yes
        Speed: 1000Mb/s
        Duplex: Full
        Port: Twisted Pair
        PHYAD: 1
        Transceiver: internal
        Auto-negotiation: on
        Supports Wake-on: g
        Wake-on: d
        Current message level: 0x000000ff (255)
        Link detected: yes
===

* The link settings for the clients:
===
ethtool eth0
Settings for eth0:
        Supported ports: [ MII ]
        Supported link modes:   10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
        Supports auto-negotiation: Yes
        Advertised link modes:  10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
        Advertised auto-negotiation: Yes
        Speed: 100Mb/s
        Duplex: Full
        Port: MII
        PHYAD: 1
        Transceiver: external
        Auto-negotiation: on
        Supports Wake-on: g
        Wake-on: g
        Link detected: yes
===

* Cables are CAT5e.

* No errors or overruns on ifconfig for both client and server.
===
          RX packets:30896793 errors:0 dropped:0 overruns:0 frame:0
          TX packets:34204246 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
===
* ping stats:
8 packets transmitted, 8 received, 0% packet loss, time 6996ms
rtt min/avg/max/mdev = 0.101/0.112/0.126/0.015 ms

* Gig backbone works great big improvement with diskless clients nfs etc.  Bonded lacp and non/bonded interfaces do not make a difference to slow dhcpd.

* I also just tried moving the dhcpd server off of a 1000 NIC and onto a 100 NIC in the server.  But this did not make a difference.  There is still at 10-30 second delay for dhcp requests.

* When we etherwake all the workstations in the school with a 2 second delay between each wake, dhcp can sometimes stall out for several minutes on the client.  At the same time nfs is fast with no lags.

* If I throw in a 10/100 switch instead of a 10/100/1000 then dhcp requests are very fast < 1second.

* for arguments sake we also tried "ping-check false;" in dhcpd.conf - but this did not help so we set it back.

* Any ideas why dhcp slows down when there is a 1G backbone?


-- 
Dean Montgomery
Network Support Tech./Programmer
dmonty at sd73.bc.ca
School District #73