FTP problem

[Edmund Brown]

Hi Everyone

I have just completed a cold install of HP-UX 11i v1 on a C3000 machine. Unfortunately my ftp connection seems to be incredibly slow ~800 kbs (I am getting 60 mbs on our C8000 machine again running 11i).

Is it likely that the ftp is being throttled by software on the C3000? If so could someone point me in the right direction to make any configuration changes necessary?

Thanks
Ed

[Tim Nelson · Accepted solution]

My first guess is incorrect lan configs.

i.e. half-duplex

check your lan connections with lanadmin -x nmid

[A. Clay Stephenson]

Almost certainly, your problem is a speed/duplex mismatch between the host and its corresponding switch port. It's not enough to check one end of the connection; both ends must be checked. Moreover, if one end is set to auto-negotiate, the other end MUST be set to auto-negotiate. Similarly, if one end is hard set to say, 100Mbit/s, full-duplex, the other end MUST be hard set to 100 MBit/s, full-duplex. Surprisingly, a mismatched duplex setting will almost work well and nothing would seem amiss is a low bandwidth application such as telnet session, for example. Applications which require high-bandwidth (such as FTP or NFS) would immediately see performance problems.

[Edmund Brown]

Hi Tim

Here are the lanadmin statistics from firstly the client and secondly the server.

If these statistics support your theory of an incorrect lan config. could you please advise what needs to be changed.

Client:
PPA Number = 0
Description = lan0 HP PCI 10/100Base-TX Core [100BASE-TX,FD,
MANUAL,TT=1500]
Type (value) = ethernet-csmacd(6)
MTU Size = 1500
Speed = 100000000
Station Address = 0x1083cf0317
Administration Status (value) = up(1)
Operation Status (value) = up(1)
Last Change = 541
Inbound Octets = 67870173
Inbound Unicast Packets = 87400
Inbound Non-Unicast Packets = 145577
Inbound Discards = 0
Inbound Errors = 7722
Inbound Unknown Protocols = 27506
Outbound Octets = 18017994
Outbound Unicast Packets = 78295
Outbound Non-Unicast Packets = 4
Outbound Discards = 0
Outbound Errors = 0
Outbound Queue Length = 0
Specific = 655367

Ethernet-like Statistics Group

Index = 1
Alignment Errors = 0
FCS Errors = 7722
Single Collision Frames = 0
Multiple Collision Frames = 0
Deferred Transmissions = 0
Late Collisions = 0
Excessive Collisions = 0
Internal MAC Transmit Errors = 0
Carrier Sense Errors = 0
Frames Too Long = 0
Internal MAC Receive Errors = 0

Server:
PPA Number = 0
Description = lan0 HP PCI 1000Base-T Release B.11.11.09
Type (value) = ethernet-csmacd(6)
MTU Size = 1500
Speed = 1000000000
Station Address = 0x110a319bf5
Administration Status (value) = up(1)
Operation Status (value) = up(1)
Last Change = 3291
Inbound Octets = 3044601021
Inbound Unicast Packets = 3437998
Inbound Non-Unicast Packets = 1048485
Inbound Discards = 0
Inbound Errors = 0
Inbound Unknown Protocols = 192644
Outbound Octets = 1283115696
Outbound Unicast Packets = 2517523
Outbound Non-Unicast Packets = 172
Outbound Discards = 0
Outbound Errors = 0
Outbound Queue Length = 15
Specific = 655367

Ethernet-like Statistics Group

Index = 1
Alignment Errors = 0
FCS Errors = 0
Single Collision Frames = 0
Multiple Collision Frames = 0
Deferred Transmissions = 17315
Late Collisions = 0
Excessive Collisions = 0
Internal MAC Transmit Errors = 0
Carrier Sense Errors = 0
Frames Too Long = 0
Internal MAC Receive Errors = 0



Thanks
Ed

[Tim Nelson]

The stats are always for the local NIC.

The simplified form for LAN0 looks like this. In my case I have gig-ether cards and they are set to auto.

lanadmin -x 0
Speed = 1000 Full-Duplex.
Autonegotiation = On.

In your case ( assuming lan0 ) you should have 100FD
autoneg should be manual

The switch end must also be set to 100FD.

100MB cards do not do well at autoneg. Always hard set on the server and the switch.

If you are not using lan0 then the lanscan command with give you the nmid ( they typically match).

[A. Clay Stephenson]

Note that when I refer to "both ends" I do not mean a client and server (except in the case of the two boxes being directly connectly via a crossover cable) but rather the settings of each NIC and its matching port on an ethernet switch. A combination that is all but guaranteed to produce your problem is a setting of auto-negotiate on one end and a hard setting on the other.

[Edmund Brown]

My ftp connection is from a C8000 with a gig. card to my older C3000 with a fast ether. card. Autoneg was switched on for the C8000 but off for the C3000. When I switched the C3000 to on (via SAM) then the ftp started to zip along. Brilliant.

I am afraid to say I have very little knowledge of networks, but would I be fairly safe in assuming that our network switches will be set to auto negotiate?

Thanks for the advice (with an 800 kbs transfer rate I had a long night ahead).

Ed

[A. Clay Stephenson]

No, that is a silly assumption. A safe assumption is that most ethernet port switch settings default to auto-negotiate but can be hard-set to any value and these values are retained across reboots of the switch.

The only truly safe assumption is that you don't know how a port is set --- unless you examine the port's settings.

[Tim Nelson]

The trick as Clay mentioned is the server/workstation and the switch need to match.

With the older fast ether in the workstation you will probably be ok set to auto. After each reboot you have a 50/50 chance of it getting set right.

To avoid the issue, if it is important, then ask that the switch be set to 100FD then set the workstation to 100FD also ( /etc/rc.config.d/hpbtlanconf ).

Regards,