what's wrong with my HP server when using ftp?

[MA Qiang]

Host A lan0's IP: 10.10.40.72
Host B lan0's IP: 10.20.20.19
Host B lan0:1's IP: 10.20.20.40

Host A ftp to Host B, when use ip 10.20.20.19, the speed is lower than 7Mbps, but when use ip 10.20.20.40, the speed is normal, about 53Mbps. Why the speed is different in the same adapter? I think there is something wrong in Host A. Because I tested many IP, and found that if the target ip adress was end with a odd number such as 10.20.20.19, the speed was low, and if end with a even number such as 10.10.20.40, the speed is normal.

I was confused.

HELP!

[Sunil Sharma_1]

As my memory help me win HP you can not have two IP address of same subnet configured on one host. System will misbehave...

just unconfigure 10.20.20.40 interface and then see performence


Sunil

[MA Qiang]

In the MC/SG environment, we must set the lan0:1 and lan0 at the same subnet. Only the host A have the problem to access host B. My PC or other hosts can access both the IP at the normal speed.

[Sunil Sharma_1]

You can not have two IP address of same subnet in one system. System will misbehave

Have a look on this thread

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=183255


Sunil

[MA Qiang]

It was HP's engineers who config the MC/SG, they use the all the IP of one host in one subnet. There is no problem that other PCs or HP-UX Servers access host B. ONLY host A has the problem.

[lawrenzo]

have you checked the settings on the lancard:

# lanadmin
# lan
# ppa 0

this lists info on the lancard, depending on the speed of your network then ensure the speed is set correctly and is also set to full-negotiate.

hope this helps

[Nguyen Anh Tien]

It is not recomment to use same IP subnet on 2 HW card without Cluster SW.

[MA Qiang]

MC/SG is HP high availability clusters software. 10.20.20.40 is one of the float IPs.

[MA Qiang]

I found that the routers are different when host A traceroutes host B.
Maybe it is the reason. But how to prevent host A to use the slower router?

# traceroute 10.20.20.19
traceroute to 10.20.20.19 (10.20.20.19), 30 hops max, 40 byte packets
1 10.10.40.1 (10.10.40.1) 0.611 ms 0.427 ms 0.421 ms
2 10.10.47.6 (10.10.47.6) 0.337 ms 0.241 ms 0.229 ms
3 10.20.2.65 (10.20.2.65) 0.387 ms 0.277 ms 0.266 ms
4 venus2 (10.20.20.19) 0.201 ms 0.173 ms 0.171 ms

# traceroute 10.20.20.40
traceroute to 10.20.20.40 (10.20.20.40), 30 hops max, 40 byte packets
1 10.10.40.1 (10.10.40.1) 0.629 ms 0.428 ms 0.420 ms
2 10.10.47.38 (10.10.47.38) 0.427 ms 0.241 ms 0.229 ms
3 10.20.2.10 (10.20.2.10) 0.375 ms 0.274 ms 0.267 ms
4 venus2 (10.20.20.19) 0.214 ms 0.185 ms 0.210 ms

[Jannik]

Both are on the same interfase so, there should be no difference, the setting for the fysical interfase is the same, so no problem.
This is a strange error...
Could you send your routing and table:
netstat -rn
and
lanadmin -sx lan0
and
check that you are running the same speed on the switch. If its other than 1Gbit you should force it to 100FD (if the card is 10/100).

do your traceroute give same information to both IPs?
Is the DNS configured correctly (try put in IP and hosts in /etc/hosts)

[MA Qiang]

I have changed the IP from lan0 to lan2 which is a stand-alone adapter. I have not configed the DNS.

# ioscan -funC lan
Class I H/W Path Driver S/W State H/W Type Description
===================================================================
lan 0 0/0/0/0 btlan CLAIMED INTERFACE HP PCI 10/100Base-TX Core
/dev/diag/lan0 /dev/ether0 /dev/lan0
lan 1 0/2/0/0 igelan CLAIMED INTERFACE HP A6847-60101 PCI 1000Base-SX Adapter
lan 2 0/4/0/0 btlan CLAIMED INTERFACE HP A5230A/B5509BA PCI 10/100Base-TX Addon
/dev/diag/lan2 /dev/ether2 /dev/lan2

# netstat -rn
Routing tables
Destination Gateway Flags Refs Interface Pmtu
127.0.0.1 127.0.0.1 UH 0 lo0 4136
10.10.40.72 10.10.40.72 UH 0 lan2 4136
10.10.40.0 10.10.40.72 U 2 lan2 1500
127.0.0.0 127.0.0.1 U 0 lo0 0
default 10.10.40.1 UG 0 lan2 0

# netstat -in
Name Mtu Network Address Ipkts Opkts
lan2 1500 10.10.40.0 10.10.40.72 4494589 62162685
lo0 4136 127.0.0.0 127.0.0.1 17278 17278

# lanadmin -sx 0
Speed = 10000000
Current Config = NO LINK AUTONEG

# lanadmin -sx 1
Speed = 1000000000
The link is down. The speed and other link state information
are undefined. The configuration specified for this card is:
Speed = 1000 Full-Duplex.
Autonegotiation = On.

# lanadmin -sx 2
Speed = 100000000
Current Config = 100 Full-Duplex AUTONEG

[MA Qiang]

the different between odd and even addrress:
# traceroute -n 10.20.20.16
traceroute to 10.20.20.16 (10.20.20.16), 30 hops max, 40 byte packets
1 10.10.40.1 0.632 ms 0.421 ms 0.406 ms
2 10.10.47.38 0.307 ms 0.223 ms 0.212 ms
3 10.20.2.10 0.398 ms 0.249 ms 0.250 ms
4 10.20.20.16 0.696 ms 0.174 ms 0.201 ms

# traceroute -n 10.20.20.17
traceroute to 10.20.20.17 (10.20.20.17), 30 hops max, 40 byte packets
1 10.10.40.1 0.600 ms 0.416 ms 0.407 ms
2 10.10.47.6 0.301 ms 0.225 ms 0.216 ms
3 10.20.2.3 0.307 ms 0.255 ms
4 10.20.20.17 0.213 ms 0.167 ms 0.156 ms

*****************************************************************************
# traceroute -n 10.20.20.17
traceroute to 10.20.20.17 (10.20.20.17), 30 hops max, 40 byte packets
1 10.10.40.1 0.575 ms 0.426 ms 0.436 ms
2 10.10.47.6 0.303 ms 0.236 ms 0.229 ms
3 10.20.2.3 0.395 ms 0.274 ms 0.268 ms
4 10.20.20.17 0.208 ms 0.175 ms 0.179 ms
*****************************************************************************

# traceroute -n 10.20.20.18
traceroute to 10.20.20.18 (10.20.20.18), 30 hops max, 40 byte packets
1 10.10.40.1 0.628 ms 0.433 ms 0.408 ms
2 10.10.47.38 0.373 ms 0.226 ms 0.215 ms
3 10.20.2.10 0.432 ms 0.253 ms 0.250 ms
4 10.20.20.18 0.215 ms 0.177 ms 0.149 ms

# traceroute -n 10.20.20.19
traceroute to 10.20.20.19 (10.20.20.19), 30 hops max, 40 byte packets
1 10.10.40.1 0.600 ms 0.426 ms 0.410 ms
2 10.10.47.6 0.311 ms 0.222 ms 0.215 ms
3 10.20.2.65 0.368 ms 0.256 ms 0.250 ms
4 10.20.20.19 0.193 ms 0.159 ms 0.151 ms

# traceroute -n 10.20.20.20
traceroute to 10.20.20.20 (10.20.20.20), 30 hops max, 40 byte packets
1 10.10.40.1 0.613 ms 0.412 ms 0.405 ms
2 10.10.47.38 0.376 ms 0.240 ms 0.226 ms
3 10.20.2.73 0.466 ms 0.253 ms 0.251 ms
4 10.20.20.20 0.212 ms 0.158 ms 0.154 ms

# traceroute -n 10.20.20.21
traceroute to 10.20.20.21 (10.20.20.21), 30 hops max, 40 byte packets
1 10.10.40.1 0.594 ms 0.426 ms 0.409 ms
2 10.10.47.6 0.274 ms 0.222 ms 0.212 ms
3 10.20.2.65 0.321 ms 0.261 ms 0.250 ms
4 10.20.20.21 0.198 ms 0.155 ms 0.149 ms

# traceroute -n 10.20.20.22
traceroute to 10.20.20.22 (10.20.20.22), 30 hops max, 40 byte packets
1 10.10.40.1 0.631 ms 0.417 ms 0.404 ms
2 10.10.47.38 0.391 ms 0.235 ms 0.213 ms
3 10.20.2.10 0.320 ms 0.250 ms 0.248 ms
4 10.20.20.22 0.567 ms 0.168 ms 0.170 ms

# traceroute -n 10.20.20.33
traceroute to 10.20.20.33 (10.20.20.33), 30 hops max, 40 byte packets
1 10.10.40.1 0.625 ms 0.435 ms 0.414 ms
2 10.10.47.6 0.276 ms 0.221 ms 0.213 ms
3 10.20.2.3 0.309 ms 0.261 ms 0.249 ms
4 10.20.20.16 0.213 ms 0.173 ms 0.166 ms

# traceroute -n 10.20.20.34
traceroute to 10.20.20.34 (10.20.20.34), 30 hops max, 40 byte packets
1 10.10.40.1 0.618 ms 0.419 ms 0.412 ms
2 10.10.47.38 0.766 ms 0.418 ms 0.326 ms
3 10.20.2.10 0.397 ms 0.250 ms 0.249 ms
4 10.20.20.18 0.166 ms 0.156 ms 0.152 ms

# traceroute -n 10.20.20.35
traceroute to 10.20.20.35 (10.20.20.35), 30 hops max, 40 byte packets
1 10.10.40.1 0.597 ms 0.427 ms 0.409 ms
2 10.10.47.6 0.301 ms 0.224 ms 0.216 ms
3 10.20.2.3 0.389 ms 0.256 ms 0.251 ms
4 10.20.20.18 0.160 ms 0.156 ms 0.141 ms

# traceroute -n 10.20.20.40
traceroute to 10.20.20.40 (10.20.20.40), 30 hops max, 40 byte packets
1 10.10.40.1 0.646 ms 1.081 ms 0.446 ms
2 10.10.47.38 0.395 ms 0.242 ms 0.291 ms
3 10.20.2.10 0.573 ms 0.339 ms 0.281 ms
4 10.20.20.19 0.227 ms 0.180 ms 0.180 ms

[rick jones · Accepted solution]

As pointed-out it is perfectly valid to have IPs in the same or different subnets assigned to logical interfaces on the same physical interface.

If you assigne IP addresses in the same subnet to different physical (from the perspective of the transport) interfaces the behaviour may not be what you expect. By default, traffic will be received on each interface, but it will only go-out one of the interfaces.

Now as to why speed differences. Any chance there is say ipfilter enabled on lan0, but not lan0:1?

Can you take before and after netstat -p tcp snapshots and subtract one from the other with beforeafter?

(ftp.cup.hp.com/dist/networking/tools/)

Do your routers try to load balance in some way?