connection problem

[Radhakrishnan Venkatara]

hi all,

I facing an intermittent problem
connecting from a workstation to the server.I am able to connect to all other servers in the same network apart from this server.This server is also able to connect to other workstations and servers only to this workstation.this problem occurs intermittently.

All servers and workstations and servers are in same network.
I tried pinging using ip , i deleted the arp entries also.
But nothing seems to work.
Please throw some inputs in which way i diagnoise it.

thanks and regards

radhakrishnan

[RAJESH GANGADHARAN]

Hi

Are you using DHCP if not check for duplicate IP address, you can use /opt/nettladm/bin/nettladm to check this.

Regards
Rajesh

[Radhakrishnan Venkatara]

Hi,

there is no DHCP used. servers and workstations are in same VLAN. at a point of time i am able to connect to other servers from the workstation and to this server.
nettladm has no messages.

thanx

radhakrishnan

[Michael Duthie]

I have seen a funny like this caused by a duplicate mac address. The mac address in /etc/rc.config.d/hp100baseconf had been modified at some stage and then this box was ignited and used to build a new system, resulting in a dupilcate mac address. This caused various strange & intermittant problems.

[Steve Labar]

What are you using for name resolution, /etc/hosts or DNS? I have seen problems if the same host and IP are in the hosts file more than once. Another thing I would try is an nslookup of the IP you want to contact when the stations are not able to connect and verify you get the hostname you are expecting. Try the nslookup from both stations because a problem at either station would cause problems.

Good Luck.

Steve

[T G Manikandan]

check that specific server for the speed and duplex settings.

check whether they mismatch with the settings of the switch to which it is connected.

Keep 100FD auto_neg off.

you can use

#lanadmin
to check up the status and the errors of the running card

Thanks

[Steven E. Protter]

Let me add to the last post.

There are problems with certain switches and HP lan cards. Cisco has been particularly troublesome to me.

If the max speed on the lan card is 100 BaseT full duplex, do hot use auto negotiate on the switch or the lan card. Have the switch admin hard code your desired speed into the switch and then edit this file.

/etc/rc.config.d/hpbtlanconf

Here is an example where lan0 is 10 BaseT Half, lan1 is 100 BaseT full fuplex.


HP_BTLAN_INTERFACE_NAME[0]=lan0
HP_BTLAN_STATION_ADDRESS[0]=
HP_BTLAN_SPEED[0]=10HD
HP_BTLAN_INTERFACE_NAME[1]=lan1
HP_BTLAN_STATION_ADDRESS[1]=
HP_BTLAN_SPEED[1]=100FD


This will insure you get the speed you expect. It will help eliminate a possible cause.

SEP

[Radhakrishnan Venkatara]

hi,

Already we have manually forced the cards to 100FD and we have set it in /etc/rc.config.d/hpbtlanconf file.
Today morning we forced switch port to 100FD manual.

While the workstation is not connecting to the particular server i can do nslookup without any problems in both server and workstation . it is returning the address for the server.
Today i found another thing , while the hosts are not getting connected , if connect to other server from the workstation and from that server if connect to the server is not connecting and from there if i start to ping the workstation it will start pinging , but ping response will start with 3000ms and it will reduce it to 0ms.

radhakrishnan

[T G Manikandan]

How about the routing table.

Does it have a long route?

check it

[Radhakrishnan Venkatara]

hi
servers and workstations are in the same subnet.


radhakrishnan

[Ron Kinner]

I have seen this same problem on our 3Com switches which allow you to easily set a static entry in a Port's MAC table. This makes it a bit quicker but you run into trouble when you forget and move the device to a different port (in our case on a different switch in the network) and don't delete the static MAC entry. The connection would work all week then over the weeked when the dynamic arp expired the static one would take over again and on Monday we had no connectivity. If the right host tried to talk to the "missing" device it would start working. You could also route it through the router and it would always work.

So next time it is not working ask your network guy to check his switches' arp tables (show cam dynamic in Cisco talk)to see if the switches know where both MACs are and to verify that they do not have another MAC entry on a different port or switch. You could also just change your MACs to something else using the /etc/rc.config.d/hpbtlanconf file and see if that fixes the problem.

Another diagnostic test that no one has mentioned is the linkloop test. You can test your connectivity without worrying about IP addresses and if you set the -v option you get nice error messages that explain anything that goes wrong. If you only have one NIC then

linkloop -n 10 -s 1400 -v

will give the LAN a good workout. I suspect it would also determine that a duplicate MAC existed. I'd run it to one of the sites that works first so you can see what a good run looks like. If you have more than one NIC you have to tell it which one to use with -i ppanumber.

Also next time that it fails, disconnect the server's network cable and have another system ping the ip address of the server. Verify that there are no responses. Without DHCP squatters are common. Someone connects up a PC and gives it an IP address which is already in use but currently not active so a ping comes back clean and when the first computer comes back on line neither can talk reliably.

As workarounds you can put static entries in your server's arp table and that of your workstation too. Also put static entries in your switch arp tables. You can also add entries in the server and workstations routing tables which say that to talk to each other they have to go via the router. That's what we did until we found the problem. Course that only works if the router is on the same switch as the device which has the problem MAC.

Ron

[Ron Kinner]

Oops. Left out the MAC. It should read:

linkloop -n 10 -s 1400 -v MAC

Where MAC is the MAC of another device on the same LAN.

Ron

[Radhakrishnan Venkatara]

hi ,

I am getting this error in linkloop when the workstation is not connecting to that server.

osshp05#linkloop -n 10 -s 1400 -v 0x0060B0F97B0A (mac address of the server)
Link connectivity to LAN station: 0x0060B0F97B0A
error: get_msg2 getmsg failed, errno = 4
error: get_msg2 getmsg failed, errno = 4
error: get_msg2 getmsg failed, errno = 4
error: get_msg2 getmsg failed, errno = 4
error: get_msg2 getmsg failed, errno = 4
error: get_msg2 getmsg failed, errno = 4
error: get_msg2 getmsg failed, errno = 4
error: get_msg2 getmsg failed, errno = 4
error: get_msg2 getmsg failed, errno = 4
error: get_msg2 getmsg failed, errno = 4
-- FAILED
frames sent : 10
frames received correctly : 0
reads that timed out : 10

this error comes only for one server while not getting connected.


then i excuted the command in server.I got this error.
[root@vindhya]# linkloop -n 10 -s 1400 -v 0x001083F9159F
Link connectivity to LAN station: 0x001083F9159F
error: expected primitive 0x30, got DL_ERROR_ACK
dl_error_primitive = 0x2d
dl_errno = 0x04
dl_unix_errno = 57
error - did not receive data part of message

this error comes for all the workstations and servers in the network but they are able to connect to this server. only one workstation is giving problem intermittently.


thanks and regards

radhakrishnan

[T G Manikandan]

ENOLINK 57 /* the link has been severed */

It still happens to be like the speed and duplex settings do not match with the other end.

1.Again confirm the speed and duplex settings
2.Try replacing the cable.


Use lanadmin to check the status and speed of the card.

Reset the card using the lanadmin.

Revert

[Ron Kinner]

"error: expected primitive 0x30, got DL_ERROR_ACK
dl_error_primitive = 0x2d
dl_errno = 0x04
dl_unix_errno = 57
error - did not receive data part of message"

This is a known problem on 10.20 systems and is fixed with PHNE_10078 which is replaced by PHNE_13650. The info from the server is not much use because of this.

Obviously if we can talk to other hosts on the network our NIC connection has to be good. Since linkloop fails from the workstation it is not an IP issue so we can rule out duplicate IP addresses.

My feeling is that we have a switch problem since the switch is responsible for getting the packet from one host to the other and is the only thing besides the NICs which even cares about the MAC addresses involved. Is the workstation which fails on a different switch from the server? Can you move it (temporarily) to the same switch? A different switch? A different port? Can you stick a different PC on the same port (with the same IP address as the workstation)?

I'd flush all of the mac tables on the switches (what kind are they by the way?) and make sure there wasn't an old static MAC hanging around
then connect up the workstation and verify that the workstation's mac showed up in all of the switches' tables.

Ron