Connection times out

[dictum9]

When I telnet/ssh to this box, the connection times out. When I ping it. I get "no answer"
When I get on it via the console, I see in syslog over and over again:


May 21 10:40:04 saptst01 vmunix: NFS server devsys8 not responding still trying


Question, does this mean the NFS server got rebooted while there was a connection? Does this mean that I have to reboot the NFS client to restore connectivity?
or can I get away with restarting the nfsd daemon?

The box is rx4640 11.23.

[A. Clay Stephenson · Accepted solution]

Your problems are larger in scope than simply NFS so trying to restart NFS is a waste of time. If you could ping the box (which you can't) then proably restarting inetd would fix everything. A reboot MAY fix you but you could have network problems outside your box. You need to do some basic network troubleshooting to locate the the problem. It could be something as trivial as a bad/loose cable but it could be something much worse.

[Pete Randall]

This means that devsys8 is either still down or, at the very least, was down, though I would expect that 11.23 NFS should have recovered when devsys8 did come back. In my experience, if the client can't recover the connection, then the only hope is to re-boot - I have never successfully recovered by stopping and starting the daemons.


Pete

[dictum9]

I should add there is also a high number of collisions as shown by netstat -i. Three million to be exact.
Is that a separate problem or a consequence of the same issue with NFS?

[Sandman!]

Have you tried ping'ing its name or IP? Wondering if it could be DNS related. Also try doing a traceroute from and to the server. The collisions reported by netstat could manifest itself anywhere i.e. switch, lan card or cable or dns...

[sachit patil]

there is problem with lan . You have to check for proper cableing in lan switch or hub . Because one lan mixes to anther then this problem caused . You have to find out which hub/swtch u r using . What is speed of lan card .

[dictum9]

I think this is the problem, it needs to reset to 100 FD, auto=off.



# lanadmin -x 0
Speed = 100 Half-Duplex.
Autonegotiation = On.

[Sandman!]

Reset it with lanadmin -X and make it permanent in /etc/rc.config.d/hpbtlanconf file or the equivalent one on your system.

[James R. Ferguson]

Hi:

If you have a 100MB LAN card, set both your network switch _and_ your LAN card to 100MB FD. This later specification is made in the appropriate '/etc/rc.config.d/hp*conf' configuration file. Failure to specify both the switch and the LAN card setting leads to problems like you are seeing. 100MB cards do not auto-negotiate very well.

Regards!

...JRF...

[Tor-Arne Nostdal]

As mentioned it is important to make a permanent setting in the /etc/rc.config.d/hp*conf configuration file.

To determine which of these files you should change you should check what kind of network card you are using.
ioscan -fk | grep lan
(or ioscan -fkClan )
will show what card you have 'btlan', 'igelan'...

example:
myhost# ioscan -fkClan
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
lan 1 0/3/0/0 igelan CLAIMED INTERFACE HP A6825-60101 PCI 1000Base-T Adapter
lan 3 0/4/0/0 igelan CLAIMED INTERFACE HP A6825-60101 PCI 1000Base-T Adapter
lan 4 0/6/0/0 igelan CLAIMED INTERFACE HP A6825-60101 PCI 1000Base-T Adapter
lan 2 0/7/0/0 gelan CLAIMED INTERFACE HP A4929A PCI 1000Base-T Adapter

With 'netstat -i' can help you determine which network card you want to permanently change the configuration for.

From my example, if I want to change the 'lan 0' it would be the /etc/rc.config.d/hpbtlanconf, but if it was 'lan 1' it should be the /etc/rc.config.d/hpigelanconf and furthermore if it was 'lan 2' I need to do it in /etc/rc.config.d/hpgelanconf

Autonegotiation
- It is very important that you have the same setting as on the switch port
- A Giga card should always be auto

The attached script lanstatus.sh might be useful for you.

/Tor-Arne

[Steven E. Protter]

Shalom,

You may need to hard code hpbtlanconf, ONLY if you have a 100 BaseT card in the system.

This system should be connected to a switch set for auto negotiate and its better to use a 1000 BaseT switch.

I think you may have a problem in /etc/rc.config.d/netconf

Could be as simple as a stray character. Take a look. I've done this to myself before with a period that didn't belong.

If thats consistent, there may be a problem with the overall network environment.

Lastly, check the hardware with mstm cstm or xstm, a NIC on the road to death might be resonsible for this.

SEP