Loosing network connectivity

[dictum9]

This is a 11.23 box (rx2600 Itanium) which suddenly lost all connectivity. Netstat -rn looked reasonable so I rebooted it. Now it has a hard time coming up, it got stuck on "Starting mail daemon". Any ideas?

[James R. Ferguson · Accepted solution]

Hi:

Check for the availablity of your DNS server.
'sendmail' relies heavily on DNS.

Regards!

...JRF...

[IT_2007]

Try to do control - pipe to break senamil process so that it will continue further. once it comes up check /etc/rc.log file for errors, failure and warnings.

[Luk Vandenbussche]

Can you ping to the server?

Have you changed a setting in your DNS

Have to tried another port on your switch.

[Marvin Strong]

/etc/rc.log might provide some information as to why it got stuck on starting mail.

If your network connections came up correctly. I would look at how your resolver is setup.
/etc/nsswitch.conf
/etc/resolv.conf

But also check the rc.log, syslog.log etc.. for any error messages. They might point you in the right direction.

[Peter Godron]

Hi,
check your total setup:
1. Network cable
2. Network card (light on?)
3. Network set up (dns etc.)
4. Network services on the machine



netstat -rn may have just read static tables, not actual network traffic.

[dictum9]

It cannot ping even the default router, I will try another port on the switch. No changes were made to the server before this sudden problem.

I think it's either the switch, or the network card is faulty.

[Arunvijai_4]

Hi ETC,

Have you modified anything in /etc/resolv.conf or /etc/nsswitch.conf files ? I could see that is one of the reason for mail daemon boot time..

-Arun

[IT_2007]

Before you attempt to change anything, check with network folks so that they can check switch side.

If there is no problem at switch side then look at netconf file for default router information.

cat /etc/rc.config.d/netconf |grep -v ^#

you would get all the information without commented fields.

[dictum9]

To reiterate, no changes at all were made to the system since the time it was up and stable... Not to resolv.conf, not to anything else.

I have a feeling it's either the port on the switch or a bad NIC or the CAT 5 cable itself went bad, if that's possible.

I will see about running cstm on the machine to see if it checks the NIC.

[Peter Godron]

Hi,
the first two tasks on my earlier post will identify any external hardware problems.
If you have traffic light on your card, that means your switch is ok and your calbe is ok.
Problem solving by elimination.

[Marvin Strong]

Yep solve by elmination, no OS changes physically look at the hardware as suggested.
OS files still entact.
link light yes/no?
network cable / card

[Sp4admin]

Hi etc,

Are yuo seeing any link lights on the NIC card? I would talk to the network people to verify that the port is good or needs to reset that port. Also start checking other Components. like cable, /etc/rc.config.d/netconf file, stop and start the network daemon, etc.

sp,

[Mel Burslan]

If I were you, I would not be so hardlined on "nothing has been changed" stance. When was the last time this server was reboot. If it is not 5 minutes ago, I would easily suspect a jump-the-gun type sysadmin doing a "temporary" change and forgetting to back it out.

hardware failure is always a possibility upon reboot. But considering the server class being so recent, I do not think that your hardware may have been worn out of many restarts, so that it will fail at this time. Again, bad hardware is still not totally out of question.

lanadmin -x lan#

may give you some detailed information on the interface as well.

[James R. Ferguson]

Hi:

You should look at '/var/adm/syslog/OLDsyslog.log' for any clues, too. Since you rebooted, the OLD log may hold the information that will decipher this riddle.

Regards!

...JRF...

[dictum9]

> Are yuo seeing any link lights on the NIC card?

Yes, the lights are there.