telnet client hangs on ES45

[Aco Blazeski]

Hi all,
I have very strange problem and I can't figure it out:
On a ES45 server (server1) telnet client hangs when I do telnet to other 3 ES40/45 (server2,3,4) ?
On the other hand telnet goes ok to any other servers (GS80, SPARCs etc.).

Name resolution is ok, that is when I do

server1:root# telnet server2
Trying correct.I.P.address ...

Any suggestions/ideas ??

Thanks

[Venkatesh BL]

did you try 'rcinet restart'?

[Aco Blazeski]

I don't think that 'rcinet restart' will do because from server1 I can do telnet to other servers (like GS80, SPARCs etc), but can't do telnet to server2, server3 and server4 which are ES40/45.
On the other hand, I can do telnet on server2, server3 and server4 from any server except from server1...

[Michael Schulte zur Sur]

Hi,

is there a filrewall involved?
Do you get any error messages in the target host errorlogs?

greetings,

Michael

[Ivan Ferreira]

1- Is there a firewall between the servers?
"Trying coneect" means that cannot reach the server.

2- Are you having the correct routes on server1?
check it with netstat -nr

3- When you run telnet from server1 to other ES40 server, run netstat -n on both servers and check the status of the connection that involves these two servers. Post the results.

[Aco Blazeski]

Hi,

All servers are in the same subnet, so no router/gateways are needed.
Second, no firewall in between.

However,this is what I have found:

The reason for my problem is, I believe:
server1 is a single-member cluster, the hostname is server1 and cluster name is srv1clus (so we have 2 different IPs).

Scenario 1: server1 - GS80
telnet works ok, GS80 pings server1 and srv1clus.

Scenario 2: server1 - server2
telnet from server1 to server2 hangs.
server2 pings server1.
server2 can't ping srv1clus --> this is very strange.
server2 resolves srv1clus name to correct IP.

Most probably problem in scenario2 is because server2 tries to respond to srv1clus,but can't reach it.

So the question is, why can't server2 ping srv1clus, like other servers do ??? This is very strange, all other servers can do it without problems, and all servers are in the same subnet.

[Michael Schulte zur Sur]

Hi,

can you give us conrete ips so that we can what which host is pinging to?

thanks,

Michael

[Ivan Ferreira · Accepted solution]

Check the /etc/ifaccess.conf file, this file is the only that may block network traffic.

Try adding the server1 IP/MAC to the arp table on server 2 and ping the cluster ip of server1, this will allow you to verify that the network switch is not preventing the access. (something with the spanning tree protocol). Use the arp command.

[Aco Blazeski]

Hi to all,

yea, the arp thing solved the problem :).
When I issue 'arp -a' on server2 I receive:
...
srv1.clus.IP.addr at (incomplete)
...

When I add server1 cluster IP on server 2 with arp then I can ping cluster IP on server1, and also successfully telnet from server1 to server2.

Thanks a lot

Now the question is why some servers dont have server1 cluster IP in arp table "srv1.clus.IP.addr at (incomplete)" ?!? :(

[Venkatesh BL]

Are those servers running routed or gated?

[Aco Blazeski]

Only server1 is running gated since it has cluster software installed on it.
Other servers (server2,3,4) have neither routed nor gated.

[Venkatesh BL]

I think you would see the cluster IP on the individual servers if you run 'gated' on them.

[Ivan Ferreira]

I'm not sure if you should use the static arp table as a permanent solution, you have a network problem, maybe you need to run tcpdump on both servers and check why the arp reply is not arriving.

You have incomplete at the arp table because you cannot contact the server.

When running tcpdump you should see something like this when trying to contact the server:

arp who-has 128.3.254.6 tell 128.3.254.68
arp reply 128.3.254.6 is-at 02:07:01:00:01:c4

To use tcpdump, see man packetfilter and tcpdump.