rcp failure:getservbyname

[David Parsons]

I am able to push or pull files using rcp from the affected servers. They do not accept rcp requests from an outside server, though. Rlogin in either direction is unaffected, as is also telnet or ftp. The "remsh " is affected, in the same manner as rcp.

The 2 systems are HP-UX 11.00 with the March 2002 Quality Pack installed. I have uncommented the kshell/ekshell services. My .rhosts files have the entries for the servers and several others. The servers with the problem are NIS clients. Nslookup works correctly for all servers. The nsswitch.conf file points to local, then DNS. Inetsvcs_sec has Kerberos disabled. I followed the directions for disabling reverse lookup for remshd mentioned in PHNE_21731, and also downloaded and installed the superceding patch PHNE_23003 to one of the machines to see if it changed the results.

I suspect that it involves the Quality Pack, since I have another HP-UX 11.0 server with the Sep. 2001 Quality Pack installed which does not have this problem. It's files were compared to the affected servers and no differences in inetd.conf or services files were found. It is an NIS master Server for another NIS domain.

Any more suggestions to try would be much appreciated...

[someone_4]

Who are you logged in the servers as and who are you trying to remsh as?

Do you have permision to do the remsh in the .rhosts?

~ Richard

[someone_4]

and in /etc/inet.conf
the following should be uncommented:

shell stream tcp nowait root /usr/lbin/remshd remshd

kshell stream tcp nowait root /usr/lbin/remshd remshd -K
klogin stream tcp nowait root /usr/lbin/rlogind rlogind -K

after you uncomment them do
inetd -c
to reread the file and accept the changes.

~ Richard

[Steffen Jaiser]

If your error looks something like:
Fatal error on "workstation": (remsh error: rcmd: primary connection shutdown)
then try the following:

Edit /var/adm/inetd.sec
and look for the line shell if the server you try to connect from is within the IP ranges.

Hope that helped.

[David Parsons]

Everyone is going to love this:

The root problem was NIS. We had recently changed our master server to a Solaris system which was little utilized or rebooted for upgrades/changes. Solaris has no Kerberos entries for kshell, klogin, etc. The March 2002 Quality Pack brought these services into play more tightly, especially with NIS. But since the services file served by the Sun machine didn't have them listed, the getservbyname error came up for any external request for those services occurred. Local requests used the local services file, found them and worked.
Host3, the other HP-UX server just proves my point since it is an NIS master and no problems have occured on that domain.

Adding the line: "services: files nis" to the nsswitch file for the problem systems forced a local lookup first and removed the error.

Problem solved.