Description of problem: I'm trying to set up monitoring of a particular host, and not really getting anywhere. Version-Release number of selected component (if applicable): Spacewalk 0.5 (just installed from the main spacewalk yum repositories). How reproducible: It is the same for all hosts we have tried to add monitoring to. None of them exhibit any other behaviour, including one which is also CentOS 5 x86_64 (identical to the server). Steps to Reproduce: 1. Add client machine to spacewalk server 2. Make sure that rhnmd is running on the client 2. Add default load monitor with no params 3. Wait. Actual results: This is the error output from the spacewalk server monitoring daemon - it's the default load monitor for Spacewalk 0.5, with a test client being added. The server is running an up to date, recently installed CentOS 5 (x86_54) (nothing else is installed besides Spacewalk - the database is on our Oracle server). SELinux is in permissive mode on the server, but disabled entirely on the client, which is running a recently installed and updated Fedora 10 (i386). The RHN Monitoring Daemon (RHNMD) is not responding: ssh: : Name or service not known. Please make sure the daemon is running and the host is accessible from the monitoring scout. Command was: /usr/bin/ssh -l nocpulse -p 4545 -i /var/lib/nocpulse/.ssh/nocpulse-identity -o StrictHostKeyChecking=no -o BatchMode=yes /bin/sh -s If I run this command myself from the spacewalk server, I quickly notice that the order of the command should be slightly different, and that the hostname appears to be missing.. (it should be before the command being run): [root@gyne repo]# /usr/bin/ssh -l nocpulse -p 4545 -i /var/lib/nocpulse/.ssh/nocpulse-identity -o StrictHostKeyChecking=no -o BatchMode=yes /bin/sh -s mapc-test.maths.bath.ac.uk ssh: /bin/sh: Name or service not known The man page indicates that if you switch the ordering so that /bin/sh comes last, then ssh executes (although there is still a problem): [root@gyne repo]# /usr/bin/ssh -l nocpulse -p 4545 -i /var/lib/nocpulse/.ssh/nocpulse-identity -o StrictHostKeyChecking=no -o BatchMode=yes -s mapc-test.maths.bath.ac.uk /bin/sh Warning: Permanently added 'mapc-test.maths.bath.ac.uk,138.38.96.47' (DSA) to the list of known hosts. Request for subsystem '/bin/sh' failed on channel 0 If I go to the host we are trying to connect to and tail /var/log/messages I see this: Jun 24 10:00:55 mapc-test rhnmd[16022]: subsystem request for /bin/sh failed, subsystem not found Jun 24 10:01:05 mapc-test rhnmd[16028]: WARNING: /etc/ssh/moduli does not exist, using fixed modulus Jun 24 10:01:05 mapc-test rhnmd[16028]: Accepted publickey for nocpulse from 138.38.33.22 port 46790 ssh2 Jun 24 10:01:19 mapc-test rhnmd[16032]: WARNING: /etc/ssh/moduli does not exist, using fixed modulus Jun 24 10:01:19 mapc-test rhnmd[16032]: Accepted publickey for nocpulse from 138.38.33.22 port 46792 ssh2 Jun 24 10:01:19 mapc-test rhnmd[16032]: fatal: chown(/dev/pts/0, 489, 5) failed: Operation not permitted Jun 24 10:01:19 mapc-test rhnmd[16032]: Attempt to write login records by non-root user (aborting) Expected results: Presumably it should tell me what the load is, or if it is too high.. Additional info:
Just in case you hadn't came across this yet, the "-s" should go after the /bin/sh. In the above it's being treated as an argument to ssh rather than sh.
Please try following on you spacewalk server: su - nocpulse /usr/bin/ssh -l nocpulse -p 4545 -i /var/lib/nocpulse/.ssh/nocpulse-identity mapc-test.maths.bath.ac.uk If this will not work, then the error is most probably not in rhnmd itself. Since rhnmd is just simple wrapper around sshd to force it start on port 4545, I doubt, there is error in rhnmd itself. My tip is some PAM or other policy in game on client.
We've upgraded since then, and it seems to be working now. I have no idea what the fix was, but would assume that it was something that changed between versions as we have made no changes.
OK. Closing. If you find the problem again, feel free to reopen this bug.
Spacewalk 0.8 has been released