No support for requested service: 49 from root

Users/Site Administrators
1

I have a v20.0.1 PBS setup on a cluster, I have a v18.2.1 on another cluster that is stuck at v18 for a bit. I would like to be able to do a qstat on the v18
pbs server against the other v20 pbs server machine.
to that end I built up the latest openpbs and its sitting in a directory on the v18 machine. (its still running v18)
when trying to use qstat from that build I get the following:

env PATH=/opt/pbs_v20/bin /opt/pbs_v20/bin/qstat @192.168.1.1
auth: error returned: -1
auth: Unable to authenticate connection (192.168.1.1:15001)
qstat: cannot connect to server 192.168.1.1 (errno=-1)

192.168.1.1 being the IP of the v20 pbs server.

in the server logs on the v20 machine:

12/15/2020 11:24:04;0100;Server@v20pbserver;Req;;Type 0 request received from root@v18pbserver, sock=15
12/15/2020 11:24:04;0080;Server@v20pbsserver;Req;?;No support for requested service: 49 from root
12/15/2020 11:24:04;0080;Server@v20pbsserver;Req;req_reject;Reject reply code=15005, aux=0, type=49, from root@v18pbserver
12/15/2020 11:24:04;0080;Server@v20pbserver;Req;?;No support for requested service: 49 from root
12/15/2020 11:24:04;0080;Server@v20pbserver;Req;req_reject;Reject reply code=15005, aux=0, type=49, from root@v18pbsserver

if I copy the v20 build I did on v18pbserver over to a directory on v20pbserver
and use that qstat all is fine. Aside from it never mentioning trying to do service 49.

12/15/2020 11:47:39;0100;Server@v20pbserver;Req;;Type 0 request received from root@v20pbserver, sock=15
12/15/2020 11:47:39;0100;Server@v20pbserver;Req;;Type 95 request received from root@v20pbserver, sock=16
12/15/2020 11:47:39;0100;Server@v20pbserver;Req;;Type 21 request received from root@v20pbserver, sock=15
12/15/2020 11:47:39;0100;Server@v20pbserver;Req;;Type 19 request received from root@v20pbserver, sock=15

thoughts?

thanks
steve

2

Bumping thread, sorry I cannot provide help, but I would also like to know where the

auth: error returned: -1
auth: Unable to authenticate connection (192.168.1.1:15001)
qstat: cannot connect to server 192.168.1.1 (errno=-1)

error comes from. I am having the same issues with the exact same pbs version on different machines (20.0.0), but when I change the PBS_SERVER value on the /etc/pbs.config to a different server, I get the same error: unable to athenticate connection.

SELINUX is disabled on all machines, DNS records are setup on /etc/hosts and pings work fine and Firewalld is disabled.

3

The problem is that the pbs_iff authentication protocol changed between version 18 and version 20. Somehow, your pbs_v20 build is using the pbs18 pbs_iff. Try setting PBS_EXEC to /opt/pbs_v20 before running qstat, etc.

I hacked our pbs19 version of pbs_iff to detect the particular error returned when talking to a pbs20 server and exec a pbs20 version of pbs_iff on top of itself.

It’s a real hack, but we don’t expect to be running mixed versions for long. I don’t have a hack for the reverse direction.

4

root can do it but not users. my user is a server manager on both ends and still not working.
but progress :slight_smile:

thanks
s

5

Do you get the same set of messages for non-root as before? Is the pbs_iff in /opt/pbs_v20/sbin setuid? Do you have flatuid set?

6

the install of v20 did not make pbs_iff setuid. changing that made things work for all users.
thanks

hopefully security wont try to break my kneecaps.

s