Re: autofs problem on rocks cluster with 2.6.12-rc4-mm2

[Ian Kent]

On Wed, 25 May 2005, Deepti Vyas wrote:

> Does anyone know of any similar problems and fixes?
> 
> Here is the problem description for Autofs after we install 2.6.12-rc4-mm2:

I'm not sure what you are describing here.
How does autofs enter into this description?
What version of autofs are you using?

> 
> 
> > on the compute node with 2.6.12 kernel I found that
> >
> > #demsg  show following error info.
> >
> > audit(1117060783.329:0): avc:  denied  { name_connect } for  dest=80
> scontext=user_u:system_r:unconfined_t
> > tcontext=system_u:object_r:http_port_t tclass=tcp_socket
> >
> > While on the nodes with 2.6.9 kernel, no such error message.
> >
> > This means frontend port 80 is close for the compute node with 2.6.12-mm
>  kernel.
> > After test on 2 nodes, following is what I got.
> >
> > [root@compute-0-1 ~]# telnet 192.168.1.250 80
> > Trying 192.168.1.250...
> > telnet: connect to address 192.168.1.250: Permission denied
> > telnet: Unable to connect to remote host: Permission denied
> >
> > [root@compute-0-0 ~]# telnet 192.168.1.250 80
> > Trying 192.168.1.250...
> > Connected to grid11.local (192.168.1.250).
> > Escape character is '^]'.
> > ^]
> > telnet> exit
> > ?Invalid command
> > telnet> quit
> > Connection closed.
> >
> > But on the frontend, iptables has no rule on private interface.
> >
> >> Looks like one of the compute nodes is advertising itself at the 411
> master. is 192.168.1.250 a compute node's IP??
> >>
> >> Follow the instructions from here:
> >>
> >> https://lists.sdsc.edu/pipermail/npaci-rocks-discussion/2004-July/006745.html
> >>
> >> Just installing the kernel RPMs under
> >> /home/install/contrib/enterprise/4/public/i386/RPMS/ is not sufficient.
> Did you add the kernel's entry in extend-compute.xml??
> >>
> >>
> >>>
> >>> This is post following my "snmpd.conf & mount /home problem " post.
> >>>
> >>> I think it is the autofs and 411 problem.
> >>>
> >>> cd /var/411
> >>>
> >>> [root@grid11 411]# make
> >>> make: Nothing to be done for `all'.
> >>>
> >>> [root@grid11 411]# cluster-fork "/opt/rocks/bin/411get --all"
> >>> compute-0-0:
> >>> Wrote: /etc/rpc
> >>> Wrote: /etc/group
> >>> Wrote: /etc/shadow
> >>> Wrote: /etc/auto/misc
> >>> Wrote: /etc/auto/master
> >>> Wrote: /etc/passwd
> >>> Wrote: /etc/auto/home
> >>> Wrote: /etc/auto/net
> >>> Wrote: /etc/services
> >>> compute-0-1:
> >>> Error: Could not reach a master server. Masters:
> >>> [http://192.168.1.250/411.d/ (-1)]
> >>> compute-0-2:
> >>> Error: Could not reach a master server. Masters:
> >>> [http://192.168.1.250/411.d/ (-1)]
> >>> compute-0-3:
> >>> Error: Could not reach a master server. Masters:
> >>> [http://192.168.1.250/411.d/ (-1)]
> >>> compute-0-4:
> >>> Error: Could not reach a master server. Masters:
> >>> [http://192.168.1.250/411.d/ (-1)]
> >>> compute-0-5:
> >>> Error: Could not reach a master server. Masters:
> >>> [http://192.168.1.250/411.d/ (-1)]
> >>> compute-0-6:
> >>> Wrote: /etc/rpc
> >>> Wrote: /etc/group
> >>> Wrote: /etc/shadow
> >>> Wrote: /etc/auto/misc
> >>> Wrote: /etc/auto/master
> >>> Wrote: /etc/passwd
> >>> Wrote: /etc/auto/home
> >>> Wrote: /etc/auto/net
> >>> Wrote: /etc/services
> >>> compute-0-7:
> >>> Error: Could not reach a master server. Masters:
> >>> [http://192.168.1.250/411.d/ (-1)]
> >>> compute-0-8:
> >>> Error: Could not reach a master server. Masters:
> >>> [http://192.168.1.250/411.d/ (-1)]
> >>> compute-0-9:
> >>> Error: Could not reach a master server. Masters:
> >>> [http://192.168.1.250/411.d/ (-1)]
> >>>
> >>> Actually compute-0-6 and compute-0-0 boot up with the default 2.6.9
> kernel.
> >>> all the others use the 2.6.12 kernel.
> >>>
> >>> By the way, I also try and build 2.6.11.10 kernel, works fine no
> problem at all.
> >>> Will this be the kernel's problem? But I use the same .config file
> when build the rpm.
> >>>
> >>> One more thing is I put both kernel 2.6.11 and 2.6.12-rc4-mm2  rpm
> under /home/install/contrib/enterprise/4/public/i386/RPMS/
> >>> I hope I could get 3 kernels installed on the compute node, but only
> 2.6.9 and 2.6.12 are installed.
> >>> Will this introduce some conflict? Or Rocks only use the newest kernel
> in that /RPMS directory.
> >>>
> >>>
> 
> 
> 
> > On Tue, 24 May 2005 09:04 am, Deepti Vyas wrote:
> >> I tried using 2.6.11.10 with patch-2.6.11-smpnice-1.diff.
> >
> > It partially worked but some bugs were present and it has been resynced
> with
> > the latest -mm kernel. Can you try
> > http://ck.kolivas.org/patches/smpnice/patch-2.6.12-rc4-mm2-smpnice-1.diff
> on top of
> > linux-2.6.12-rc4-mm2
> >
> > Cheers,
> > Con
> >
> > P.S. please don't email me 3 different ways and separate the mailing
> list email from the personal ones and so on. One copy cc'ed to relevant
> places is
> > enough :|
> >
> 
> 
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/