On 20 September 2012 17:04, Arif Ali <[email protected]> wrote:
> I am currently installing a cluster with SL6.3, and my comments on the
> questions in line below, with my 2 cents
>
>>> #0 - Yes, xcatconfig is trying to restart the nfs service for the first
>>> installing to export the /install directory. But the problem is that could
>>> you run 'service nfs restart' successfully without touch of 'rpcbind'
>>> service? If no, is it an issue of SL6?
>>
>> I don't think any RHEL based systems do this kind of dependency checking
>> (although later Fedora's do) - certainly RHEL5 and before will not start
>> portmapper for you.
>
> I've not had this issue, as I just do a standard base install, and do
> a "yum install xCAT" maybe it was somehow totally missed

I just retried this on a SL6 machine I just installed with xCAT, the
kickstart %package definition was basically:
%package --nobase
@core

with some additional stuff for the commands I see xcat using un some
of the postscript. After doing this, and "yum install xCAT", I see in
/var/log/messages:
Sep 20 17:43:55 test32 kernel: RPC: Registered udp transport module.
Sep 20 17:43:55 test32 kernel: RPC: Registered tcp transport module.
Sep 20 17:43:55 test32 kernel: RPC: Registered tcp NFSv4.1 backchannel
transport module.
Sep 20 17:43:55 test32 kernel: Installing knfsd (copyright (C) 1996
[email protected]).
Sep 20 17:43:55 test32 kernel: svc: failed to register nfsdv2 RPC
service (errno 111).
Sep 20 17:43:55 test32 kernel: svc: failed to register nfsaclv2 RPC
service (errno 111).
Sep 20 17:43:55 test32 kernel: nfsd: last server has exited, flushing
export cache
Sep 20 17:43:55 test32 kernel: svc: failed to register nfsdv2 RPC
service (errno 97).
Sep 20 17:43:55 test32 kernel: svc: failed to register nfsaclv2 RPC
service (errno 97).
Sep 20 17:43:55 test32 kernel: nfsd: last server has exited, flushing
export cache

it doesn't hang, but it doesn't work either. Starting rpcbind allows
nfs to start.

So, given that xcatconfig already starts nfs, I suggest it should also
start rpcbind.

>>> #1 - The rcons worked in pxe phase was because the console was control by
>>> firmware. Could you check the parameter for the console was set correctly.
>>> And what's it help if not add the console parameter to the kernel parameter?
>>
>> Using xnba I sometimes get the console through rcons - sometimes I don't.
>
> I use XNBA, have done since it's inception. But the "console=ttyXXX"
> is required for the purpose of viewing the console of the machine via
> rcons/wcons. Without this you have no console. Have you tried the
> recommended settings on you machine. I still have a test cluster with
> x3455's, and I had no problems with that, except that the serialspeed
> needed to be 19200 instead of 115200

My serial settings for the node are:
    serialport=0
    serialspeed=19200

and the kcmdline is:
    quiet repo=http://xcat.fe.up.pt/install/SL6/x86_64/
ks=http://xcat.fe.up.pt/install/autoinst/test32 ksdevice=bootif
cmdline console=tty0 console=ttyS0,19200

I don't need the console ATM, so I think I'll just remove the
configuration to get it installing consistently.

>>> #4 - Stopping loop install does NOT depend on any postscript.
>>
>> How does xcat know to change the boot status from "install" to "boot" after
>> the install is completed?
>
> The postscript, i.e.
>
> #INCLUDE:#ENV:XCATROOT#/share/xcat/install/scripts/post.rh#
>
> is required for the machine to communicate with the MN so that it
> knows about the install finishing, and change the relevant xnba/pxe
> settings for it to boot back again. This also grabs the relevant
> postscripts defined in postscripts.postscripts. but if you don't have
> any postscripts, then you should be ok, but the main one you would
> want to keep is the remoteshell; so that you can log into the machine
> without any passwords, and able to do you own post install stuff.
>
> The pre script is required, so that you can monitor the state of the
> install while it is installing, via nodestat.

Ok, that makes more sense, and explains why my installs keep looping
when I don't include the post.rh.

Do you know which postscript script contacts the MN to change the install state?

I just discovered that xdsh requires /usr/bin/ssh, so that should
probably be added as a Requires in the xcat.spec file as well.

Many thanks.

> regards,
> Arif
-- 
Jonathan Barber <[email protected]>

------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://ad.doubleclick.net/clk;258768047;13503038;j?
http://info.appdynamics.com/FreeJavaPerformanceDownload.html
_______________________________________________
xCAT-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/xcat-user

Reply via email to