[GE users] SGE 6.0 and shadow master troubles

Marco Donauer Marco.Donauer at Sun.COM
Thu Aug 12 16:33:08 BST 2004


Bogdan,

sorry for my late response, but I was on vacation the last days!


Bogdan Lobodzinski wrote:
> Hello Marco,
> 
> On Tue, 20 Jul 2004, Marco Donauer wrote:
> 
> 
>>Bogdan,
>>
>>normally, the installation script installs the shadowd on local maschine.
>>Using the -host option forces the script to install on a remote
>>maschine. In this case
>>any shell is used to login into the remote host. The default shell is ssh.
>>Did you start the inst_sge with -host option?
> 
> No, I did not start "-host" option. I tried to install it locally after
> login as root on the machine.
> 
> 
>>I had a look onto the ths script and found out, that the shadowhostname
>>is compared to the return value of hostname command.
>>If this is not equal, the script uses also the ssh to rlogin. Is it
>>possible that your hostname command returns any different name like the
>>fully qualified name?
> 
> The response of hostname command is (as should be):
> [balin] ~SGE_ROOT # hostname
> balin
> 
> 
>>I think the problem is the hostname comaprison.
>>
>>Would it be possible for you to file a bug? I guess this is a bug!
> 
> you mean to send a new issue into "Project Issue Tracking: gridengine" ?

Yes, right!

> 
> And I just realize possible another problem connected with sgeshadowd.
> Let we suppose such a configuration (this is exactly what I would like to
> start):
> Host A: primary sge master (NOT exec host);
> Host B: shadow master      (NOT exec host);
> 
> Installing sgeshadowd separatelly on Host B the startup script is not
> created. I understand it must be sgemaster. Just now I have to create this
> separatelly.
> I guess it will be nice to add creation of startup scripts (like in case
> of sgemaster installation) into pure shadow master installation procedure.
> 
> 
>                     Thanks a lot for your help,
> 
>                            Bogdan
I guess your, right! In 5.3 it was necessary that the SGE_ROOT dir was 
mounted on
the shadow host. -> The startup scripts was available for the shadow host.
Now, If you are using a berkeley db rpc server the mount is not 
mandatory for
a running shadowd and so the rc-script is missing!

We should think about this point!

Thanks!

Regards,
Marco
> 
> 
> 
>>Best regards and thanks for your help,
>>
>>Marco
>>
>>Bogdan Lobodzinski wrote:
>>
>>
>>>Hi Marco,
>>>
>>>On Tue, 20 Jul 2004, Marco Donauer wrote:
>>>
>>>
>>>
>>>
>>>>Hi Bogdan,
>>>>
>>>>I have currently no experience with afs, but as I know
>>>>only the qmaster must be installed with the afs switch.
>>>>If the shadow_master tries to take over he will read the bootstrap file,
>>>>which
>>>>contains the information about the security level and so the qmaster
>>>>will automatically start
>>>>in right mode.
>>>>
>>>>The command you have use is a littlebit wrong.
>>>>
>>>>To install the qmaster you can use install_qmaster -afs that's right.
>>>>To install the shadow please use inst_sge -sm
>>>>
>>>>The install_qmaster is a wrapperscript which execs this command: inst_sge -m
>>>>
>>>>I guess that you alread have installed a qmaster with install_qmaster -afs
>>>>With your command the qmaster installation will be started once again.
>>>>This should result in an error!
>>>>
>>>>
>>>
>>>Thanks for this hint.
>>>
>>>
>>>
>>>
>>>>Bogdan Lobodzinski wrote:
>>>>
>>>>
>>>>
>>>>>Hello,
>>>>>
>>>>>    I have some questions about shadow master installaton:
>>>>>1.
>>>>>When I try to install shadow master I am always failed because at
>>>>>some point the command "install_qmaster -afs -sm" tries to use ssh to
>>>>>contact with second master.
>>>>>Is it necessary to use ssh ?
>>>>>How to avoid it and make the installation fully succesfull ?
>>>>>
>>>>>I am starting all daemons as root. All
>>>>>ssh connections via root to another hosts are disabled.
>>>>>All necessary directories: ../common, .../spool/qmaster  are shared (rw)
>>>>>via AFS on both servers: master and shadow master and are available by
>>>>>root.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>Hm, yes your are right!
>>>>You think about the question about SGE_ROOT and SGE_CELL, right?
>>>>But if you install the shadow on a different host than the qmaster,
>>>>which makes more sense, you
>>>>can't get this information from anywhere. We have to ask for the
>>>>SGE_ROOT and SGE_CELL.
>>>>These path are save within the bootstrap file, that's right, but the
>>>>script can't find the bootstrap file without
>>>>SGE_ROOT and SGE_CELL
>>>>
>>>>
>>>
>>>You are right: I forgot to export SGE_CELL variable.
>>>But the problem with ssh connection during shadow master instalation still
>>>exist. So I decided to remove ssh parts from the script inst_sge,
>>>what finally allows me to start the sge_shadowd.
>>>
>>>
>>>
>>>
>>>>>2.
>>>>>according to the manual qmaster must be installed before shadow master.
>>>>>But installing shadow master I have to repeat exactly thge same procedure
>>>>>like during qmaster installation.
>>>>>again: Is it really necessary ?
>>>>>I guess something is wrong with this.
>>>>>
>>>>>3.
>>>>>Is it possible to modify the time after which sgeqmaster is started on
>>>>>shadow master host ?
>>>>>
>>>>>
>>>>
>>>>This is a compiled in option. Do you compile your own distribution or do
>>>>you use the binaries?
>>>>
>>>>
>>>
>>>I am using a source compilation.
>>>
>>>
>>>
>>>
>>>>If you compile your own binaries you can change it, this way:
>>>>
>>>>open source/daemons/shadowd/shadowd.c
>>>>
>>>>you can find these defines:
>>>>
>>>>#define CHECK_INTERVAL      60
>>>>#define GET_ACTIVE_INTERVAL 240
>>>>#define DELAY_TIME          600
>>>>
>>>>Here you can change the interval and the delaytime till sge_qmaster will
>>>>be started.
>>>>In this case: check every 60 sec's and start after 600 sec's
>>>>
>>>>
>>>
>>>Thanks for this pointer !
>>>
>>>                 Best Regards,
>>>
>>>                      Bogdan
>>>
>>>
>>>
>>>
>>>>Regards,
>>>>
>>>>Marco
>>>>
>>>>
>>>>
>>>
>>>---------------------------------------------------------------------
>>>To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>>>For additional commands, e-mail: users-help at gridengine.sunsource.net
>>>
>>>
>>>
>>
>>
>>
>>---------------------------------------------------------------------
>>To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>>For additional commands, e-mail: users-help at gridengine.sunsource.net
>>
> 
> 
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net
> 

-- 

Marco Donauer            Tel: +49 941 3075-211  (x60211)
Software Engineer        Fax: +49 941 3075-222  (x60222)
Sun Microsystems GmbH
Dr.-Leo-Ritter-Str. 7    mailto:marco.donauer at sun.com
D-93049 Regensburg       http://www.sun.com/gridware


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list