[GE users] Automated installation & BerkeleyDB

Marco Donauer Marco.Donauer at Sun.COM
Tue Apr 19 16:08:44 BST 2005


Alan,

thanks, I will check and change it!

Regards,
Marco

Alan Carriou wrote:

> Marco,
>
> okay then. By the way, README-Autoinstall.txt is wrong when it says, 
> in the paragraph 2., that the template configuration can be found in 
> $SGE_ROOT/$SGE_CELL/util/install_modules . It is in 
> $SGE_ROOT/util/install_modules .
>
> Best regards,
> Alan
>
> Marco Donauer wrote:
>
>> Alan,
>>
>> yes, that right. It should be local directory on berkeley db server.
>> This is a bug in update 3. It will be also fixed in update 4.
>> The berkeley db installation doesn't care about existing spooldb 
>> directories.
>> => the berkeley db server installation works well, but the qmaster 
>> installation finds
>> an existing spooldb directory with db files and stops.
>>
>> Regards,
>> Marco
>>
>> Alan Carriou wrote:
>>
>>> Marco,
>>>
>>> thank you for your answer.
>>>
>>> I think I had stopped the servers and completely deleted the 
>>> directory, but I'll give it another try.
>>>
>>> > The autoinstallation stops, if the entered <sge-cell> or the spooldb
>>> > directory already exists.
>>> > This is done to make sure, that already installed clusters can't be
>>> > destroyed.
>>> But doesn't the BerkeleyDB installation create those directories too 
>>> ? Or should I place the spooldb directory for the BerkeleyDB host 
>>> elsewhere ?
>>>
>>>
>>> Best regards,
>>> Alan
>>>
>>> Marco Donauer wrote:
>>>
>>>> Hi Alan,
>>>>
>>>> no you are on the right mailing list!
>>>>
>>>> Before you started the second (fresh) installation did you delete 
>>>> your old spooldb directory?
>>>> The autoinstallation stops, if the entered <sge-cell> or the 
>>>> spooldb directory already exists.
>>>> This is done to make sure, that already installed clusters can't be 
>>>> destroyed.
>>>>
>>>> Please before restarting a autoinstallation check and delete the 
>>>> spooldb directory and the cell directory.
>>>> This is the most common reason, why the scripts fails.
>>>>
>>>> Your approach using the autoinstallation was right.
>>>>
>>>> First install the berkeley db insteractively, then use the 
>>>> autoinstallation to install qmaster and execd's
>>>>
>>>> Within the next few days the update 4 will be released and provides 
>>>> a better autoinstallation.
>>>> The berkeley db installation will be automatic also.
>>>>
>>>> Best Regards,
>>>> Marco
>>>>
>>>> Alan Carriou wrote:
>>>>
>>>>> Hi everybody,
>>>>>
>>>>> (disclaimer: It's my first mail on this mailing-list, so tell me 
>>>>> if it's not the good one)
>>>>>
>>>>> I have a problem using the automated installation with a 
>>>>> BerkeleyDB for
>>>>> SGE 6.0u3.
>>>>>
>>>>> In our cluster, a NFS server, say host1, is set up, and exports 
>>>>> /sge-root/ to every computer. What we'd like to do is to use:
>>>>> - host1 as a Berkeley DB server,
>>>>> - another host, say host2, as the master host
>>>>> (and of course a few shadow and execution hosts)
>>>>> As the NFS is already a single point of failure, I guess 
>>>>> installing the BerkeleyDB on the same server won't be much harmful.
>>>>>
>>>>> When I try to install manually the master host, with "inst_sge -m" on
>>>>> host2, everything goes fine. I install the BerkeleyDB on host1 
>>>>> when it tells me that I can, and it completes without any problem. 
>>>>> Then I set up some execution hosts with the automated 
>>>>> installation, everything's fine.
>>>>>
>>>>> But, as there are quite a few computers that we're going to use in 
>>>>> the
>>>>> cluster, I'd prefer using the automated installation. So I tried 
>>>>> it on a fresh installation. As the manual tells me, I've first 
>>>>> installed the Berkeley DB on host1 with the interactive command 
>>>>> "inst_sge -db". I have put the spooldb directory in 
>>>>> /sge-root/testgrid/spooldb/. The server is alive, and seems to work.
>>>>>
>>>>> Then I wanted to use the automated procedure for the master and the
>>>>> execution hosts. So, I tried
>>>>>
>>>>> inst_sge -m -x -auto myfile.conf
>>>>>
>>>>> on host2. But the script fails with "The spooling directory 
>>>>> /sge-root/testgrid/spooldb/ already exists! Exiting 
>>>>> installation!". As the BerkeleyDB server has already been set up, 
>>>>> the directory indeed exists.
>>>>>
>>>>> So, either I am missing something, either there is a bug in the 
>>>>> automated installation. It seems that the installation of the 
>>>>> master host tries to create DB_SPOOLING_DIR locally, or at least 
>>>>> checks is presence, even if a BerkeleyDB is configured on a remote 
>>>>> host.
>>>>>
>>>>> Should I set the spooldb DB_SPOOLING_DIR to another location, not 
>>>>> shared on the NFS ? Or have I forgotten something ?
>>>>>
>>>>> Thanks,
>>>>> Alan
>>>>>
>
> ---------------------------------------------------------------------
> 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




More information about the gridengine-users mailing list