[GE users] two cluster with the same reporting

jana jana.olivova at sun.com
Fri Nov 28 14:47:40 GMT 2008


    [ The following text is in the "UTF-8" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some characters may be displayed incorrectly. ]

Hi Sergio,

So, just to make it clear you cannot have multi-cluster support unless  
you have at least the 6.2 reporting installed. The reporting is just a 
front-end to the database, and it is separate from the cluster, you only 
have to unpack the binaries in the sge_root directory. So technically, 
if you would download the 6.2 ARCo binaries and just reinstall the 
reporting (not the dbwriters) you could have multi-cluster support now, 
with the limitations that I had outlined before. But of course is better 
to just upgrade everything.

Jana

sergio diaz wrote:
> Hi Jana,
>
> We have two clusters with 6.1u4. In one of them I have the dbwriter 
> working and I want install the dbwriter in the other cluster.
> A part of these clusters, I have a machine in which I installed sge 
> (only qmaster) and then I installed the reporting and webconsole. 
> Usually, the tomcat overload the machine too much and I wouldn't want 
> that this affect to performance of the front-end of the cluster.
>
> Then I have to wait for our migration to reinstall the dbwriters and 
> reporting to avoid possibles conflicts or fails.
>
> Thanks for all,
> Regards
>
> Sergio
>
>
> jana escribió:
>   
>> Hi Sergio,
>>
>> Let me first get clear if I understand what  are you saying. You have 
>> one cluster that is 6.1 and the second that is 6.2?
>>
>> If you have the 6.2 reporting sources you can easily add the database 
>> connection parameters from clusters that use 6.1 dbwriter and have the 
>> 6.1 database schema structure.  Since in the 6.2 reporting  the 
>> predefined-queries  are for  6.2 some of them will not work on 6.1 
>> database. For example none of the AdvanceReservation queries will work 
>> since there are no  ar tables in the 6.1 database schema, some others 
>> might not work either, because some unused fields were dropped in 6.2 
>> and thus you will get SQL Exception when trying to reference 
>> non-existent field.
>>
>> There were also some changes to the actual queries to accommodate for 
>> other changes that were made in the table structures or the data's 
>> that were reported, they might not throw Exception but they might just 
>> not provide the most correct results.  You can of course copy the 6.1 
>> xml query files from 6.1 reporting spool directory to the spool 
>> directory of 6.2 reporting and renaming them so you know these are to 
>> be used on 6.1 cluster.
>>
>> Do you have 6.1 version or 6.1u4 and >?
>>
>> Jana
>>
>> sergio diaz wrote:
>>     
>>> Hi Jana,
>>>
>>> I have just remember that I have the reporting in other machine 
>>> different that these two cluster. If I have the two cluster with sge 6.1 
>>> and I have sge 6.2 in the machine with the reporting, Can I have 
>>> multi-cluster support, can't I?
>>>
>>> Regards,
>>> Sergio
>>>
>>> jana escribió:
>>>   
>>>       
>>>> Hi Sergio,
>>>>
>>>> Unfortunately, no you cannot do this with 6.1. The multi-cluster 
>>>> support is a feature of 6.2. In 6.1 you must have one reporting for 
>>>> each dbwriter, there is no support in the 6.1 reporting for  multiple 
>>>> clusters, it is able to handle only one.
>>>>
>>>> When you migrate to 6.2 you will have to reinstall both dbwriters and 
>>>> reporting. Please refer to  the Upgrade procedure in the manual and 
>>>> don't forget to backup your databases. :)
>>>>
>>>> Regards,
>>>>
>>>> Jana Olivova
>>>>
>>>> sergio diaz wrote:
>>>>     
>>>>         
>>>>> Thank very much Jana.
>>>>>
>>>>> I'm using 6.1 but in a few days we are going to migrate to 6.2. Can I do 
>>>>> this with 6.1, can't I? And when the cluster are migrated, we will have 
>>>>> to reinstall the both dbwriter?
>>>>>
>>>>> Regards,
>>>>> Sergio
>>>>>
>>>>> jana escribió:
>>>>>   
>>>>>       
>>>>>           
>>>>>> Hi Sergio,
>>>>>>
>>>>>> I am assuming you are using 6.2, aren't you? Just reinstall the 
>>>>>> reporting  and enter both of the databases parameters. After you enter 
>>>>>> the first parameters you will be asked:
>>>>>>
>>>>>>  Do you want to add another cluster? (y/n) [n] >>
>>>>>>
>>>>>> Editing the config.xml directly  would not work, since also the 
>>>>>> password has to be generated.
>>>>>>
>>>>>> To your other part of a question, if you are referring to 
>>>>>> cross-cluster queries yes you onbly have to use fully qualified names 
>>>>>> if you want to combine output from the tables from 2 different 
>>>>>> databases (database schemas). Generally, you will just want to run the 
>>>>>> existing queries on a specified cluster.
>>>>>>
>>>>>> Regards,
>>>>>>
>>>>>> Jana Olivova
>>>>>>
>>>>>> sergio diaz wrote:
>>>>>>     
>>>>>>         
>>>>>>             
>>>>>>> Hi all,
>>>>>>>
>>>>>>> I want install other dbwriter in other cluster using the same reporting 
>>>>>>> and also the same webconsole.
>>>>>>> If I create other database for the new cluster, the webconsole won't see 
>>>>>>> this database because the reporting was set with the other database. 
>>>>>>> Could I add to the file /gridware/sge/default/arco/reporting/config.xml 
>>>>>>> two configuration of database?
>>>>>>>
>>>>>>> I.E. I have the following configuration and we could add the same only 
>>>>>>> changed <database name="arco" host="machine.arco" port="3306" 
>>>>>>> schema="arco"> for <database name="arco1" host="machine.arco" 
>>>>>>> port="3306" schema="arco1">
>>>>>>>
>>>>>>> <configuration>
>>>>>>>   <!--
>>>>>>>     Configure the database connection to be used by the application
>>>>>>>   -->
>>>>>>>   <database name="arco" host="machine.arco" port="3306" schema="arco">
>>>>>>>      <driver type="mysql">
>>>>>>>         <javaClass>com.mysql.jdbc.Driver</javaClass>
>>>>>>>      </driver>
>>>>>>>      <user name="arco_read" passwd="XXXXXXX" maxConnections="10"/>
>>>>>>>   </database>
>>>>>>>             <applUser>default</applUser>
>>>>>>>   <storage>
>>>>>>>     <root>/var/spool/arco</root>
>>>>>>>     <queries>queries</queries>
>>>>>>>     <results>results</results>
>>>>>>>   </storage>
>>>>>>> </configuration>
>>>>>>>
>>>>>>>
>>>>>>> If I read the following paragraph in the web page of arco, I guess that 
>>>>>>> it isn't necessary and only If I want do a query using the two 
>>>>>>> databases, I have to do for example "select arco.host, arco1.host from 
>>>>>>> arco, arco1.sge_job;"
>>>>>>> Would be correct this suppose?
>>>>>>>
>>>>>>>   
>>>>>>>       
>>>>>>>           
>>>>>>>               
>>>>>>>> If you configure MySQL multiple databases using just one pair of users 
>>>>>>>> (arco_write, arco_read) and grant the privileges as described in step 
>>>>>>>> 7 of How to Configure the ARCo Database on MySQL 
>>>>>>>> <http://wikis.sun.com/display/GridEngine/Installing+ARCo#InstallingARCo-mysqlgrant>, 
>>>>>>>> you can perform cross_cluster queries. You must use the 
>>>>>>>> fully-qualified names when accessing objects, for example 
>>>>>>>> /database_name/./table_name/.
>>>>>>>>     
>>>>>>>>         
>>>>>>>>             
>>>>>>>>                 
>>>>>>> Many Thanks,
>>>>>>> Regards
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>   
>>>>>>>       
>>>>>>>           
>>>>>>>               
>>>>>   
>>>>>       
>>>>>           
>>>   
>>>       
>
>

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=90259

To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].



More information about the gridengine-users mailing list