Ok. I understand how it works. Thank you very much!!.
Best regards, Diego Martínez
El 8 de febrero de 2011 09:01, Frazer Clement
<frazer.clement@stripped>escribió:
> Hi Diego,
> Ndb Cluster has row level locking. Transactions take row locks before
> updating or deleting existing rows. Two transactions attempting to update
> the same row will be serialised by the system. The second transaction will
> see the effects of the first. The final effect will be as if the
> transactions were run in some serial order.
>
> I am not sure how / if Hibernate uses table locks directly, or just
> expects that the DBMS 'somehow' avoids races between different users. If it
> uses Table locks directly, and is using them for some purpose other than
> serialising updates to the database, then this may not work with Cluster.
> If it just expects that the DBMS present a consistent view of the database,
> and serialises competing transactions, then this can work fine with Cluster.
>
> Perhaps if you can explain how/why Hibernate uses Table locks, then we can
> determine whether or not it will work unmodified with Cluster.
> The 'Single User Mode' which Wagner Bianchi mentions is probably not what
> you want to use, except on very rare occasions (e.g. DBA maintenance windows
> etc.).
>
> Frazer
>
>
> Diego Martínez wrote:
>
>> Thanks for your response. My question are what happens when several
>> threads
>> update the same tables. For example, if I have a web service application
>> where multiple threads are updating the product table and no exist table
>> locking, the result would be completely unexpected. Each thread would
>> update
>> the same table, but in a different node?. What would be the end result?.
>>
>> Best regards, Diego Martínez
>>
>>
>> El 7 de febrero de 2011 12:52, Wagner Bianchi
>> <wagnerbianchijr@stripped>escribió:
>>
>> Hi Martinez,
>>>
>>> I understand the unique and efficient way to lock tables on MySQL Cluster
>>> SQL NODES is using "SINGLE-USER MODE". LOCK TABLES [WRITE | READ] will
>>> work
>>> as expected, but, using only one SQL NODE. You'll LOCK TABLE t1 TO WRITE
>>> on
>>> SQL NODE 1, but your application will can access the same table through
>>> the
>>> SQL NODE 2 and write to that table.
>>>
>>> Let us know more about your project and perhaps we can help you.
>>>
>>> Best regards.
>>> --
>>> Wagner Bianchi
>>>
>>> 2011/2/7 Diego Martínez <tabarem@stripped>
>>>
>>> Hi,
>>>
>>>> Members of the list have experience using MySQL Cluster through
>>>> Hibernate?
>>>> I
>>>> am worried, because I understand that MySQL Cluster does not support
>>>> locking
>>>> tables for example.
>>>>
>>>> Do you know design patterns to use this new engine?
>>>>
>>>> Best regards, Diego Martínez
>>>>
>>>>
>>>
>>
>
> --
> Frazer Clement, Senior Software Engineer,
> MySQL Cluster / Oracle - www.mysql.com
> Office: Edinburgh, UK
>
> Are you MySQL certified? www.mysql.com/certification
>
