On Thu, 02 Feb 2006 18:11:05 -0500, "Art S. Kagel"
<kagel@bloomberg.net> wrote:
>cacunet@hotmail.com wrote:
>>>2+2 RAID10 I hope!
>>
>>
>>>OK, so the desired sequence of events is:
>>
>>
>>
>>>- BEGIN WORK;
>>>- SET PDQPRIORITY 100;
>>>- LOCK TABLE <tabname> IN EXCLUSIVE MODE;
>>>- DROP INDEX <indexname>;
>>>- ALTER TABLE <tabname> TYPE (RAW);
>>>- ALTER FRAGMENT ON TABLE tablename INIT FRAGMENT BY ...
>>>- ALTER TABLE <tablename> TYPE (STANDARD);
>>>- CREATE ... INDEX <indexname ON <tablename>...
>>>- COMMIT WORK;
>>
>>
>> Yes, it's RAID10.
>>
>> So this means that the table being reorg will not be accessible by the
>> users?
>> How long do you estimate this process will run? I understand that
>> there's a lot of different variables that can impact the run time, but
>> I just want a rough idea of how long it will take assuming the rest of
>> the chain in our system is on par with the rest of the world.
>
>Neil's right this is going to take time. If you absolutely cannot have this
>table offline for more than a brief period, then you'll have to copy the
>data to a new table in the new dbspace, create new matching indexes and
>constraints with different names, then when you can get exclusive access and
>keep the users off for a short time, rename the old table then rename the
>new table to the old name. Once that's done you can drop the old table or
>scan it for records added deleted and update since you made the copy.
>
Those deltas can be the hard part, though . . . .
Would this be a good feature request . . . . online reorgs??
JWC