[index review] Index rebuilding/creation

[Emmanuel Lecharny <el...@apache.org>]

I have renamed the thread, because I think we need to discuss everything 
about index construction or reconstruction when the server already 
contains data.

Kiran Ayyagari wrote:
>
>
> Emmanuel Lecharny wrote:
>> Ok, let's go deep into each system index now. We don't really care 
>> about user index right now.
>>
>> First, we have to know that a system index can still be configured in 
>> the Server.xml file, because the user can set the index size. We need 
>> to clarify exactly what can be configured, what is the default and 
>> also discuss about what can be improved.
>>
>> 1) What can be configured
>>
>> Let's list the parameter we can work on :
>> - wkDirPath : the working directory path used to store the index
>
> <paranoia>
>  What if the user configures it to '/tmp' and reboots the machine 
> after the server has loaded some data.
>  All the indexed data would be lost. May be we should warn users about 
> it in the doc
> </paranoia>
We can't protect users against themselves... Someone deciding to put an 
index on /tmp either knows what he does or deserves to see his index 
being deleted on reboot ;)

>
> We should have some mechanism where by the server notifies the user 
> either to rebuild the index or
> the server itself does it.
I was thinking about the index rebuilding problem, or more probably, 
index building after addition of one index after data has been injected 
into the server. We need such a tool, the thing is how to implement it ? 
Is it possible to have a on-line index construction system ? (ie, a 
mechanism able to build an index while the server is running).

IMHO, I think we should offer a simple index creation system. It can be 
activated as a separate process, or probably better, as a startup option 
for the server, so we can invoke it through Studio.
>
> This should be easy as we already have a command line tool to rebuild 
> the index the only thing is to fix
> this partially broken tool (I have started to fix this tool sometime 
> back but it hasn't come out of
> workspace may be for 2.0 it will :) and add some code to the server to 
> notify the user.
That could be a good start.


-- 
--
cordialement, regards,
Emmanuel Lécharny
www.iktek.com
directory.apache.org