[Building Sakai] Can I Upgrade Search from 1.2.5 to 1.3.5 in Sakai 2.7

[David Horwitz]

Hi Mike,

yes you can delete all and restart but this is more drastic than you 
need to do. There will be a bit of load during a full index rebuild - 
but should be non drastic. I would recommend:

1) Check the status of your journals, are there any in merge-prepare? If 
so update them to match the others (probably 'committed' slightly 
misspelled)
2) Remove the old node names in search_node_status

You then should see one of the nodes locking the first 200 journals and 
starting the optimization operation. What you want to see is a steady 
decrease in the number of items in search_journal till its > 100 and 
then start to build up again till there are enough to optimize.  You 
don't need to worry about the node id in the search_journal table - this 
is only of historic interest once the node is written and will be 
changed when the journal is locked for optimization.

Regards

David




On 01/26/2012 10:06 PM, Mike Jennings wrote:
> I have been looking around on the net and found this response from David
> where he says describes how to rebuild the entire index...
>
> http://collab.sakaiproject.org/pipermail/production/2011-February/001215.html
>
> I was wondering if I did the following, then started the servers, and
> clicked on Rebuild Whole Index in production, would there be a huge hit
> in performance to the users while this action was taking place?
>
> Thanks,
>
> Mike
>
> On 1/26/2012 2:49 PM, Mike Jennings wrote:
>> Seth,
>>
>> The bad news is that there are rows in the search_journal that have
>> indexwriter values with the same names as the old server ids in the
>> search_node_status table.
>>
>> Mike
>>
>> On 1/26/2012 2:39 PM, Seth Theriault wrote:
>>> Mike Jennings wrote:
>>>
>>>> Can I just delete the entries for the servers that are not needed in the
>>>> search_node_status table?
>>> I have done this without incident. You may want to verify that none of
>>> the rows in the search_journals reference an old node in the indexwriter
>>> column.
>>>
>>> Seth
>>>