Re: Is there a problem with NOT ending journaling before doing a reorg?

Comments inline.

Pete Massiello wrote:
>
> When I said at least two, that was because when you move a
> record from one Relative Record to another, you will have a
> delete of the current record and then a put (write) of the
> record into the new location. That is two.

Yep. I concur. For every row that is /moved/ by the reorganize request. Yes, at least two log entries -- one entry for the delete of an active row and one entry for the (re)insert of that previously active row in place of a deleted row.

> Pete Massiello wrote:
>> <<SNIP>> If you use a Reorg while Active, then each record
>> as it is moved to a new Relative Record would be written to
>> the journal,

This part of the original response is /the same/ and so again, I am in total agreement.

>> and this would be an HUGE amount of entries written to the journal.

And that is where the statement goes astray. At this point there is an implication having used /would/ instead of /could/ for the number of entries that would transpire. There could in fact be possibly, nearly zero entries written to the journal. I am not really _that_ picky; I read on. And then...

>> Basically at least two for every record in the file.
>> So, if you had a million records, you would have at
>> least 2 million entries written to your journal.

Patently false as written. Sorry, but here a lack of having repeated explicitly a reference to /moved records/ had me feeling obligated to reply. These two statements are very deceptive. IMO most everyone will read them for what is stated, not for what was the intended implication.

I felt it necessary to clarify that an online reorganize does not mean conclusively, that any 1M row file will have at least 2M logged entries. So I was trying to show by an example that the 2X number of log entries, for the 1X number of rows, does not always apply; that the results are dependent on the layout of the deleted rows within the dataspace. As a stated result for a file, the 2X is effectively a /worst case scenario/ that few reorganize requests would ever exhibit.

> What would happen if you were also journaling access paths of
> the logical files that were over that physical file. Would
> there not be another update to the for the LF, therefore the
> statement at least 2.

Very worthwhile to suggest there may be more than 2X per /moved row/. There are even more log entries, like for periodic commits.

Regards, Chuck