On Mon, Jan 18, 2010 at 5:07 PM, Terry <td3201@gmail.com> wrote:
> On Mon, Jan 18, 2010 at 4:48 PM, Andy Colson <andy@squeakycode.net> wrote:
>> On 1/18/2010 4:08 PM, Terry wrote:
>>>
>>> Hello,
>>>
>>> Sorry for the poor subject. Not sure how to describe what I need
>>> here. I have an application that logs to a single table in pgsql.
>>> In order for me to get into our log management, I need to dump it out
>>> to a file on a periodic basis to get new logs. I am not sure how to
>>> tackle this. I thought about doing a date calculation and just
>>> grabbing the previous 6 hours of logs and writing that to a new log
>>> file and setting up a rotation like that. Unfortunately, the log
>>> management solution can't go into pgsql directly. Thoughts?
>>>
>>> Thanks!
>>>
>>
>> How about a flag in the db, like: dumped.
>>
>> inside one transactions you'd be safe doing:
>>
>> begin
>> SET TRANSACTION ISOLATION LEVEL REPEATABLE READ;
>> select * from log where dumped = 0;
>> -- app code to format/write/etc
>> update log set dumped = 1 where dumped = 0;
>> commit;
>>
>> Even if other transactions insert new records, you're existing transaction
>> wont see them, and the update wont touch them.
>>
>> -Andy
>>
>
> I like your thinking but I shouldn't add a new column to this
> database. It's a 3rd party application.
>
Although. I really like your idea so I might create another table
where I will log whether the data has been dumped or not. I just need
to come up with a query to check this with the other table.
