Re: [Info-Ingres] Checkpoint journaling question

On Jun 26, 2008, at 3:41 PM, Mike Ossing wrote:

> Hello,
>
> I hadn't checked up on the checkpointing this week, until today,
> when I found I have several journal files, from the last few days
> in my journal folder for the ERP application Database. I also have a
> couple of days of checkpoint folders.
>
> I run a nightly checkpoint with ckpdb +j -d dbname. In the past,
> this has always taken the current ckp, then deleted the old.

The journal files contain a running log of updates that have occurred
since a checkpoint.
Having extra journal files around is not necessarily an indication of
trouble.

I'd do an "infodb databasename" in a command prompt window and
examine the section that
is titled something like "checkpoint history for journal". It should
show you the
last checkpoint (just one if you're doing -d), and the valid flag
should be 1.
If valid is 0, something is going wrong during the checkpoints and
you need
to find out what. (presumably the ckpdb output says what.)

If there's no problems, look at the first_jnl and last_jnl sequence
numbers
and match them up with your leftover journal files. If your leftover
files are older, that just means that somehow they got missed and you
can just delete them by hand. (If your journals are in the sequence
listed,
then they are current and you should just leave them alone.)

By the way, I don't care for the -d option to ckpdb myself. Only
having one
preceding checkpoint in the database checkpoint history limits your
recovery options when things go pear-shaped. It's better to not use -d,
and manage disk space by deleting old checkpoints and journals yourself.
A middle way is to build up some safe number of checkpoints, and then
start
using alterdb -delete_oldest_ckp which deletes the oldest checkpoint.
This is more automatic but you do have to keep an eye on it to make
sure that it doesn't go haywire and leave you with insufficient
disk space.

Karl