what is holding the log in this session
Floyd ran a 24+ hour batch job (ESQL/C) against a non-logged IDS 10 database that kept aborting with a long transaction error, even though other sessions were filling the logs, and he wanted to know what was holding the oldest log open and which onstat/onlog output could identify it. Replies suggested the culprit was logged DDL/metadata activity — index creation and extent additions, which are logged even in a non-logged database — and recommended setting proper initial/next extent sizes and using onstat -lx to find the log where the transaction began. No confirmed resolution is recorded in the thread.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
I have a session that is running stuff from ec code in an ids10 non logged database. I know that at the beginning of the program, some indexes are created, and lots of inserts happen to a table.
This is a batch process that takes over 24 hours to run. It has continued to abort with long transaction.
So I have added more logs and am watching it. It is not actually filling the logs, other processes are, but what I want to know is, what thing is causing a log to act like this transaction is going on, that keeps this whole process tied to that log ?
Are there any onstats I can do to look in the logs or anyway find out what is happening here ?
I really can't make heads or tails out of onlog to be able to tie it to a sid or see if it's an open transaction etc..
Thanks for your advice.
Floyd
Floyd Wellershaus wrote:
> I have a session that is running stuff from ec code in an ids10 non
> logged database. I know that at the beginning of the program, some
> indexes are created, and lots of inserts happen to a table.
> This is a batch process that takes over 24 hours to run. It has
> continued to abort with long transaction.
>
> So I have added more logs and am watching it. It is not actually filling
> the logs, other processes are, but what I want to know is, what thing is
> causing a log to act like this transaction is going on, that keeps this
> whole process tied to that log ?
>
> Are there any onstats I can do to look in the logs or anyway find out
> what is happening here ?
> I really can't make heads or tails out of onlog to be able to tie it to
> a sid or see if it's an open transaction etc..
>
> Thanks for your advice.
>
> Floyd
>
DDL?
For example extent additions have to be logged.
wild guess... hmmm could be that it is extending extents of a table/
index....
only thing i can think of is make sure that the initial/next extent
size is properly set...
you may have to do a onstat -lx and search for the log where the trx
began.
Superboer
way fast=http://www.clipjes.nl/clip/nederlands/n/normaal_-
_oerend_hard.html
On 3 mrt, 13:57, "Floyd Wellershaus" <fl...@fwellers.com> wrote:
> I have a session that is running stuff from ec code in an ids10 non logged database. I know that at the beginning of the program, some indexes are created, and lots of inserts happen to a table.
> This is a batch process that takes over 24 hours to run. It has continued to abort with long transaction.
>
> So I have added more logs and am watching it. It is not actually filling the logs, other processes are, but what I want to know is, what thing is causing a log to act like this transaction is going on, that keeps this whole process tied to that log ?
>
> Are there any onstats I can do to look in the logs or anyway find out what is happening here ?
> I really can't make heads or tails out of onlog to be able to tie it to a sid or see if it's an open transaction etc..
>
> Thanks for your advice.
>
> Floyd