Log full when overwrite mode is on

[Former Member]

Hi,

as described in this thread I'm experiencing a log full message when overwrite mode is on. It seems that the db "pauses" when the message is shown because a loader task stops inserting and resumes when the message goes away.

I've seen this message with more than one db, one was about 10GB data space 512MB log space, the other 4GB data space, 128MB log space. Log queue is set to 2000 pages.

Here's a screenshot of Database studio : http://files.me.com/avettori/vmnnb8

MaxDB version 7.7.06.09 on Linux 64bit.

Thank you !!

[Former Member · Accepted solution]

Hi,

Overwrite log mode means that the database is writing log information onto the log volume but will overwrite it when the end of the circle is reached.

But the log can only be overwritten when there is loginformation which is allowed to overwrite.

When you have only open transactions on the log volume then it cannot be overwritten.

You told us that you are using the loader when this log full situation happens.

If you only load data you could switch OFF the log completely to avoid the log full situation.

Does this answer your question?

Regards, Christiane

[lbreddemann · Accepted solution]

GOSH!

Christiane is of course right - as usual

Basically this means for you: commit more often during the import.

Switching off the logwriter alltogether may be a bit risky, since than now automatic savepoints are written anymore.

But of course, this will speed up things as well a bit.

regards,

Lars

[Former Member · Accepted solution]

Dear Christiane, Dear Lars,

I understand

So when the log was full what happened ? Was my data loaded correctly ?

I'll try to check the parameters to find how to commit more frequently. Probably in "normal" use of the database this will not be a problem.

Thank you !!

[lbreddemann · Accepted solution]

Hi Andrea,

Sorry I've been a bit unprecise with my explanation.

The log of transaction become overwritable, when the changed data has been written to disk - that is, when a savepoint has occured.

With the massive load of data you generate a whole lot of redo information which filled up the log area. When a savepoint has been written, the new synchronization point between log and data is clear and the pointer in the log area can move on.

I really recommend to read the MaxDB internals course on this.

Regards,

Lars

[lbreddemann]

Hi Andreas,

I tried to reproduce this but didn't succeed. Maybe somebody else will !?

regards,

Lars