Alter statements, Checkpoint Duration, Log File

[Former Member]

On IQ 15.4 with Solaris 10, Checkpoint Duration is exceeding 15 minut

Does anyone know of a way to improve (reduce) duration of a checkpoint? I've truncated the transaction log, and started a new iqmsg log, but still they are taking long.

Each alter causes a checkpoint, correct?  Is there a way to have a single checkpoint after a series of alters to reduce the amount of checkpointing occuring with alters?

[jong-kil_park]

Hi Steve,

IQ server runs the checkpoint whenever every single alter command is completed.

And, a set of the "alter " commands are not be able to be run as a single transaction.

Even though you run them within "begin tran ... commit", IQ runs the alter command one by one followed by checkpoint.

This behavior has not been changed in IQ 16.0, too.

There are some internal steps for completing the checkpoint operation.

A step out of the steps is to construct RF(Roll Forward)/RB(Roll Back) for ACTIVE transactions and write it to the checkpoint log.

I believe IQ server spends the most of the time for checkpoint to complete the above step.

The size of the transaction log has nothing to do with the elapse time for checkpoint.

So, I think you had better run it based on schedule when there are not many running ACTIVE transactions.

Best Regards

Jerry

[tayeb_hadjou]

Hi,

Have you changed a parameter prior to this issue?

What is value for -gc server option ?

You may need have a look in OS errolog and Iqmsg.

Disk I/O performance could be low.

You said you truncated the log. You mean IQ transaction log (.log) ?