II_RCP open/close

> I've got two identical installations of INGRES 6.4/05 running on two
> identical Silicon Graphics workstations running IRIX 5.3.  Databases
> on both installations are not too large, but they do keep me busy!
> 
> Some general information first:
> Database A (on machine A) has many tables, many users and is about 350MB
> in size.
> Database B (on machine B) has 2 tables, many users and is about 200MB in
> size.
> Transaction log files for both installations are a little less than
> 150MB in size.
> 
> The problem:
> The II_ACP.LOG and II_RCP.LOG files for Database B are HUGE and growing
> by leaps and bounds!  Yesterday they were both around 350MB in size and
> I had to delete them.  Today, they are back up around 8MB--after only
> 24 hours in operation!
> Please note that, despite the documentation, the II_ACP.LOG and II_RCP.LOG
> files *are not* being overwritten after each reboot of INGRES.  For
> example, the files for Database A still contain records from November
> 1995!  Therefore to manage these files I must delete them manually.
> 
> My questions:
> What could be causing the log files for Database B to grow so large so
> quickly?  Is there some flag I need to set?
> What kinds of tuning do I need to do to change how often lines are 
> written to these log files?  
> Does it have to do with the number and frequency of consistency points
> that are being taken?  
> Does it have to do with journaling?  Both databases are 
> journaled--Database A has many more tables than Database B.  Wouldn't that
> mean Database A would be the one with so much log activity?
> 
> It's my impression, although I could be wrong (haven't run any metrics
> lately), that Database A has much more activity than Database B.  But
> transactions on Database B do tend to be much larger in scope than those
> on Database A.
> 
> Am I making any sense here?  I have a call in to Tech Support but I 
> thought I'd check with you all as well.  I suspect this could be an
> interesting tuning exercise and want to know what you all think.
> 
> Thanks in advance,
> Karen Robinson
> 
> P.S.  I can send examples of the contents of the log files to anyone
> interested.  Basically there is nothing out of the ordinary that I can
> see, except for the fact that Database B has approximately one dozen
> records per second being written to it at any given time.
> 
> P.S.S.  In my initial call to INGRES Tech Support, I was told that the
> problem isn't related to usage or configuration of the server.  I then
> asked, "But what else could it be?"  She then decided to have someone
> call me back about it tomorrow...
> 
> 
> -- 
> Karen Robinson
> Database Administrator
> National Snow and Ice Data Center
> University of Colorado at Boulder
> 303.492.1504
> robinson@shiver.colorado.edu



HI Karen,

    Check how often you are opening and closing databases. Each time you do
    this another entry is written into the II_RCP.LOG. Plus you get a PURGEDB
    on the database close that causes a consistency point to be taken and
    hence the Archiver will come up (Depends upon the setting of No. of CP's
    before invoking archiver). Once the Archiver is up you get more stuff in
    II_ACP.LOG

    You should also compare the output of showrcp from both platforms. Look at
    CP interval and No. of CP's before invoking archiver.

    If the databases are being constantly opened and closed then you can
    prevent this by simply having a session hold a permanent connection to the
    databases of interest (inc. iidbdb). Depending upon levels of usage such a
    permanent hold can help performance connecting to the databases. NB. these
    permanent connections should only be in place during business hours -
    they'll screw up any attempt to sysmod a database for instance. You can do
    them from a cron entry Shell/PERL script for convenience.

    Martin Bowes.
Ingres Q & A
Back to William's Home Page

© William Yuan 2000

Email William