This site is currently read-only as we are migrating to Oracle Forums for an improved community experience. You will not be able to initiate activity until January 31st, when you will be able to use this site as normal.

    Forum Stats

  • 3,890,592 Users
  • 2,269,776 Discussions
  • 7,916,825 Comments

Discussions

Recommended Redo Log size

13»

Answers

  • Srini Ramaswamy
    Srini Ramaswamy Member Posts: 68 Blue Ribbon
    edited May 9, 2016 9:46AM

    May be 1PM to 2PM.

    Starting backwards where 23 is assumed to be 11 PM to 12 AM.

    T.

    Srini

  • Jonathan Lewis
    Jonathan Lewis Member Posts: 10,142 Blue Diamond
    edited May 9, 2016 10:01AM

    Wrong !

    But thanks for co-operating.

    It's a reasonable guess, of course - clearly the system must have seen a higher average rate of change over that hour (and the following one) - but that doesn't mean the system was under stress at any particular time.  The only thing this report can tell you is that the pattern isn't the same as usual - and when the pattern hasn't changed you might have a problem, and when the pattern has changed you might not have a problem.

    The point at which the most significant redo waits appeared was from 20:58 to 21:02 when there were 10 log file switches in just over 3 minutes. As Cary Millsap puts it: you can't derive detail from a summary.

    Regards

    Jonathan Lewis

    Srini Ramaswamy
  • John Stegeman
    John Stegeman Member Posts: 24,269 Blue Diamond
    edited May 9, 2016 9:54AM

    I knew I would be wrong, Jonathan - I was wincing awaiting the inevitable rebuke

  • 2995489
    2995489 Member Posts: 114
    edited May 9, 2016 10:04AM

    Hi,

    yes i used the query ( V$ARCHIVED_LOG ). I find mininum 6 switschs/hour all the day.

  • 2995489
    2995489 Member Posts: 114
    edited May 9, 2016 10:10AM

    Hi,

    i have said i must because i noticed locks sometimes in the DB. and when i have checked (diagnostic) i have found that requests wait the LGWR process.

    i have understand that it wait the switch to can write in the next redo log file..

    Med.

  • EdStevens
    EdStevens Member Posts: 28,778 Gold Crown
    edited May 9, 2016 10:16AM
    2995489 wrote:
    
    Hi,
    
    yes i used the query ( V$ARCHIVED_LOG ). I find mininum 6 switschs/hour all the day.
    
    

    But you still haven't demonstrated that log switches -- regardless of how frequently they occur -- actually causes a problem in your business.  The key is NOT the rate at which log switches occur, the key is measured wait events related to log switches.  Take some meaningful statspack reports and see how often those wait events appear in the 'top 5' list.  And even then you need to apply a bit of reason, as there will always be a 'top 5' events.

    How much effort is justified to reduce the average response time of an OLTP transaction by 90%?   (Really?  what if the averages response time is already 0.05 seconds?)

    How much effort is justified to reduce the run time of a nightly batch job from 5 minutes to 2 minutes?

    How much effort is justified to reduce the run time of a nightly batch job from 2 hours to 30 minutes, if no other process is found to ever be waiting on completion of the batch job?

    AndrewSayer
  • Jonathan Lewis
    Jonathan Lewis Member Posts: 10,142 Blue Diamond
    edited May 9, 2016 12:12PM

    Which types of locks did you notice, how much time did they spend waiting, and what diagnostic did you use to find them and then to decide that the processes were waiting for LGWR ?

    Regards

    Jonathan Lewis

This discussion has been closed.