5 Replies Latest reply: Oct 8, 2013 7:49 AM by Jonathan Lewis RSS

    Redo Log Writer Problem

    Street

      Hello

       

      What can I do, when Average redo log write time is 17'300ms (averaged over 30ms)?

      I have only one redolog writer. Should I start more then one redolog writer? We have 3 Redolog Groups (64MB) and we work with oracle dataguard. Its Oracle 11.2 (Unix Solaris 10).

      The system switch every 45 min the redogroup.

       

      Thanks for your support...

       

      Best regards...

       

      Roger

        • 1. Re: Redo Log Writer Problem
          Aman....

          Street wrote:

           

          Hello

           

          What can I do, when Average redo log write time is 17'300ms (averaged over 30ms)?

          I have only one redolog writer. Should I start more then one redolog writer? We have 3 Redolog Groups (64MB) and we work with oracle dataguard. Its Oracle 11.2 (Unix Solaris 10).

          The system switch every 45 min the redogroup.

           

          Thanks for your support...

           

          Best regards...

           

          Roger

          Why do you think that this time, 30ms is not good enough for your database ? Did you get any redo log related issues in the Statspack/AWR report?

           

          There is only one LGWR possible, you can't have more than one LGWR.

           

          Aman....

          • 2. Re: Redo Log Writer Problem
            Street

            The question is, why I have 17'300ms redo log write time? I think this is a very high value.

            What can I do? We have new oracle solaris T4 machines and a new disk san.

            • 3. Re: Redo Log Writer Problem
              Aman....

              What does the Statspack says? Any event is reported there? The fact is, there is very less that you can do. Your options are, generate less redo or use faster disks. The first option is not really possible if you are a heavy OLTP db and the 2nd one you need to check by yourself. A bad configuration may lead to more time spent for the change vector writing, for example using RAID-5 . So don't just think that the time is high(or low) . As long as there is no performance issue reported due to it , which can be really measured, you are good. Else, you can go for Exadata and use the Flash for the storage of ORL's or in a normal machine, use SSD.

               

              Aman....

              • 4. Re: Redo Log Writer Problem
                EdStevens

                Street wrote:

                 

                The question is, why I have 17'300ms redo log write time? I think this is a very high value.

                The question is why you "think this is a very high value."

                 

                What can I do? We have new oracle solaris T4 machines and a new disk san.

                 

                At this point it hasn't be determined you need to do anything at all.

                • 5. Re: Redo Log Writer Problem
                  Jonathan Lewis

                  Street wrote:

                   

                   

                  What can I do, when Average redo log write time is 17'300ms (averaged over 30ms)?

                   

                   

                  I can't make sense of your figures - is this 17.3 seconds of "log file parallel write" time over 30 minutes ?

                  Or is it 17,300 milliseconds of time with an average time of 30 milliseconds ?

                   

                  Whatever it is, information about the volume of redo written in the time is also relevant.

                   

                  Regards

                  Jonathan Lewis