9 Replies Latest reply on Jan 16, 2017 4:25 PM by mdtaylor

    ebs upgrade downtime estimate

    user13109564

      Hello,

       

       

      Need your help is approx. downtime requirement for below upgrade.

       

       

      Upgrade Requirement : EBS 12.1.3 to 12.2.5

      Database Size : 4+ TB

      Server : Linux 5 64 bit (Virtual Machine)

      Database : 11.2.0.4

      NFS protocol for shared appl top

      NFS server ( 2 CPU and 2 GB RAM )

       

       

       

       

      Middle Tier ( node 1 )

      - CPU 20 ( Virtual CPU )

      - RAM 96 GB

       

       

      Middle Tier ( node 2 )

      - CPU 8 ( Virtual CPU )

      - RAM 48 GB

       

       

      Middle Tier ( node 3 )

      - CPU 8 ( Virtual CPU )

      - RAM 48 GB

       

       

      External node ( node 4)

      - CPU 8 ( Virtual CPU )

      - RAM 48 GB

       

       

      Database Tier

      - CPU 30 ( Virtual CPU )

      - RAM 144 GB

       

       

      Upgrade Steps :

      1) Install database patches from ETCC

      2) File system install

      3) Patches before 12.2.0.

      4) 12.2.0 patch

      5) Pre-steps/checks before adop enablement

      6) Adop enablement patch

      7) Post Step/checks after adop enablement

      8) AD and TXK patches then install a custom top

      6) Patches before 12.2.5

      7) 12.2.5 patch then add custom top

      8) Shared appl top configration on node 2

      9) Shared appl top configuration on node 3

      10) External appl top configuration on DMZ node

      11) Wls managed server failover testing

        • 1. Re: ebs upgrade downtime estimate
          mdtaylor

          It is completely dependent on the speed of your storage.  If it is 50 MB/s storage in the VM, it will take well over a hundred hours.  If you have over 1000 MB/s storage in your VM, you can maybe do it in 50-80 hours depending on how many times you have done it before in test instances and how well you have documented the process for yourself.  Steps 2 and 3 can be done before the downtime, and you can do the DB ETCC patches some weekend before the upgrade too if you want.

          • 2. Re: ebs upgrade downtime estimate
            Srini Chavali-Oracle

            It also depends on what modules you are using, what what functionality within those modules.

             

            The only way to get an accurate estimate is to practice the upgrade on identical hardware to production.

            • 3. Re: ebs upgrade downtime estimate
              VishnuVinnakota

              I would recommend a downtime of around 2-3 Days. This might vary based on the Post-Upgrade Patches, Workers you use during patches/upgrade, DB SGA etc.

               

              Have you done any iterations till now? If not, start a TEST iteration for the estimate which is the only way to find the estimate for your environment.

               

              I have done several 12.2.5 upgrades and ideally if you plan well, should not take more than 48 hours even for 4TB+ DB with major modules implementation.

               

              Thanks,

              Vishnu

              • 4. Re: ebs upgrade downtime estimate
                user13109564

                Thanks for Reply.

                Production hardware is similar to this.

                All on VM and NFS for shared appl top.

                 

                Just want to check if there are any additional DB parameters , OS parameters or NFS parameters to reduce the overall process time.

                • 5. Re: ebs upgrade downtime estimate
                  user13109564

                  Hi Vishu,

                   

                  Thanks for sharing.

                  Please let me know even in Virtual Server ( VM ) . it will be completed in 48 hours ?

                  Any purging etc...is done before upgrade ?

                  Please share if additional DB parameters , OS parameters or NFS parameters to reduce the overall process time.

                  • 7. Re: ebs upgrade downtime estimate
                    user13109564

                    already reviewed

                    • 8. Re: ebs upgrade downtime estimate
                      user13109564

                      how to find the storage speed

                      • 9. Re: ebs upgrade downtime estimate
                        mdtaylor

                        Here is what I get on my Thinkpad with a 1TB 500 MB/s Samsung SSD:

                         

                        plato:~ # df -h

                         

                        Filesystem      Size  Used Avail Use% Mounted on

                        rootfs           14G  7.9G  5.4G  60% /

                        devtmpfs         16G   16K   16G   1% /dev

                        tmpfs            16G  175M   16G   2% /dev/shm

                        tmpfs            16G  780K   16G   1% /run

                        /dev/sdb6        14G  7.9G  5.4G  60% /

                        tmpfs            16G     0   16G   0% /sys/fs/cgroup

                        tmpfs            16G     0   16G   0% /media

                        tmpfs            16G  780K   16G   1% /var/lock

                        tmpfs            16G  780K   16G   1% /var/run

                        /dev/sdb5        20G   16G  4.7G  77% /lvw01

                        /dev/sda4       587G  511G   76G  88% /lv01

                        /dev/sdb7        11G  5.1G  5.0G  51% /home

                        /dev/sdb1       380M  158M  203M  44% /boot

                        /dev/sdb2        11G  4.3G  5.8G  43% /mnt/home

                         

                        plato:~ # hdparm -t --direct /dev/sda4

                         

                        /dev/sda4:

                        Timing O_DIRECT disk reads: 1500 MB in  3.00 seconds = 499.43 MB/sec

                        plato:~ # hdparm -t --direct /dev/sda4

                         

                        /dev/sda4:

                        Timing O_DIRECT disk reads: 1498 MB in  3.00 seconds = 499.18 MB/sec

                        plato:~ # hdparm -t --direct /dev/sda4

                         

                        /dev/sda4:

                        Timing O_DIRECT disk reads: 1498 MB in  3.00 seconds = 499.33 MB/sec

                        plato:~ #

                         

                        ensure filsystemio_options=setall

                         

                        SQL> SET SERVEROUTPUT ON

                         

                        DECLARE

                              l_latency  PLS_INTEGER;

                              l_iops     PLS_INTEGER;

                              l_mbps     PLS_INTEGER;

                            BEGIN

                               DBMS_RESOURCE_MANAGER.calibrate_io (num_physical_disks => 20,

                                                                   max_latency        => 20,

                                                                   max_iops           => l_iops,

                                                                   max_mbps           => l_mbps,

                                                                  actual_latency     => l_latency);

                         

                             DBMS_OUTPUT.put_line('Max IOPS = ' || l_iops);

                             DBMS_OUTPUT.put_line('Max MBPS = ' || l_mbps);

                             DBMS_OUTPUT.put_line('Latency  = ' || l_latency);

                           END;

                           /

                        Max IOPS = 33575

                        Max MBPS = 526

                        Latency  = 5

                         

                        PL/SQL procedure successfully completed.

                         

                        host:/lv01 # dd if=/dev/zero of=/lv01/test1 bs=2G count=1 oflag=dsync

                        0+1 records in

                        0+1 records out

                        2147479552 bytes (2.1 GB) copied, 4.15402 s, 517 MB/s

                        host:/lv01 # dd if=/lv01/test1 of=/dev/null bs=2G count=1 oflag=dsync

                        0+1 records in

                        0+1 records out

                        2147479552 bytes (2.1 GB) copied, 1.32954 s, 1.6 GB/s

                        host:/lv01 # rm -Rf /lv01/test1

                         

                         

                        Enterprise systems should be able to get > 500 MB/s.  I have EBS databases that get up to 100,000 IOPS and over 1600 MBPS in a VM.

                         

                        You can can also install iotop with yum install iotop and run it while taking an RMAN backup.