Saturday, February 13, 2021

Oracle Database 19c now supports DBMS_CLOUD.

 If you have been wondering why I've spent so much time blogging about how to configure ZFS as an object store, today is the day you get the answer.




Today MOS note 2748362.1 - How To Setup And Use DBMS_CLOUD Package was published.


You are probably saying, "so what?, 21c supports DBMS_CLOUD, and that's a long way off for me"..

This note goes through the steps to configure DBMS_CLOUD for 19c. Yes. it's backported !

NOTE: it is only supported for multitenant

If you were very astute, you might have noticed that the 19.9 release of the software contained scripts in the $ORACLE_HOME/rdbms/admin directory with names like dbms_cloud.sql.

Well today, this published notes explains how to install DBMS_CLOUD packages so that you can use it with your 19.9 + database.

I'm going to take this a step further, and show you how to use these scripts and connect the ZFS appliance.  Keep in mind, there is a ZFS simulator you can use, and do the same steps.

Here is some information on how to do this if you don't know where to start

Configuring ZFS as an object store

Step 1. Install DBMS_CLOUD in the CDB

I am going through the MOS, and I am following the same series of steps.  Creating a script in my /home/oracle/dbc directory, so that I can run this again for all my databases.

Just as in the note, I ran a perl script, and looked at the logs. Everything was successful.

I then ran the 2 queries. First against the CDB, then against the PDB.


So far so good.


Step 2 Create SSL Wallet with Certificates.

The next step is to create a wallet, and download the certificates using the link.  The certificates come in a zip file containing all 3 certificates.

VeriSign.cer
BaltimoreCyberTrust.cer
DigiCert.cer

These are the Certificate Authorities that will be used to authenticate the SSL certificates.  DBMS_CLOUD uses HTTPS, and requires that a valid certificate is used.



ZFS NOTE BEGIN : *************************************

At this point there is an additional step for using ZFS (or your own object store). You need to add the certificate to the wallet if it is a self-signed certificate (which is what ZFS will use normally).

In order to get the certificate you need to display it with the following command (filling in your IP address).

openssl s_client -showcerts -connect 10.136.64.85:443

From the output I want to grab the certificate which is between the BEGIN and END


-----BEGIN CERTIFICATE-----
MIIEWTCCA0GgAwIBAgIIXJAYBgAAAAIwDQYJKoZIhvcNAQELBQAwcDEtMCsGA1UE
AwwkenM3LTJjYXAtMjAwZi12bTAyLnVzLm9zYy5vcmFjbGUuY29tMT8wPQYDVQQN
DDZodHRwczovL3pzNy0yY2FwLTIwMGYtdm0wMi51cy5vc2Mub3JhY2xlLmNvbToy

..

oH4pa4Hv4/s0GKJcjQDTlhyyAQXHD+EDfa0KSqP6+Rcwv9+pzXhTJu6IYJLanKo

uM6RxG2XAIH82blU+A==
-----END CERTIFICATE-----

Once I put it in a file, I perform the same command to load these certificates from my file.

ZFS NOTE END : *****************************************

Once added I display what is in the wallet.

orapki wallet display -wallet .
Oracle PKI Tool Release 21.0.0.0.0 - Production
Version 21.0.0.0.0
Copyright (c) 2004, 2020, Oracle and/or its affiliates. All rights reserved.

Requested Certificates:
User Certificates:
Trusted Certificates:

Subject:        2.5.4.13=https://zs7-2cap-200f-vm02.bgrenn.com:215/\#cert,CN=zs7-2cap-200f-vm02.bgrenn.com
Subject:        2.5.4.13=https://zs7-2cap-200f-vm01.bgrenn.com:215/\#cert,CN=zs7-2cap-200f-vm01.bgrenn.com
Subject:        CN=DigiCert Global Root CA,OU=www.digicert.com,O=DigiCert Inc,C=US
Subject:        CN=Baltimore CyberTrust Root,OU=CyberTrust,O=Baltimore,C=IE
Subject:        CN=VeriSign Class 3 Public Primary Certification Authority - G5,OU=(c) 2006 VeriSign\, Inc. - For authorized use only,OU=VeriSign Trust Network,O=VeriSign\, Inc.,C=US


You can see that it captured the server names of the ZFS ports I am using.


Step 3 Configure the Oracle environment for the Wallet.


I followed the next set of instructions to update the sqlnet.ora file with the location of wallet.

A few items to note on this step
  • I am in a RAC environment so I need to make the change to ALL nodes in my RAC cluster, and I also need to copy the wallet to the nodes in the same location on all hosts.
  • The WALLET_LOCATION is also used by ZDLRA. if you are using a ZDLRA for backups, you need add the certificates to wallet that is used by the ZDLRA.
  • If you using Single Sign On which may use the WALLET_LOCATION, be especially careful since they often default to $ORACLE_BASE, but will get over ridden when this is set.
I completed these steps, and I now have the same sqlnet.ora file on all nodes, and my wallet is on all nodes in the same location under $ORACLE_BASE.

Step 4 Configure the Database with ACEs for DBMS_CLOUD.


The next step is to create Access Control Entries (ACEs) to allow communication. This only needs to be in CDB$ROOT.

I stored the script in a file and changed the script in 2 ways

  • define sslwalletdir= {my wallet locatioin}  --> I set this.
  • I removed all lines around proxy. I didn't need a proxy since I am only using a ZFS internal to my datacenter.
I verified with the query and it returned the location of SSL_WALLET.

Step 5 Verify the configuration of the DBMS_CLOUD

I put the script in a file and made a few changes.

  • wallet_path => {my wallet path}
  • wallet_password => {my wallet password}
  • get_page('https://zs7-2cap-200f-vm02.bgrenn.com') --> I put in the URL for my first ZFS network name (from the certificate) followed by the second name from the certificate.
I got a "valid response" backup.. 
I can also check the ACLs with the script below
SELECT host, lower_port, upper_port, acl
FROM   dba_network_acls;

Step 6 Configure users or roles to use  DBMS_CLOUD

I changed the script to use my username which I created in my PDB to create tables etc. and utilize DBMS_CLOUD, and ran it in my pdb.

I took the second script, removed the proxy information , entered the wallet path and executed in my  PDB.

Step 6 Configure users or roles to use DBMS_CLOUD


I changed the script and added my username in my PDB.


Step 7 Configure ACEs for a user role to use DBMS_CLOUD


Again I removed the proxy information since there was no proxy. I also entered the SSL_wallet directory.

Step 8  Configure the credential for OCI (or S3 if you prefer).


Using the create credential and the parameters I have pointed out in previous posts.

 I create a credential to point to my OCI bucket on ZFS.


exec DBMS_CLOUD.CREATE_CREDENTIAL ( -
    CREDENTIAL_NAME => 'ZFS', -
    USER_OCID => 'ocid1.user.oc1..oracle', -
    TENANCY_OCID => 'ocid1.tenancy.oc1..nobody', -
    PRIVATE_KEY => 'MIIEogIBAAKCAQEAnLe/u2YjNVac5z1j/Ce7YRSd6wpwaK8elS+TxucaLz32jUaDCUfMbzfSBP0WK00uxbdnRdUAss1F1sRUm+GqyEEvT2c1LRJ0FnfSFEXrJnDZfEVe/dFi90fctbx4BUSqRroh0RQbQyk24710zO2C3tev66eHEvfxxXGUqI+jrDKOJ7sFdGE42R9uRhhWxaWS4e43OEZk41gq2ykdVFlNp...mXU6w6blGpxWkzfPMJKuOhXYoEXM41uxykDX3nq/wPWxKJ7TnShGLyiFMWiuuQF+s29AbwtlAkQRcHnnkvDFHwE=', -
    FINGERPRINT => '1e:6e:0e:79:38:f5:08:ee:7d:87:86:01:13:54:46:c6');

Note the parameters for ZFS

  • CREDENTIAL_NAME - Name of the credential
  • USER_OCID - 'ocid1.user.oci..' || {ZFS user id}
  • TENANCY_ID - 'ocid1.tenancy.oci1..nobody' - hardocded in
  • PRIVATE_KEY - Private key matching the public key on the ZFS
  • FINGERPRINT - fingerprint for the public key on the ZFS.

Step 9  Load raw data to the object store.


First I am going to open a file, and put some data into it.. Upload the file to my OCI bucket and then create an external table on it.

Below is the input file.

 16TS$                           TABLE                    1904172019041720190417VALID
        20ICOL$                         TABLE                    1904172019041720190417VALID
         8C_FILE#_BLOCK#                CLUSTER                  1904172019041720190417VALID
        37I_OBJ2                        INDEX                    1904172019041720190417VALID
        22USER$                         TABLE                    1904172019041720190417VALID
        33I_TAB1                        INDEX                    1904172019041720190417VALID
        40I_OBJ5                        INDEX                    1904172019041720190417VALID
        31CDEF$                         TABLE                    1904172019041720190417VALID
        41I_IND1                        INDEX                    1904172019041720190417VALID
         3I_OBJ#                        INDEX                    1904172019041720190417VALID
         6C_TS#                         CLUSTER                  1904172019041720190417VALID
        51I_CON1                        INDEX                    1904172019041720190417VALID
        34I_UNDO1                       INDEX                    1904172019041720190417VALID
        11I_USER#                       INDEX                    1904172019041720190417VALID
        29C_COBJ#                       CLUSTER                  1904172019041720190417VALID
        49I_COL2                        INDEX                    1904172019041720190417VALID
        32CCOL$                         TABLE                    1904172019041720190417VALID
        14SEG$                          TABLE                    1904172019041720190417VALID
        23PROXY_DATA$                   TABLE                    1904172019041720190417VALID
        44I_FILE2                       INDEX                    1904172019041720190417VALID
        46I_USER1                       INDEX                    1904172019041720190417VALID
        56I_CDEF4                       INDEX                    1904172019041720190417VALID
        21COL$                          TABLE                    1904172019041720190417VALID
        47I_USER2                       INDEX                    1904172019041720190417VALID
        26I_PROXY_ROLE_DATA$_1          INDEX                    1904172019041720190417VALID
        18OBJ$                          TABLE                    1904172019041720190417VALID
        42I_ICOL1                       INDEX                    1904172019041720190417VALID
        19IND$                          TABLE                    1904172019041720190417VALID
        39I_OBJ4                        INDEX                    1904172019041720190417VALID
        59BOOTSTRAP$                    TABLE                    1904172019041720190417VALID
        36I_OBJ1                        INDEX                    1904172019041720190417VALID
        15UNDO$                         TABLE                    1904172019041720190417VALID
        10C_USER#                       CLUSTER                  1904172019041720190417VALID
         4TAB$                          TABLE                    1904172019041720190417VALID
         2C_OBJ#                        CLUSTER                  1904172019041720190417VALID
        28CON$                          TABLE                    1904172019041720190417VALID
         5CLU$                          TABLE                    1904172019041720190417VALID
        27I_PROXY_ROLE_DATA$_2          INDEX                    1904172019041720190417VALID
        24I_PROXY_DATA$                 INDEX                    1904172019041720190417VALID
        45I_TS1                         INDEX                    1904172019041720190417VALID
        13UET$                          TABLE                    1904172019041720190417VALID
        12FET$                          TABLE                    1904172019041720190417VALID
        17FILE$                         TABLE                    1904172019041720190417VALID

I created a file locally (/tmp/objects.csv), created a bucket (using the OCI CLI tool) and uploaded the file.

Create the bucket on zfs

oci os bucket create --endpoint http://zs7-2cap-200f-vm02.bgrenn.com/oci --namespace-name export/objectstoreoci --compartment-id export/objectstoreoci --name bucketoci  


And copy my file to my bucket.

oci os object put --endpoint http://zs7-2cap-200f-vm02.bgrenn.com/oci ---namespace-name export/objectstoreoci --bucket-name bucketoci  - --file /tmp/objects.csv  --name objects.csv


Step 10  Create an external table on the object.


Now we have the file in the bucket we are ready to create the external table.

ZFS NOTE BEGIN : ***************************************

There is an additional step to access the ZFS. There is a table owned by C##CLOUD$SERVICE which contains the objects store that can be accessed, and how to authenticated. By looking at the current entries you can see the types for OCI and S3.

until I do this you will an error like this..


ERROR at line 1:
ORA-20006: Unsupported object store URI -
https://zs7-2cap-200f-vm01.bgrenn.oracle.com/export/objectstoreoci/bucketoci
/objects.csv
ORA-06512: at "C##CLOUD$SERVICE.DBMS_CLOUD", line 917
ORA-06512: at "C##CLOUD$SERVICE.DBMS_CLOUD", line 2411
ORA-06512: at line 1

Here is the table that we need to change. You can see that it contains 
  • CLOUD_TYPE - authentication to use
  • BASE_URI_PATTERN - URI pattern to identify and allow
  • VERSION - This is used if different authentication versions exist for an object store
  • STATUS - Not sure, but they are all '1'


 desc C##CLOUD$SERVICE.dbms_cloud_store;
 Name                       Null?    Type
 ----------------------------------------- -------- ----------------------------
 CLOUD_TYPE                        VARCHAR2(128)
 BASE_URI_PATTERN                    VARCHAR2(4000)
 VERSION                        VARCHAR2(128)
 STATUS                         NUMBER

I add a row to this table for my object store.  ORACLE_BMC is the OCI authentication

SQL> insert into C##CLOUD$SERVICE.dbms_cloud_store values ('ORACLE_BMC','%.bgrenn.com',null,1);

1 row created.

SQL> commit;

Commit complete.


ZFS NOTE END : *****************************************


We are ready, now let's create the table and give it a go !!

Create the external table on the object
exec DBMS_CLOUD.CREATE_EXTERNAL_TABLE( -
    table_name      =>'CHANNELS_EXT_ZFS', -
    credential_name =>'ZFS', -
    file_uri_list   =>'https://zs7-2cap-200f-vm01.bgrenn.com/oci/n/export/objectstoreoci/b/bucketoci/o/objects.csv', -
    format          => json_object('trimspaces' value 'rtrim', 'skipheaders' value '1', 'dateformat' value 'YYYYMMDD'), -
    field_list      => 'object_id      (1:10)   char' || -
                      ', object_name    (11:40)  char' || -
                      ', object_type    (41:65)  char' || -
                      ', created_date1  (66:71)  date mask "YYMMDD"' || -
                      ', created_date2  (72:79)  date' || -
                      ', last_ddl_time  (80:87)  date' || -
                      ', status         (88:97)', -
   column_list     => 'object_id      number' || -
                      ', object_name    varchar2(30)' || -
                      ', object_type    varchar2(25)' || -
                      ', status         varchar2(10)' || -
                      ', created_date1  date' || -
                      ', created_date2  date' || -
                      ', last_ddl_time  date');

Select from the table.

OBJECT_ID OBJECT_NAME              OBJECT_TYPE            STATUS
---------- ------------------------------ ------------------------- ----------
CREATED_D CREATED_D LAST_DDL_
--------- --------- ---------
    20 ICOL$              TABLE             VALID
17-APR-19 17-APR-19 17-APR-19

     8 C_FILE#_BLOCK#          CLUSTER            VALID
17-APR-19 17-APR-19 17-APR-19

    37 I_OBJ2              INDEX             VALID
17-APR-19 17-APR-19 17-APR-19



That's all there is to it.

Enjoy !

Wednesday, February 10, 2021

Oracle Cloud object Store access with rlone.

 If you are using the Oracle Object Store as part of the Oracle Public Cloud, "rclone" is an open source tool you can use to make things easier.


One of the things I really like about RCLONE is that it provides a command line like interface that is easy to use.  If you have looked at the OCI cli tool, it requires a myriad of parameters.  Below is the command I was using with OCI to view my list of buckets (I obfuscated some of the values).

oci os bucket list --endpoint https://objectstorage.us-ashburn-1.oraclecloud.com  --namespace-name id20xxxxxofo --compartment-id ocid1.compartment.oc1..aaaaaaxxxxxxxxxxxxxxxxxcpqyvzzb4ykd3tyq --config-file ~/.oci/natdconfig 

In order to use the OCI tool, I had to constantly keep a text file open to copy and paste commands.

In comparison, this is the command to list the buckets in my object store using rlcone.

rclone ls oci_bucket:


1) Configure compatibility for an S3 interface in the Public cloud.

In your public cloud council, in the top right hand corner, click on the "silhouette" that controls you settings. in the pull down menu click on "user settings" to bring up the window to configure you resources.  Once there, click on "Customer Secret keys" and then "Generate Secret Key" bring up the window to add a secret key.



On this window give your secret key a name (like S3Key" in my case).  When you click the "Generate Secret key" button, it will give you secret associated with key. SAVE THIS.

Once complete, you will have 2 items associated with your account

NAME:            S3Key                                                        or whatever you named your key.
Access Key:    ddddddddddddeeeeeeeeeffffffffggggg     A system generated "secret"

2) Download rclone.

     This can easily be done from the  RCLONE.ORG site.


2) Configure Rclone.Once rclone is installed.

    You start by executing "rclone config". this will create a configuration file in ~/.config/rclone/rclone called rclone.conf.  This is an interactive interface that will set the correct configuration parameters to be used.

This is an example of what I entered to connect to my Object Store using the 3 interface.

--> rclone config
 
Name> oci_s3

Type of storage to configure.
Enter a string value. Press Enter for the default ("").
Choose a number from below, or type in your own value
..
 4 / Amazon S3 Compliant Storage Provider (AWS, Alibaba, Ceph, Digital Ocean, Dreamhost, IBM COS, Minio, Tencent COS, etc)
..
Storage> 4

Choose your S3 provider.
Enter a string value. Press Enter for the default ("").
Choose a number from below, or type in your own value
..
13 / Any other S3 compatible provider
..
provider> 13
Get AWS credentials from runtime (environment variables or EC2/ECS meta data if no env vars).
Only applies if access_key_id and secret_access_key is blank.
Enter a boolean value (true or false). Press Enter for the default ("false").
Choose a number from below, or type in your own value
 1 / Enter AWS credentials in the next step
   \ "false"
env_auth> 1
AWS Access Key ID.
Leave blank for anonymous access or runtime credentials.
Enter a string value. Press Enter for the default ("").
access_key_id> S3_Key   --> This is what I created in the public cloud
AWS Secret Access Key (password)
Leave blank for anonymous access or runtime credentials.
Enter a string value. Press Enter for the default ("").
secret_access_key> ddddddddddddeeeeeeeeeffffffffggggg  --> The system generated key associated with my access key
Region to connect to.
Leave blank if you are using an S3 clone and you don't have a region.
Enter a string value. Press Enter for the default ("").
Choose a number from below, or type in your own value
 1 / Use this if unsure. Will use v4 signatures and an empty region.
   \ ""
region>
Endpoint for S3 API.
Required when using an S3 clone.
Enter a string value. Press Enter for the default ("").
Choose a number from below, or type in your own value
endpoint> {namespace}.compat.objectstorage.us-ashburn-1.oraclecloud.com --> Note that you will need to fill in your namespace from your account, and ensure the region is correct for the URL.
Location constraint - must be set to match the Region.
Leave blank if not sure. Used when creating buckets only.
Enter a string value. Press Enter for the default ("").
location_constraint>
Canned ACL used when creating buckets and storing or copying objects.

This ACL is used for creating objects and if bucket_acl isn't set, for creating buckets too.

For more info visit https://docs.aws.amazon.com/AmazonS3/latest/dev/acl-overview.html#canned-acl

Note that this ACL is applied when server side copying objects as S3
doesn't copy the ACL from the source but rather writes a fresh one.
Enter a string value. Press Enter for the default ("").
Choose a number from below, or type in your own value
 1 / Owner gets FULL_CONTROL. No one else has access rights (default).
   \ "private"
acl>
Edit advanced config? (y/n)
y) Yes
n) No (default)
y/n>

Remote config
--------------------
[oci_s3]
type = s3
provider = Other
env_auth = false
access_key_id = S3_Key
secret_access_key = ddddddd...
endpoint = xxxxxxx.compat.objectstorage.us-ashburn-1.oraclecloud.com
--------------------
y) Yes this is OK (default)
e) Edit this remote
d) Delete this remote
y/e/d> y

Now let's verify what got create.

> cat ~/.config/rclone/rclone.conf

[oci_s3]
type = s3
provider = other
env_auth = false
access_key_id =s3_key
secret_access_key = dddddxxxxxx
endpoint = xxxxxxx.compat.objectstorage.us-ashburn-1.oraclecloud.com
acl = authenticated-read

That's It.  In my case the entry is oci_s3


Now with rclone I can execute commands against my object store that are more linux like.

rclone mkdir oci_s3:mybucket  --> will create a bucket named "mybucket"
rclone ls oci_s3:  --> will list all my buckets
rclone ls oci_s3:mybucket --> will list all the objects in my bucket.

I can also use it to copy to and from my bucket.

rlcone copy /home/oracle/myfile.txt oci_s3:mybucket   --> copies the file to the bucket.

Finally, a great command is sync to synchronize the contents of my on-prem to the cloud

 rlcone sync /home/oracle/mydir/ oci_s3:mybucket  --> this will sync the 2

Now how fun with it !!







Tuesday, February 2, 2021

ZDLRA - Using Protection Policies to manage databases that have migrated or to be retired

 One the questions that keeps coming up with ZDLRA is how to manage the backups for a database that has either

  • Been migrated to another ZDRA
  • Been retired, but the backup needs to be kept for a period of time












The best way to deal with this by the use of Protection Policies.

How Protection Policies work:


If you remember right, Protection Policies are way of grouping databases together that have the same basic characteristics.

The most important of which are :

Name/Description             - Used to identify the Protection Policy
Recovery Window Goal    - How many days of recovery do you want to store at a minimum 
Max Retention Window    - (Optional) Maximum number of days of backups you want to keep
Unprotected Window        - (Optional) Used to set alerts for databases that are no longer receiving recovery data.

One of the common questions I get is.. What happens if I change the Protection Policy associated with my database ?

Answer :  By changing the Protection Policy a database is associated with, you are only changing the metadata.  Once the change is made, the database follows the Protection Policy rules it is now associated with, and no longer is associated with the old Protection Policy

How this plays out with a real example is... 
My Database (PRODDB) is a member of a Protection Policy (GOLD) which has a Recovery Window Goal of 20 days, and a Max Retention Window of 40 days (the default value being 2x the Recovery Window Goal).
My Database (PRODDB) currently has 30 days of backups, which is right in the middle. 



 What would normally happen for this database is (given enough space), backups will continue to be kept until PRODDB has 40 days of backups.  On day 41, a maintenance job (which runs daily) will execute, and find that my database, PRODDB, has exceeded it's Recovery Window Goal.  This job will remove all backups (in a batch process for efficiency) that are older than 20 days.

BUT ........................

Today, I moved my database, PRODDB, to a new protection policy (Silver) which only has a 10 day Recovery Window Goal, and a Max Recovery Window of 20 Days.


As I pointed out, the characteristics of the NEW Protection Policy will be used, and the next time the daily purge occurs, this database will be flagged, and all backups greater than the Recovery Window Goal will be purged.





Retiring databases: - 

One very common question how to handle the retiring of database.  As you might know, when you remove a database from the ZDLRA, ALL backups are removed from ZDLRA.
When a database is no longer sending backups to the ZDLRA,  the backups will continue to be purged until only a single level 0 backup remains.  This is to ensure that at least one backup is kept, regardless of the Max Recovery Window.
The best way to deal with Retiring database (and still keep the last Level 0 backup) through the use of Protection Policies.
In my example for my database PRODDB, I am going to retire the database instead of moving it to the Silver policy.  My companies standard is to  keep the final backup for my database available for 90 days, and on day 91 all backups can be removed.

These are requirements from the above information.
  • At least 1 backup is kept for 90 days, even though my Max Recovery Window was 40 days.
  • I want to know when my database has been retired for 90 days so I can remove it from the ZDLRA.
In order to accomplish both of these items, I am going to create a Protection Policy named RETIRED_DB with the following attributes
  • Recovery Window Goal of 2 days
  • Max Recovery Window of 3 Days
  • Unprotected Data Window of 90 days
  • New Alert in OEM to tell me when a database in this policy violates its Unprotected Data Window
If you look closely at the attributes, you will noticed that I decreased the Recovery Window Goal to allow backups to be removed after 3 days.  I also set the Unprotected Data Window to be 90 days.
What this looks like over  time is 




As you can see by moving it to the new policy, within a few days, all backups except for the most recent Full back is removed.  You can also see that on day 91 (when it's time to remove this database) I will be getting an alert.

Migrating Databases:

Migrating databases is very similar to retiring databases, except that I don't want remove the old backups until they naturally expire.  For my example of PRODB with a Recovery Window Goal of 20 days, as soon as I have a new Level 0 on the new ZDLRA, I will move this database to a new policy (GOLD_MIGRATED) with the following attributes.
  • Recovery Window Goal of 20 days, since I still need to preserve old backups
  • Max Recovery Window goal of 21 days. This will  remove the old backups as they age off.
  • Unprotected Data Window of 21 days, which will alert me that it time to remove this database.
How this would look over time time is.




Conclusion:

When retiring or migrating databases, Protection Policies can be leveraged to both
  • Ensure backups are removed as they age out until only a single L0 (Full) remains
  • Alert you when it is time to remove the database from the ZDLRA.

Thursday, January 21, 2021

ZFS as a swift object store

 This blog post goes through a feature of the ZFS Appliance that has been around for at least 3 years now. The Openstack Swift Object store.


When looking at the S3 API, and the OCI API, I forgot all about where it started.. With the Swift API.

I will go through the 3 APIs, and how they came about (from what I can find by reading through articles)..

It all started with the Swift API. Swift (V1) has simple authentication and a simple interface.

A URI to manage/access objects has the format of

HTTP://{object store server}/object/v1/{Account}/{bucket name}/{object name}.

In the case of ZFS, 

  • Account - this is the share name.. "/export/swiftshare" for example
  • Bucket name - The name of the bucket that was created
  • Object name - name of the object.
Authentication with Swift while using curl is typically a 2 step process.  
First the authorization URI is called
HTTP://{object store server}/auth/v1.0/{Account}

The username and password is sent with the authentication URI.  The URI then returns an auth token which is used in the curl command line to manage buckets/objects.

Username/password authentication (v1.0) is one of the 3 choices.
  1. Local username/password created on the ZDLRA.
  2. LDAP user ZFS ties to
  3. Keystone authentication server.
For all the testing I am doing on my ZFS simulator, I use a local user.

Before I go into how to configure and use the Swift interface on ZFS, I'll share what I was able to find out.

The Swift API has some limitations, and these limitations is what drove the move to S3.
As you probably noticed, the authentication and tracking of objects does not have enough details to support the segregation of users, and billing.
The S3 API takes the Swift API, and adds the ability to create separate tenants, set up billing, etc. All the things an enterprise needs to do.

With S3, you probably noticed that the authentication layer changed. It is based on secret name/secret rather than a username/password returning an auth token..

Well lets go through what it takes to configure the Swift interface.

First, most of the steps around configuring ZFS for an object store, I documented in my previous blog posts. 
If you look the posts below you see the steps on configuring a share,creating a local user on ZFS, and configuring the http service.

ZFS Appliance - Your on-premise cloud store


For Swift, I will just go the steps specific to Swift.

All I need to do is enable swift. That's it !



Swift gets enabled just like enabling the S3 API, and the OCI API. Because I do not have a Keystone Authentication Server (which would be the OpenStack Identity Service), I didn't fill those values in.

NOTE: Authentication for swift is a little different from S3, or OCI. Both of the other APIs do not tie directly to the local user.  S3 uses "secrets", and OCI uses a PEM file, and a Fingerprint.

Accessing my Swift bucket.

First some links to documentation that will give you examples of these ways of connecting.
Swift Guide for ZFS OS 8.8 release  -- Current release as of writing.
Using ZFS as an object store  -- This is old, but has a lot of detail and great detail
API Guide OS 8.8 for Swift Docs -- Current documentation guide

Also, since the Swift implementation is OpenStack, there is a lot of examples and documentation (non-oracle) available across the web.


I was able to access my bucket one of 3 ways

The first 2 ways are very similar

Swift command line tool - python based tool to connect to swift and manage buckets

CURL - Command line took similar to swift.



In order to create a bucket, and upload an object ......

First I execute the  curl command to get the authentication token.


NOTE: my ZFS emulator is 10.0.0.110 and my share is /export/short

curl -i http://10.0.0.110/auth/v1.0/export/short -X GET -H "X-Auth-User: oracle" -H "X-Auth-Key: oracle123"

HTTP/1.1 200 OK
Date: Thu, 21 Jan 2021 18:54:38 GMT
Server: Apache
X-Content-Type-Options: nosniff
X-Storage-Url: http://10.0.0.110:80/object/v1/export/short
X-Auth-Token: ZFSSA_522d6355-9056-4a95-9060-c88648007993
X-Storage-Token: ZFSSA_522d6355-9056-4a95-9060-c88648007993
Content-Length: 0
X-Trans-Id: tx62e2f031f21640c29a2bf-006009cdee
Content-Type: text/html; charset=utf-8

Next I execute create a bucket .

From the output  above I can get the "Auth Token", and the Storage URL to manage the object store in curl. Note that the Auth Token will expire.

Create a container in curl

curl -i http://10.0.0.110:80/object/v1/export/short/bucketswift -X PUT -H "Content-Length: 0" -H "X-Auth-Token: ZFSSA_522d6355-9056-4a95-9060-c88648007993"

Create a container in swift

swift post container -A http://10.0.0.110:80/object/v1/export/short -U oracle -K oracle123


That's all there is to it with the Swift Object Store on ZFS.




Tuesday, January 5, 2021

Managing an Object Store on ZFS

 This blog post will cover how to access the object store on ZFS to create buckets and upload files. For S3, I am using Cloudberry, which I downloaded here. For OCI, I am using the OCI cli tool.












S3 access to ZFS

This is the easiest, since the S3 object store on ZFS is an S3 compatible interface.

In cloudberry add a new account and use the following for input.





Note that you need to enter the 4 fields above.

  • Display Name -- What you want to call the new account entry
  • Service Point  -- This is the ZFS interface for S3 in the form of 
HTTP://{the IP of the ZFS}/s3/v1/{share name}
  • Access Key  -- This is the name you gave the S3 access key, when you added it to the ZFS
  • Secret Key   -- This is the long string of characters that was returned by ZFS when you created the key.
That's it ! You can now use cloudberry to create buckets, upload files, sync object stores etc.


OCI access to ZFS

Install the CLI

Like the Oracle cloud, there is currently (as of me writing the blog post), no GUI tool like cloudberry that will connect to an OCI object store.  When connecting to the Oracle cloud, you can access the OCI object store through the S3 interface, but this is not possible on ZFS. Both the OCI and S3 object store are independent and cannot access buckets etc. in the other object store.

In order to access ZFS through OCI we start with downloading the OCI cli tool. Documentation on how do this can be found here.

In my install, I took the easy route  (and since I had a Ubuntu client with root access to play with). I installed it directly using "sudo pip install oci-cli"

Create a config file.


Once you have the OCI cli installed we need to set up a configuration file to be used.
The default file is ~/.oci/config, but this location can be changed when using the command if you access multiple OCI installations.

This is the contents of my file.


1
2
3
4
5
6
7
[DEFAULT]
user=ocid1.user.oc1..oracle
fingerprint=1e:6e:0e:79:38:f5:08:ee:7d:87:86:01:13:54:46:c6
key_file=/home/oracle/opc/oracle_private.pem
tenancy=ocid1.tenancy.oc1..nobody
region=us-phoenix-1
pass_phrase = oracle

Now to walk through each line.

1. This identified the entry. Since the config file can contain entries for multiple OCI locations, this entry is identified as the default entry to use (If I don't specify one).
2. This is the user ID.  Since I am using ZFS, the format is "ocid1.user.oc1..{zfs user}"
3. This is the fingerprint. I mentioned in the last blog post that this will be needed. This fingerprint identifies the API public_key entry on ZFS to use when matching the private API key being sent
4. This is the private key file. This contains the private API key that matches the public key that was added to the ZFS.
5. This is unimportant to ZFS, but is required to be set. Use the entry above.
6. Like #5. this is not used by ZFS but is needed by the OCI client.
7. This is optional. If the API private key was created with a pass_phrase, this the pass_phrase that matches the private key.


Create a bucket on OCI.

Almost there now ! We have everything in place for authentication, and we are ready to create an OCI bucket on ZFS for storing data.

The command is 

oci os bucket create --endpoint {OCI object store location} --namespace-name {location on the object store} --compartment-id {compartment in OCI} --name {new bucket name}


Now let's walk through what the parameters will be for ZFS

--endpoint               -> For my ZFS appliance, it is the url + oci
--namespace-name  -> This is the share on the ZFS.  "/export/short" in my config.
--compartment-id    -> This is also the share on the ZFS.  "/export/short" in my config.
--name                     -> the name of the bucket I want to create.

For my configuration below is the command and the output.. I now have a bucket created, and I am able upload data !


oci os bucket create --endpoint http://10.0.0.110/oci --namespace-name export/short --compartment-id export/short --name mynewbucket 

{
  "data": {
    "approximate-count": null,
    "approximate-size": null,
    "compartment-id": "export/short",
    "created-by": "oracle",
    "defined-tags": null,
    "etag": "a51c8ecbf1429f95b446c4413df9f494",
    "freeform-tags": null,
    "id": null,
    "is-read-only": null,
    "kms-key-id": null,
    "metadata": null,
    "name": "mynewbucket",
    "namespace": "export/short",
    "object-events-enabled": null,
    "object-lifecycle-policy-etag": null,
    "public-access-type": "NoPublicAccess",
    "replication-enabled": null,
    "storage-tier": "Standard",
    "time-created": "2021-01-05T16:15:05+00:00",
    "versioning": null
  },
  "etag": "a51c8ecbf1429f95b446c4413df9f494"
}



ADVANCED TOPIC -- SSL with OCI CLI


Now let's say I want to encrypt my connections to OCI and use the HTTPS server available on ZFS.
First I need to create a file containing the certificate. I can get the certificate by executing.

openssl s_client -showcerts -connect 10.0.0.110:443

This returns a lot of information, but within the output I can see the certificate, and I can copy and paste into a file.

Certificate chain
 0 s:CN = 10.0.0.110, description = https://10.0.0.110:215/#cert
   i:CN = 10.0.0.110, description = https://10.0.0.110:215/#cert
-----BEGIN CERTIFICATE-----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-----END CERTIFICATE-----
---
Server certificate
subject=CN = 10.0.0.110, description = https://10.0.0.110:215/#cert

I want to copy the certificate including the "BEGIN CERTIFICATE" and "END CERTIFICATE" lines into a file. 

I now need to set my environment to see the certificate file and use it. In my case "/home/oracle/opc/wallet_cloud/zfs.cer"

export REQUESTS_CA_BUNDLE=/home/oracle/opc/wallet_cloud/zfs.cer

I can now view the buckets in my object store, and upload files encrypted.

oci os bucket list --endpoint https://10.0.0.110/oci --namespace-name export/short --compartment-id export/short 
{
  "data": [
    {
      "compartment-id": "export/short",
      "created-by": "oracle",
      "defined-tags": null,
      "etag": "a51c8ecbf1429f95b446c4413df9f494",
      "freeform-tags": null,
      "name": "mynewbucket",
      "namespace": "export/short",
      "time-created": "2021-01-05T16:15:05+00:00"
    }
  ]
}


The OCI documentation should give you everything you need to upload/download objects within a bucket.

Managing authentication for a ZFS Object Store

 As promised, I am continuing my blog series on how to work with ZFS as a cloud store.



My first blog post went through the steps of how to configure ZFS as an object store.

This post will go through how to create the authentication keys/secrets to access  Object Store.

OCI/S3 user management

The first thing to do is to create a user on the ZFS that will be used as the owner of the object store.

In my case I am going to use the "oracle" account, and ensure that the GUID is the same as the GUID I use on all my DB servers.


NOTE: Most of the information I used to go through the process was from this document.

Create the user

Start by logging into the web interface for the ZFS appliance and navigate to Configuration -> users .

Once on this page, click on the + sign next to users to create a new user. In the example, I had already created the Oracle user.





Now on the create user page, ensure the user is a "local" user, and the "User ID" is the same as the GUID I normally use for Oracle. After entering the information, click on "ADD" in the upper right hand corner to add the user.











Change share ownership

Now that I have the "oracle" user created, I am going to change ownership on my share that will be my object store.

In order to do this, I am going navigate to Shares -> SHARES . I see my object store share and highlight it and click on the pencil icon to edit the share.






I am now on the detail screen for my share, and I navigate to the Shares -> SHARES -> Access page.
On this page, I change the user to be Oracle, and ensure the permissions are open enough. Once this change is made click on the apply button in the top right hand corner.







OCI Authentication.

Create the API keys

Now we need to add to add the API key to authenticate the user to the Object store.
In the case of an OCI bucket, authentication is performed by using an X.509 certificate.
This is the same authentication used for an OCI bucket in the Oracle Cloud.

Instructions on how to create an API signing key can be found here.

In my case I used the linux command instructions, and the openssl command to create both a private and public key.  When completed, I had 2 files.

/home/oracle/opc/oracle_public.pem  
        -----BEGIN PUBLIC KEY-----
        sdfa23
        ....
        -----END PUBLIC KEY-----

/home/oracle/opc/oracle_private.pem  
       -----BEGIN RSA PRIVATE KEY----
        5dfgsret345
        ....
        -----END RSA PRIVATE KEY-----

Add the API key

Now that we generated the Keys, lets add them to the share so we can access the OCI object store.

We start by going to Configuration-> Services -> HTTP. Click on HTTP to bring up the next page.















On this page, we want to go to the OCI tab and add a new key.














On the "New Key" window that popped up, add the Oracle user, and paste in the public key. Once everything is entered click on add to create the key.



















Once added make note of the Fingerprint.






S3 Authentication.

Create the Secret.

Unlike OCI, S3 authentication is done through a "secret".  The use of a "secret" is similar to the idea of a Key and a Passcode. You create a new access  key for the user, and you then you are provided a long string that is the "passcode" for this key.

Like creating API key for OCI, we start by going to the HTTP service.















This time we go the S3 tab under HTTP and click on the + sign to add a key.
















Enter the oracle user, and give your key a name. Once complete, click on ADD to create the key.











Now you will see a window with the Secret Key.
SAVE THIS KEY. you will not be given this key again. You can copy and paste it, but better yet, save it in a file.













Authentication for S3 and OCI.


When you completed both of these actions you will have 2 authentication pieces that we will use to create buckets and access the object store in future blog posts.

S3 - You have an "ACCESSS_KEY" and a "SECRET_KEY" that will be used.

OCI - You have a file containing the private_key, public_key, and the fingerprint associated with the public_key to identify it.


Wednesday, December 2, 2020

Advanced Compression and TDE

 This blog post details what happens when you utilize ACO to mitigate the effects of TDE.




I started researching this topic as I thought about what happens when ACO is implemented on a table.

The first thought on Implementing ACO is that it will mitigate the effects of TDE by reducing the size of the data, and thus reducing the size of the backup RMAN creates (without using RMAN compression). Most backup strategies use compression, and TDE data will not encrypt. Implementing ACO to reduce the database size, reduces the backup size and mitigates the effect of TDE.

This all sounds good, and at first glance makes sense.

Of course there is a lot more to ACO, then just mitigating the loss of compression on my backup strategy.

  • The size of the level 0 is smaller, thus my restore time can be reduced AND the amount of storage needed for my Level 0 is lessened.
  • I am reading fewer blocks for the same amount of data. Less physical reads typically means faster queries (that perform disk reads).

If I am able to achieve 2x compression through ACO, my backup size is reduced to be 1/2 the size it would be without ACO right ?

Well .... There are few things to think about.

First - Indexes may be a large part of your database. I have seen applications where the indexes take up more space than the data itself.  Also, you can only compress indexes that have more than 1 column. In a multi-column index, the prefix columns can be compressed. Single column indexes cannot be compressed.


Second- Let's take a close look at how ACO works with my table. What happens with my level 0 backup is pretty clear. I would expect to get at least a 2x compression ratio. What I wanted to explore is what happens to my daily incrementals and archive logs ?

Here is the documentation that best describes it.  Below is a picture that shows what happens with OLTP compression. This is where I looked close at.


What caught my eye in this is that the block starts out initially uncompressed. As the block gets full, a background process will compress the rows and the resulting block will be compressed. The thoughts I had were

  • That's great for my incremental backup size.  More data in each block means less blocks to backup.
  • With my redo, the row is initially inserted in the block at the current scn, and the batch process then changes the block and that row is moved. The movement of the row needs to be captured in the redo, resulting in more redo.
Now to put this to the test and see what happens.

I started by creating my own copy of DBA_OBJECTS and continued to replicate the data until it was ~ 100MB. This seemed like a good size to work with. In order to facility updates to this dataset (since there were duplicates), I added a column that is the rownum. This gives each row a unique key from 1 to the number of rows.

My table is named "bgrenn.myobjects"
My tablespace is ENCRYPTED_DATA and was created with the ENCRYPTION clause.

Step 1 - I created a table to hold the results of the changes to analyze.  Below is the definition (if you want to follow along at home).

create table bgrenn.tests(test_type varchar(10),
              test_table varchar(20),
              test_number integer,
              test_start_seq# number,
              test_end_seq# number,
                          rows_affected number,
                          blocks_changed number,
                          incr_size number,
                          archive_size number,
                          total_size number) tablespace encrypted_data;
Step 2 - I created 2 empty tables in the encrypted tablespace. One defined with OLTP compression, one without.

create table bgrenn.uncompressed tablespace encrypted_data as select * from bgrenn.myobjects where 1=0;
create table bgrenn.compressed tablespace encrypted_data  ROW STORE COMPRESS ADVANCED as select * from bgrenn.myobjects where 1=0;

INSERTS


Step 3
- Execute my procedure to insert into the uncompressed table committing every 100 rows. At the end of the procedure I insert into my "tests" table the information about this action.

DECLARE
   TYPE mytest_cur IS REF CURSOR;
   commit_count number := 0;
   insert_count number := 0;
   start_seq_number number;
   end_seq_number number;
   blocks_changed number;
   start_scn number;
   archive_size number;
  intable bgrenn.myobjects%ROWTYPE;
  CURSOR c1 IS SELECT
     * from bgrenn.myobjects;

BEGIN
   OPEN c1;
   execute immediate 'alter system archive log current';
   select max(sequence#) into start_seq_number from v$log;
   select current_scn into start_scn from v$database;
   LOOP
      FETCH c1 INTO intable;
      EXIT WHEN c1%NOTFOUND;
       insert into bgrenn.uncompressed values intable ;
       if commit_count=100 then
     commit;
     commit_count:=0;
     end if;
    commit_count := commit_count + 1;
    insert_count := insert_count + 1;
   END LOOP;
   dbms_output.put_line('# of rows inserted = ' || insert_count);
   execute immediate 'alter system archive log current';
   dbms_stats.gather_table_stats('BGRENN','UNCOMPRESSED');
   select sum(blocks) into blocks_changed from dba_segments where owner= 'BGRENN' and segment_name in ( 'UNCOMPRESSED','UNCOMPRESSED_IDX');
   select count(distinct dbms_rowid.rowid_block_number(rowid)) into blocks_changed from bgrenn.uncompressed where ora_rowscn >= start_scn;
   select max(sequence#) - 1 into end_seq_number from v$log;
   select sum(blocks*block_size) into archive_size from v$archived_log where sequence#>=start_seq_number and sequence# <=end_seq_number;
   insert into bgrenn.tests values('Insert','Uncompressed',1,start_seq_number,end_seq_number,insert_count,blocks_changed,blocks_changed*8192,archive_size,blocks_changed*8192+archive_size);
   commit;
   CLOSE c1;
END;
/

Step 4 - Execute my procedure to insert into the compressed table committing every 100 rows. 



set serveroutput on;
DECLARE
   TYPE mytest_cur IS REF CURSOR;
   commit_count number := 0;
   insert_count number := 0;
   start_seq_number number;
   end_seq_number number;
   blocks_changed number;
   archive_size number;
   start_scn number;
  intable bgrenn.myobjects%ROWTYPE;
  CURSOR c1 IS SELECT
     * from bgrenn.myobjects;

BEGIN
   OPEN c1;
   execute immediate 'alter system archive log current';
   select max(sequence#) into start_seq_number from v$log;
   select current_scn into start_scn from v$database;
   LOOP
      FETCH c1 INTO intable;
      EXIT WHEN c1%NOTFOUND;
       insert into bgrenn.compressed values intable ;
       if commit_count=100 then
     commit;
     commit_count:=0;
     end if;
    commit_count := commit_count + 1;
    insert_count := insert_count + 1;
   END LOOP;
   dbms_output.put_line('# of rows inserted = ' || insert_count);
   execute immediate 'alter system archive log current';
   dbms_stats.gather_table_stats('BGRENN','COMPRESSED');
   select sum(blocks) into blocks_changed from dba_segments where owner= 'BGRENN' and segment_name in ( 'COMPRESSED','COMPRESSED_IDX');
   select count(distinct dbms_rowid.rowid_block_number(rowid)) into blocks_changed from bgrenn.compressed where ora_rowscn >= start_scn;
   select max(sequence#) - 1 into end_seq_number from v$log;
   select sum(blocks*block_size) into archive_size from v$archived_log where sequence#>=start_seq_number and sequence# <=end_seq_number;
   insert into bgrenn.tests values('Insert','Compressed',1,start_seq_number,end_seq_number,insert_count,blocks_changed,blocks_changed*8192,archive_size,blocks_changed*8192+archive_size);
   commit;
   CLOSE c1;
END;
/

Now that we have done our inserts, let's take a look and see what happened.

TEST_TYPE  TEST_TABLE           ROWS_AFFECTED  INCR_SIZE ARCHIVE_SIZE TOTAL_SIZE
---------- -------------------- ------------- ---------- ------------ ----------
Insert     Uncompressed                585640   96075776    282468864  378544640
Insert     Compressed                  585640   50331648    476089856  526421504

I inserted the same number of rows into both tables, in fact I inserted the same exact rows.

When looking at the size of the incremental backup (these will become part of the level 0 backup) the comparison is.
    Compressed      - 50331648    
    Uncompressed - 96075776    
    Compression ratio  1.9X
Perfect ! My incremental backup size of my compressed table size is about 1/2 of that of uncompressed data. This makes up for the loss of compression in my backup.

Now let's take a look at the Archive Log Backup. These will be kept for the retention window.

Compressed        -    476089856  
Uncompressed    -    282468864  
Compression ratio    .60x

Wow, it's less than 1. The archive logs for compressed data is almost double what they are for uncompressed.

Finally let's take a look at the Total Change Backup size.

Compressed        -    526421504  
Uncompressed    -    378544640
Compression ratio    .71x

INSERTS - The daily backup size for inserts is bigger for compressed data.

UPDATES


Step 5 - Execute my procedure to update 1% of the data in  the uncompressed table committing every 100 rows. 


DECLARE
   commit_count number := 0;
   update_count number := 0;
   max_updates number := 0;
   max_rows number := 0;
   random_row number :=0;
   start_seq_number number;
   end_seq_number number;
   blocks_changed number;
   archive_size number;
   start_scn number;
BEGIN
   select current_scn into start_scn from v$database;
   select count(1) into max_rows from bgrenn.uncompressed;
   max_updates := max_rows * .01;
  execute immediate 'alter system archive log current';
   select max(sequence#) into start_seq_number from v$log;

   LOOP
      EXIT WHEN update_count > max_updates;
       select trunc(dbms_random.value(1,max_rows),0) into random_row from dual;
       UPDATE bgrenn.uncompressed SET SUBOBJECT_NAME= OBJECT_NAME WHERE myrownum = random_row;
       insert into bgrenn.myblockchanges select dbms_rowid.rowid_block_number(rowid) from  bgrenn.uncompressed where myrownum = random_row;
       if commit_count=100 then
     commit;
     commit_count:=0;
     end if;
    commit_count := commit_count + 1;
    update_count := update_count + 1;
   END LOOP;
   execute immediate 'alter system archive log current';
   dbms_stats.gather_table_stats('BGRENN','UNCOMPRESSED');
   select count(distinct block_number) into blocks_changed from bgrenn.myblockchanges;
   select count(distinct dbms_rowid.rowid_block_number(rowid)) into blocks_changed from bgrenn.uncompressed where ora_rowscn >= start_scn;
   select max(sequence#) - 1 into end_seq_number from v$log;
   select sum(blocks*block_size) into archive_size from v$archived_log where sequence#>=start_seq_number and sequence# <=end_seq_number;
   insert into bgrenn.tests values('Update','Uncompressed',1,start_seq_number,end_seq_number,update_count,blocks_changed,blocks_changed*8192,archive_size,blocks_changed*8192+archive_size);
   commit;
   dbms_output.put_line('# of rows updated = ' || update_count);

 END;
/

Step 6 - Execute my procedure to update 1% of the data in  the compressed table committing every 100 rows. 

DECLARE
   commit_count number := 0;
   update_count number := 0;
   max_updates number := 0;
   max_rows number := 0;
   random_row number :=0;
   start_seq_number number;
   end_seq_number number;
   blocks_changed number;
   archive_size number;
   start_scn number;
BEGIN
   select current_scn into start_scn from v$database;
   select count(1) into max_rows from bgrenn.compressed;
   max_updates := max_rows * .01;
  execute immediate 'alter system archive log current';
   select max(sequence#) into start_seq_number from v$log;

   LOOP
      EXIT WHEN update_count > max_updates;
       select trunc(dbms_random.value(1,max_rows),0) into random_row from dual;
       UPDATE bgrenn.compressed SET SUBOBJECT_NAME= OBJECT_NAME WHERE myrownum = random_row;
       insert into bgrenn.myblockchanges select dbms_rowid.rowid_block_number(rowid) from  bgrenn.compressed where myrownum = random_row;
       if commit_count=100 then
     commit;
     commit_count:=0;
     end if;
    commit_count := commit_count + 1;
    update_count := update_count + 1;
   END LOOP;
   execute immediate 'alter system archive log current';
   dbms_stats.gather_table_stats('BGRENN','COMPRESSED');
   select count(distinct block_number) into blocks_changed from bgrenn.myblockchanges;
   select max(sequence#) - 1 into end_seq_number from v$log;
  select count(distinct dbms_rowid.rowid_block_number(rowid)) into blocks_changed from bgrenn.compressed where ora_rowscn >= start_scn;
    select sum(blocks*block_size) into archive_size from v$archived_log where sequence#>=start_seq_number and sequence# <=end_seq_number;
   insert into bgrenn.tests values('Update','Compressed',1,start_seq_number,end_seq_number,update_count,blocks_changed,blocks_changed*8192,archive_size,blocks_changed*8192+archive_size);
   commit;
   dbms_output.put_line('# of rows updated = ' || update_count);

 END;
/

Now that we have done our updates, let's take a look and see what happened.

TEST_TYPE  TEST_TABLE           ROWS_AFFECTED  INCR_SIZE ARCHIVE_SIZE TOTAL_SIZE
---------- -------------------- ------------- ---------- ------------ ----------
Update     Uncompressed                  5857   37699584      5346816   43046400
Update     Compressed                    5857   30203904     11993600   42197504

I updated the same number of rows into both tables.

When looking at the size of the incremental backup (these will become part of the level 0 backup) the comparison is.
    Compressed      - 30,203,904         
    Uncompressed -  37,699,584      
    Compression ratio  1.2X

Perfect ! My incremental backup size of my compressed table size isn't much different than that of uncompressed data. This isn't surprising. with a 1% change rate, the same block might not be updated more than once. Same number updates = same number of blocks backed up.


Now let's take a look at the Archive Log Backup. These will be kept for the retention window.

Compressed        -    11,993,600   
Uncompressed    -      5,346,816     
Compression ratio    .44x

Wow, it's less than 1/2. The archive logs for compressed data is about double what they are for uncompressed.

Finally let's take a look at the Total Change Backup size.

Compressed        -     42197504
Uncompressed    -     43046400
Compression ratio    1x

UPDATES - The daily backup size for updates is about the same for both compressed data and uncompressed data.

SUMMARY.


Implementing ACO does have a large number of benefits. One of which is decreasing the size of a full backup. When implementing TDE, this can help mitigate the loss of compression in your backup strategy.
However . If you have a high change rate, and a long retention window, the backups may end up being the same size as they were before, and could possibly be bigger. This is especially relevant in a backup strategy that includes deduplication/incremental merge/virtual fulls.