Exadata Storage Snapshots

This post describes how to implement Oracle Database Snapshot Technology on Exadata Machine.

Because Exadata Storage Cell Smart Features, Storage Indexes, IORM and Network Resource Manager work at level of ASM Volume Manager only, (and they don’t work on top of ACFS Cluster File System), the implementation of the snapshot technology is different compared to any other non-Exadata environment.

At this purpuse Oracle has developed a new type of ASM Disk Group called SPARSE Disk Group. It uses ASM SPARSE Grid Disk based on Thin Provisioning to save the database snapshot copies and the associated metadata, and it supports non-CDB and PDB snapshot copy.

The implementation requires the following minimal software versions :

  • Exadata Storage Software version 12.1.2.1.0.
  • Oracle Database version 12.1.0.2 with bundle patch 5.
One major restriction applies to Exadata Storage Sanpshot compared to ACFS;
the source database must be a shared copy open on read only and called Test Master. The Test Master Database can not be modified or deleted as long the latest child snapshot is in use.
This restriction exists because Exadata Snapshot technology uses “allocate on first write”, and not “copy on write” (like for ACFS), and the snapshot is per-database-datafile.
When a child snapshot issue a write, the write goes to a private copy of that block inside the snapshot, preserving the original block value which can be accessed by other child snapshots of the same Test Master.

How to Implement Exadata Storage Snapshots in a PDB Environment

Check the celldisks for available free space to allocate to a new SPARSE Disk Group

[root@strgceladm01 ~]# cellcli -e list celldisk attributes name,freespace
 CD_00_strgceladm01 853.34375G
 CD_01_strgceladm01 853.34375G
 CD_02_strgceladm01 853.34375G
 CD_03_strgceladm01 853.34375G
 CD_04_strgceladm01 853.34375G
 CD_05_strgceladm01 853.34375G
 CD_06_strgceladm01 853.34375G
 CD_07_strgceladm01 853.34375G
 CD_08_strgceladm01 853.34375G
 CD_09_strgceladm01 853.34375G
 CD_10_strgceladm01 853.34375G
 CD_11_strgceladm01 853.34375G
 FD_00_strgceladm01 0
 FD_01_strgceladm01 0
 FD_02_strgceladm01 0
 FD_03_strgceladm01 0
[root@strgceladm01 ~]#


[root@strgceladm02 ~]# cellcli -e list celldisk attributes name,freespace
 CD_00_strgceladm02 853.34375G
 CD_01_strgceladm02 853.34375G
 CD_02_strgceladm02 853.34375G
 CD_03_strgceladm02 853.34375G
 CD_04_strgceladm02 853.34375G
 CD_05_strgceladm02 853.34375G
 CD_06_strgceladm02 853.34375G
 CD_07_strgceladm02 853.34375G
 CD_08_strgceladm02 853.34375G
 CD_09_strgceladm02 853.34375G
 CD_10_strgceladm02 853.34375G
 CD_11_strgceladm02 853.34375G
 FD_00_strgceladm02 0
 FD_01_strgceladm02 0
 FD_02_strgceladm02 0
 FD_03_strgceladm02 0
[root@strgceladm02 ~]#


[root@strgceladm03 ~]# cellcli -e list celldisk attributes name,freespace
 CD_00_strgceladm03 853.34375G
 CD_01_strgceladm03 853.34375G
 CD_02_strgceladm03 853.34375G
 CD_03_strgceladm03 853.34375G
 CD_04_strgceladm03 853.34375G
 CD_05_strgceladm03 853.34375G
 CD_06_strgceladm03 853.34375G
 CD_07_strgceladm03 853.34375G
 CD_08_strgceladm03 853.34375G
 CD_09_strgceladm03 853.34375G
 CD_10_strgceladm03 853.34375G
 CD_11_strgceladm03 853.34375G
 FD_00_strgceladm03 0
 FD_01_strgceladm03 0
 FD_02_strgceladm03 0
 FD_03_strgceladm03 0
[root@strgceladm03 ~]#

For each Storage Cell Create a SPARSE Grid Disks as described below

[root@strgceladm01 ~]# cellcli -e CREATE GRIDDISK ALL PREFIX=SPARSE, sparse=true, SIZE=853.34375G
Cell disks were skipped because they had no freespace for grid disks: FD_00_strgceladm01, FD_01_strgceladm01, FD_02_strgceladm01, FD_03_strgceladm01.
GridDisk SPARSE_CD_00_strgceladm01 successfully created
GridDisk SPARSE_CD_01_strgceladm01 successfully created
GridDisk SPARSE_CD_02_strgceladm01 successfully created
GridDisk SPARSE_CD_03_strgceladm01 successfully created
GridDisk SPARSE_CD_04_strgceladm01 successfully created
GridDisk SPARSE_CD_05_strgceladm01 successfully created
GridDisk SPARSE_CD_06_strgceladm01 successfully created
GridDisk SPARSE_CD_07_strgceladm01 successfully created
GridDisk SPARSE_CD_08_strgceladm01 successfully created
GridDisk SPARSE_CD_09_strgceladm01 successfully created
GridDisk SPARSE_CD_10_strgceladm01 successfully created
GridDisk SPARSE_CD_11_strgceladm01 successfully created
[root@strgceladm01 ~]#

For each Storage Cell List all Grid Disks

[root@strgceladm01 ~]# cellcli -e list griddisk attributes name,size
 DATAC1_CD_00_strgceladm01 6.294586181640625T
 DATAC1_CD_01_strgceladm01 6.294586181640625T
 DATAC1_CD_02_strgceladm01 6.294586181640625T
 DATAC1_CD_03_strgceladm01 6.294586181640625T
 DATAC1_CD_04_strgceladm01 6.294586181640625T
 DATAC1_CD_05_strgceladm01 6.294586181640625T
 DATAC1_CD_06_strgceladm01 6.294586181640625T
 DATAC1_CD_07_strgceladm01 6.294586181640625T
 DATAC1_CD_08_strgceladm01 6.294586181640625T
 DATAC1_CD_09_strgceladm01 6.294586181640625T
 DATAC1_CD_10_strgceladm01 6.294586181640625T
 DATAC1_CD_11_strgceladm01 6.294586181640625T
 FGRID_FD_00_strgceladm01 2.0717315673828125T
 FGRID_FD_01_strgceladm01 2.0717315673828125T
 FGRID_FD_02_strgceladm01 2.0717315673828125T
 FGRID_FD_03_strgceladm01 2.0717315673828125T
 RECOC1_CD_00_strgceladm01 1.78143310546875T
 RECOC1_CD_01_strgceladm01 1.78143310546875T
 RECOC1_CD_02_strgceladm01 1.78143310546875T
 RECOC1_CD_03_strgceladm01 1.78143310546875T
 RECOC1_CD_04_strgceladm01 1.78143310546875T
 RECOC1_CD_05_strgceladm01 1.78143310546875T
 RECOC1_CD_06_strgceladm01 1.78143310546875T
 RECOC1_CD_07_strgceladm01 1.78143310546875T
 RECOC1_CD_08_strgceladm01 1.78143310546875T
 RECOC1_CD_09_strgceladm01 1.78143310546875T
 RECOC1_CD_10_strgceladm01 1.78143310546875T
 RECOC1_CD_11_strgceladm01 1.78143310546875T
 SPARSE_CD_00_strgceladm01 853.34375G
 SPARSE_CD_01_strgceladm01 853.34375G
 SPARSE_CD_02_strgceladm01 853.34375G
 SPARSE_CD_03_strgceladm01 853.34375G
 SPARSE_CD_04_strgceladm01 853.34375G
 SPARSE_CD_05_strgceladm01 853.34375G
 SPARSE_CD_06_strgceladm01 853.34375G
 SPARSE_CD_07_strgceladm01 853.34375G
 SPARSE_CD_08_strgceladm01 853.34375G
 SPARSE_CD_09_strgceladm01 853.34375G
 SPARSE_CD_10_strgceladm01 853.34375G
 SPARSE_CD_11_strgceladm01 853.34375G
[root@strgceladm01 ~]#

From ASM Instance Create a SPARSE Disk Group

SQL> CREATE DISKGROUP SPARSEC1 EXTERNAL REDUNDANCY DISK 'o/*/SPARSE_CD_*'
ATTRIBUTE
'compatible.asm' = '12.2.0.1',
'compatible.rdbms' = '12.2.0.1',
'cell.smart_scan_capable'='TRUE',
'cell.sparse_dg' = 'allsparse',
'AU_SIZE' = '4M';

Diskgroup created.

Set the following ASM attributes on the Disk Group hosting the Test Master Database

ALTER DISKGROUP DATAC1 SET ATTRIBUTE 'access_control.enabled' = 'true';

Grant access to the OS RDBMS user used to access to the Disk Group

ALTER DISKGROUP DATAC1 ADD USER 'oracle';

From an ASM Instance Set ownership permissions for every file that belongs solely to the PDB being snapped cloned as per example below

alter diskgroup DATAC1 set ownership owner='oracle' for file '+DATAC1/CDBT/<xxxxxxxxxxxxxxxxxxx>/DATAFILE/system.xxx.xxxxxxx';
alter diskgroup DATAC1 set ownership owner='oracle' for file '+DATAC1/CDBT/<xxxxxxxxxxxxxxxxxxx>/DATAFILE/sysaux.xxx.xxxxxxx';
alter diskgroup DATAC1 set ownership owner='oracle' for file '+DATAC1/CDBT/<xxxxxxxxxxxxxxxxxxx>/DATAFILE/users.xxx.xxxxxxx';
...
..

Restart the Master Test PDB in Read Only

alter pluggable database PDBTESTMASTER close immediate instances=all;
alter pluggable database PDBTESTMASTER open read only;

Create the first PDB Snapshot Copy on Exadata SPARSE Disk Group

Create pluggable database PDBDEV01 from PDBTESTMASTER tempfile reuse create_file_dest='+SPARSEC1' snapshot copy;

Feedback of the Exadata Storage Snapshots

The ability to create storage efficient database copies in a few seconds, independently from the size of the Test Master is very useful for today IT departments; but such extreme velocity and flexibility is not entirely free. In fact performance tests on a I/O bound workload have highlighted important performance degradation. This reminds us that as defined by Oracle Corporation, the Snapshot Technology, included on Exadata Machine remains a non-production option.

Advertisements

Feedback of Modern Consolidated Database Environment

 

Since the launch of Oracle 12c R1 Beta Program (August 2012) at Trivadis, we have been intensively testing, engineering and implementing Multitenant architectures for our customers.

Today, we can provide our feedbacks and those of our customers!

The overall feedback related to Oracle Multitenant is very positive, customers have been able to increase flexibility and automation, improving the efficiency of the software development life cycles.

Even the Single-tenant configuration (free of charge) brings few advantages compared to the non-CDB architecture. Therefore, from a technology point of view I recommend adopting the Container Database (CDB) architecture for all Oracle databases.

 

Examples of Multitenant architectures implemented

Having defined Oracle Multitenant a technological revolution on the space of relational databases, when combined with others 12c features it becomes a game changer for flexibility, automation and velocity.

Here are listed few examples of successful architectures implemented with our customers, using Oracle Container Database (CDB):

 

  • Database consolidation without performance and stability compromise here.

 

  • Multitenant and DevOps here.

 

  • Operating Database Disaster Recovery in Multitenant environment here.

 

 


 

Oracle Multitenant supports database DevOps standards

As a consultant I constantly speak with my customers, and among a big number of them I noticed that the speed and flexibility of all database provisioning activities generate huge concern.

Hence I decide to describe on this post few Oracle Multitenant options to resolve those problems.

If production is the most critical environment to maintain, it is definetly not the one generating the greatest efforts in term of provisioning. The applications are more and more complex, and require continuous delivery;  to satify those needs the infrastructure has few provisioning challengers to overcome.

Now with the Oracle version 12.2 and the Mutitenet option, the DBaaS model becomes simpler than ever.

 

Clone PDB

The Clone PDB operation has been enhanced from Cold to Hot Clone. This improvement requires the usage of  PDB Local Undo. The Hot Clone is now the default method and can be devided in three major steps:

  1. PDB source datafile copy, because the PDB remains open in read/write at this stage the cloned datafiles are physically inconsistent (fuzzy data files).
  2. The Redo Log entries generated on the source PDB during the copy are applied to the targed PDB. This step makes the source and target PDBs two exact physical copies.
  3. Because the Redo Log entries coming from the source PDB contain committed and uncommitted transactions, to make the target PDB transactionally consistent, the undo entries of all uncommitted transations must be applied.

 

The command below shows how to clone a PDB open in read/write:

Create Pluggable Database ERP_Hot_Clone from ERP;

 

Refreshable PDB

Refreshable PDB leverages the Hot Clone PDB capability, creating an initial copy of the source PDB refreshed over the time at scheduled interval or on-demand.

To better understand the possible use cases, the graphical example below covers the development’s request to have every morning a copy of production data.

 

Refreshabe_PDB_all.png

 

How to create a Refreshable PDB

Syntax to create an automatic refreshable PDB:

Create Pluggable Database CRM_Test from CRM_Prod@db_link refresh mode every 720; -- (12H)

 

Syntax to create a manual PDB refresh:

Create Pluggable Database CRM_Test from CRM_Prod@db_link refresh mode manual;

 

After the clone the refreshable PDB should then be opened in read-only:

Alter Pluggable Database CRM_Test read only;

 

How to invoke a manual PDB refresh:

Alter Pluggable Database CRM_Test refresh;

 

Creation of the snapshot databases:

Create Pluggable Database CRM_TEST_Snap01 FROM CRM_Test
FILE_NAME_CONVERT = ('/u03/oradata/CDB122/CRM_Test/','/u03/oradata/CDB122/CRM_Test_Snap01/')
SNAPSHOT COPY;

 

 


 

 

New Oracle version (12.2.0.1) old BUG!

 

In June 2016 I posted the following BUG: Bug on Oracle 12c Multitenant & PDB Clone as Snapshot Copypromising to post an update once the version 12cR2 is available, because in the service request, originally opened with the version 12.1.0.2 Oracle stated that the bug would be fixed in 12cR2.

I was so impatient, that just few hours after the general availability of the Oracle Database 12c Release 2  I created a new cluster and tested the resolution.

 

For the record, it states that the resolution of this bug is important for one of my clients, where we have implemented the snapshot PDB on the application development lifecycle.

 

So let’s see if the bug has been fixed!

SQL*Plus: Release 12.2.0.1.0 Production on Wed Mar 1 21:06:54 2017

Copyright (c) 1982, 2016, Oracle. All rights reserved.

Connected to:
Oracle Database 12c Enterprise Edition Release 12.2.0.1.0 - 64bit Production


SQL> CREATE PLUGGABLE DATABASE PDBACFS1_SNAP1 from PDBACFS1 SNAPSHOT COPY;

Pluggable database created.

SQL> ALTER PLUGGABLE DATABASE PDBACFS1_SNAP1 OPEN instances=all;

Pluggable database altered.

SQL> select CON_ID, NAME, OPEN_MODE, SNAPSHOT_PARENT_CON_ID from v$pdbs where NAME in ('PDBACFS1','PDBACFS1_SNAP1');

 CON_ID     NAME               OPEN_MODE  SNAPSHOT_PARENT_CON_ID
---------- ------------------- ---------- ----------------------
 5          PDBACFS1            READ WRITE
 6          PDBACFS1_SNAP1      READ WRITE               <-- This should be 5 but is NULL

2 rows selected.

 

To a certain point of view progress has been made, in version 12.1.0.2 the column SNAPSHOT_PARENT_CON_ID was always zero (0) now is null!

I’m sorry for my customer, I’ll keep testing hoping …

 

 

Oracle DB stored on ASM vs ACFS

Nowadays a new Oracle database environment with Grid Infrastructure has three main storage options:

  1. Third party clustered file system
  2. ASM Disk Groups
  3. ACFS File System

While the first option was not in scope, this blog compares the result of the tests between ASM and ACFS, highlighting when to use one or the other to store 12c NON-CDB or CDB Databases.

The tests conducted on different environments using Oracle version 12.1.0.2 July PSU have shown controversial results compared to what Oracle  is promoting for the Oracle Database Appliance (ODA) in the following paper: “Frequently Asked Questions Storing Database Files in ACFS on Oracle Database Appliance

 

Outcome of the tests

ASM remains the preferred option to achieve the best I/O performance, while ACFS introduces interesting features like DB snapshot to quickly and space efficiently provision new databases.

The performance gap between the two solutions is not negligible as reported below by the  AWR – TOP Timed Events sections of two PDBs, sharing the same infrastructure, executing the same workload but respectively using ASM and ACFS storage:

  • PDBASM: Pluggable Database stored on  ASM Disk Group
  • PDBACFS:Pluggable Database stored on ACFS File System

 

 

PDBASM AWR – TOP Timed Events and Other Stats

topevents_asm

fg_asm

 

 

PDBACFS AWR – TOP Timed Events and Other Stats

TopEvents_ACFS.png

fg_acfs

 

Due to the different characteristics and results when ASM or ACFS is in use, it is not possible to give a generic recommendation. But case by case the choise should be driven by business needs like maximum performance versus fast and efficient database clone.

 

 

 

 

Bug on Oracle 12c Multitenant & PDB Clone as Snapshot Copy

While automating the refresh of the test databases on Oracle 12c Multitenant environment with ACFS and PDB snapshot copy, I encountered the following BUG:

The column SNAPSHOT_PARENT_CON_ID of the view V$PDBS shows 0 (zero) in case of PDBs created as Snapshot Copy.

This bug prevents to identify the parent-child relationship between a PDB and its own Snapshots Copies.

The test case below explains the problem:

SQL> CREATE PLUGGABLE DATABASE LARTE3SEFU from LARTE3 SNAPSHOT COPY; 
 
 Pluggable database created. 
 
 SQL> select CON_ID, NAME, OPEN_MODE, SNAPSHOT_PARENT_CON_ID from v$pdbs where NAME in ('LARTE3SEFU','LARTE3'); 
 
 CON_ID      NAME          OPEN_MODE  SNAPSHOT_PARENT_CON_ID 
 ---------- -------------- ---------- ---------------------- 
 5          LARTE3         READ ONLY  0 
 16         LARTE3SEFU     MOUNTED    0  <-- This should be 5
 
 2 rows selected. 

A Service Request to Oracle has been opened, I’ll update this post once I have the official answer.

Update from the Service Request: BUG Fixed on version 12.2

How to Create and Clone PDBs

################################################
## How to create a PDB Database from Seed DB  ##
################################################

CREATE PLUGGABLE DATABASE pdb01
  ADMIN USER pdb_adm IDENTIFIED BY <password> ROLES=(DBA)
  PATH_PREFIX = '/u01/'
  STORAGE (MAXSIZE 20G MAX_SHARED_TEMP_SIZE 2048M)
  FILE_NAME_CONVERT = ('+DATA01','+DATA02')
  DEFAULT TABLESPACE users DATAFILE '+DATA02' SIZE 10G AUTOEXTEND ON MAXSIZE 20G
  TEMPFILE REUSE;

ALTER PLUGGABLE DATABASE pdb01 OPEN;  
 


 
##################################################
## How to clone a PDB Database running on ASM   ##
##################################################

ALTER PLUGGABLE DATABASE pdb01 CLOSE;  
ALTER PLUGGABLE DATABASE pdb01 OPEN READ ONLY;

CREATE PLUGGABLE DATABASE pdb02 FROM pdb01;

ALTER PLUGGABLE DATABASE pdb01 OPEN READ WRITE;
ALTER PLUGGABLE DATABASE pdb02 OPEN READ WRITE;

 
 
 
##################################################
## How to clone a PDB Database using ACFS Snapshot Copy
##################################################
 
ALTER PLUGGABLE DATABASE pdb03 CLOSE;
ALTER PLUGGABLE DATABASE pdb03 OPEN READ ONLY;
 
 
CREATE PLUGGABLE DATABASE pdb04 FROM pdb03
FILE_NAME_CONVERT = ('/u03/oradata/CDB2/pdb03/','/u03/oradata/CDB2/pdb04/')
SNAPSHOT COPY;

ALTER PLUGGABLE DATABASE pdb03 CLOSE;
ALTER PLUGGABLE DATABASE pdb03 OPEN READ WRITE;
ALTER PLUGGABLE DATABASE pdb04 OPEN READ WRITE;