Backing up a RAC database with all nodes

Backing up a RAC environment from multiple DB nodes

I've been working with many customers who have never performed a multi-node backup for a RAC database.

If you didn't already know, in a RAC environment you can backup from more than 1 DB node.

Backing up from multiple DB nodes can be very important for a large multi-terabyte database to give you faster backup times.


** Keep in mind these throughput numbers would be the same for restore capability.

PART 1 - WHY

Why would I want to go through the trouble of backing up on multiple nodes ?

When looking at backup rates, the number to keep in mind is the backup rate for each network speed.

1   Gb/s  == 350 GB/hr
10 Gb/s == 3.5 TB/hr
25 Gb/s == 8.8 TB/hr

**  Note that networks are measured in small "b" which is bits/seconds
**  Database size is measured in large "B" which is Bytes.

The chart below gives you an idea of the impact of adding active network ports, and how the speed of the ports affects the backup throughput.

Using this chart you can calculate how long it will take to backup your large database.

Here is the example of the number of hours it would take to backup a 50 T database..

PART 2 -  How ?

When performing a backup, the backup process uses channels.  I'm guessing you probably already knew that.  

Each allocated channel connects to a database instance and sends the backup through that channel to the backup location (disk, media manager, etc).

By default the allocated channel connects to the database instance that the RMAN process is executed from.

What this means is... if you are "inst_id 1" and execute a backup, all channels will be connected to "inst_id 1".

This can be changed within the channel allocation by using a channel parameter named "CONNECT".

The connect (as you expect) uses an ID and Password (or a wallet) to connect to a specific service, or by specifying an EZ Connect string.

For my example I am going to use a configure channel setting and a common service that is shared across all nodes and let it balance. the other options are

• Create a specific service to control which instances are available for backing up.  
• Manually allocate channels and specify the instance name in the connection.

Step 1 - Create a user for backing up

I'm going assume you are already on 12c, which contains the sysbackup role.

I'm going to create a user ,"backup_only", specifically for backing up, and I am going to use a complicated password.
Next,  I am going to grant that user sysbackup privlege.

SQL  create user backup_only identified by kkk3223490sf9090fjkljkl3590;

User created.

SQL   grant sysbackup to backup_only;

Grant succeeded.

Step 2 - Change the channel configuration to use this user

Now I am going to configure my channel to use this user for the connect clause.

Below is the statement I'm executing.  ** Notice I added the "as sysbackup" to the connection.

CONFIGURE CHANNEL DEVICE TYPE 'SBT_TAPE' connect 'backup_only/kkk3223490sf9090fjkljkl3590@bryan as sysbackup' FORMAT   '%d_%U' PARMS  "SBT_LIBRARY=/u01/app/oracle/product/12.1.0.2/dbhome_1/lib/libra.so, ENV=(RA_WALLET='location=file:/u01/app/oracle/product/12.1.0.2/dbhome_1/dbs/zdlra credential_alias=zdl2ing-scan:1521/zdlra:dedicated')";

This is for a ZDLRA so I have the additional information it needs for the library etc.

Notice that the I am including the id/password and tnsnames entry in the connection string.

This is what happens when I execute the configure command.

RMAN  
RMAN CONFIGURE CHANNEL DEVICE TYPE 'SBT_TAPE' connect * FORMAT   '%d_%U' PARMS  "SBT_LIBRARY=/u01/app/oracle/product/12.1.0.2/dbhome_1/lib/libra.so, ENV=(RA_WALLET='location=file:/u01/app/oracle/product/12.1.0.2/dbhome_1/dbs/zdlra credential_alias=zdl2ing-scan:1521/zdlra:dedicated')";
old RMAN configuration parameters:
CONFIGURE CHANNEL DEVICE TYPE 'SBT_TAPE' FORMAT   '%d_%U' PARMS  "SBT_LIBRARY=/u01/app/oracle/product/12.1.0.2/dbhome_1/lib/libra.so, ENV=(RA_WALLET='location=file:/u01/app/oracle/product/12.1.0.2/dbhome_1/dbs/zdlra credential_alias=zdl2ing-scan:1521/zdlra:dedicated')";
new RMAN configuration parameters:
CONFIGURE CHANNEL DEVICE TYPE 'SBT_TAPE' CONNECT '*' FORMAT   '%d_%U' PARMS  "SBT_LIBRARY=/u01/app/oracle/product/12.1.0.2/dbhome_1/lib/libra.so, ENV=(RA_WALLET='location=file:/u01/app/oracle/product/12.1.0.2/dbhome_1/dbs/zdlra credential_alias=zdl2ing-scan:1521/zdlra:dedicated')";
new RMAN configuration parameters are successfully stored
starting full resync of recovery catalog
full resync complete

One thing that stands out is that the connection information doesn't appear in the channel.  it is obfuscated by the '*' even though it is stored.

Step 3 - Execute backup

I can now execute my backup from either OEM, or from the command line.

This can be monitored (to ensure multiple nodes are used) by looking at the channel allocation in the output log of the backup.

Starting backup at 2018-08-08 09:27:17
allocated channel: ORA_SBT_TAPE_1
channel ORA_SBT_TAPE_1: SID=218 instance=bryan2 device type=SBT_TAPE
channel ORA_SBT_TAPE_1: RA Library (ZDLRA) SID=72EFFBF5EAF19225E05376AF880A0BA7
allocated channel: ORA_SBT_TAPE_2
channel ORA_SBT_TAPE_2: SID=245 instance=bryan2 device type=SBT_TAPE
channel ORA_SBT_TAPE_2: RA Library (ZDLRA) SID=72EFFBFF3A0E9235E05376AF880A1C6D
allocated channel: ORA_SBT_TAPE_3
channel ORA_SBT_TAPE_3: SID=615 instance=bryan1 device type=SBT_TAPE
channel ORA_SBT_TAPE_3: RA Library (ZDLRA) SID=72EFFC0907A51E40E05375AF880AA173
allocated channel: ORA_SBT_TAPE_4
channel ORA_SBT_TAPE_4: SID=299 instance=bryan2 device type=SBT_TAPE
channel ORA_SBT_TAPE_4: RA Library (ZDLRA) SID=72EFFC128EDD9244E05376AF880A6B6B
allocated channel: ORA_SBT_TAPE_5
channel ORA_SBT_TAPE_5: SID=64 instance=bryan2 device type=SBT_TAPE
channel ORA_SBT_TAPE_5: RA Library (ZDLRA) SID=72EFFC1BD2EF924AE05376AF880A68B2
allocated channel: ORA_SBT_TAPE_6
channel ORA_SBT_TAPE_6: SID=64 instance=bryan1 device type=SBT_TAPE
channel ORA_SBT_TAPE_6: RA Library (ZDLRA) SID=72EFFC2595C71E53E05375AF880AA8D1
allocated channel: ORA_SBT_TAPE_7
channel ORA_SBT_TAPE_7: SID=74 instance=bryan2 device type=SBT_TAPE
channel ORA_SBT_TAPE_7: RA Library (ZDLRA) SID=72EFFC2FA4979254E05376AF880AF834
allocated channel: ORA_SBT_TAPE_8

Above is the example from my output.. You can see the that there were 8 channels allocated, and 5 of the channels were on the first instance and 3 were allocated on the second instance.

That's all there is to it !!

Next blog post is going to be restoring across multiple nodes in a RAC cluster.