A first look at RAC 12c (part I)

I have recently upgraded my RAC 12.1.0.1.3 system to RAC 12.1.0.2 including the RDBMS installation. Currently I am updating my skills with information relevant to what I would normally have called 12c Release 2 (so that would also answer the question: when is 12c Release 2 coming out?). Then I realised I haven’t posted a first look at RAC 12c post yet-so here it comes.

There are a few things that aren’t specifically mentioned in the new features guide that caught my eye. First of all, RAC 12 does a few really cool things. Have a look at the srvctl command output:

[oracle@rac12node1 ~]$ srvctl
Usage: srvctl <command> <object> [<options>]
    commands: enable|disable|export|import|start|stop|relocate|status|add|remove|modify|getenv|setenv|unsetenv|
       config|convert|update|upgrade|downgrade|predict
    objects: database|instance|service|nodeapps|vip|network|asm|diskgroup|listener|srvpool|server|scan|scan_listener|
        oc4j|home|filesystem|gns|cvu|havip|exportfs|rhpserver|rhpclient|mgmtdb|mgmtlsnr|volume|mountfs
For detailed help on each command and object and its options use:
  srvctl <command> -help [-compatible] or
  srvctl <command> <object> -help [-compatible]
[oracle@rac12node1 ~]$

Quite a few more than with 11.2.0.3:

[oracle@rac112node1 ~]$ srvctl
Usage: srvctl <command> <object> [<options>]
 commands: enable|disable|start|stop|relocate|status|add|remove|modify|getenv|setenv|unsetenv|config|convert|upgrade
 objects: database|instance|service|nodeapps|vip|network|asm|diskgroup|listener|srvpool|server|scan|scan_listener|oc4j|home|filesystem|gns|cvu
For detailed help on each command and object and its options use:
 srvctl <command> -h or
 srvctl <command> <object> -h

I will detail the meaning of some of these later in this post or another one to follow.

Evaluation and Prediction

When you are working with policy managed databases RAC 12c already gave you a “what if” option in form of the -eval flag. If for example you wanted to grow your server pool from 2 to 3 nodes:

[oracle@rac12node1 ~]$ srvctl modify srvpool -serverpool pool1 -max 3 -eval -verbose
Database two will be started on node rac12node3
Server rac12node3 will be moved from pool Free to pool pool1
[oracle@rac12node1 ~]$

Now you will be able to predict a resource failure as well:

[oracle@rac12node1 ~]$ srvctl predict -h

The SRVCTL predict command evaluates the consequences of resource failure.

Usage: srvctl predict database -db <database_name> [-verbose]
Usage: srvctl predict service -db <database_name> -service <service_name> [-verbose]
Usage: srvctl predict asm [-node <node_name>] [-verbose]
Usage: srvctl predict diskgroup -diskgroup <diskgroup_name> [-verbose]
Usage: srvctl predict filesystem -device <volume_device> [-verbose]
Usage: srvctl predict vip -vip <vip_name> [-verbose]
Usage: srvctl predict network [-netnum <network_number>] [-verbose]
Usage: srvctl predict listener -listener <listener_name> [-verbose]
Usage: srvctl predict scan -scannumber <scan_ordinal_number> [-netnum <network_number>] [-verbose]
Usage: srvctl predict scan_listener -scannumber <scan_ordinal_number> [-netnum <network_number>] [-verbose]
Usage: srvctl predict oc4j [-verbose]

So what would happen if a disk group failed?

[oracle@rac12node1 ~]$ srvctl predict diskgroup -diskgroup DATA -verbose
Resource ora.DATA.dg will be stopped
Resource ora.DATA.ORAHOMEVOL.advm will be stopped
[oracle@rac12node1 ~]$

What it doesn’t do at this stage seems to be an assessment of cascading further problems. If +DATA went down, it would pretty much drag the whole cluster with it, too.

Status

Interestingly you can see a lot more detail with 12.1.0.2 than previously. Here is an example of a policy-managed RAC One Node database:

[oracle@rac12node1 ~]$ srvctl config database -d RONNCDB
Database unique name: RONNCDB
Database name: RONNCDB
Oracle home: /u01/app/oracle/product/12.1.0.2/dbhome_1
Oracle user: oracle
Spfile: +DATA/RONNCDB/PARAMETERFILE/spfile.319.854718651
Password file: +DATA/RONNCDB/PASSWORD/pwdronncdb.290.854718263
Domain:
Start options: open
Stop options: immediate
Database role: PRIMARY
Management policy: AUTOMATIC
Server pools: ronpool1
Disk Groups: RECO,DATA
Mount point paths:
Services: NCDB
Type: RACOneNode
Online relocation timeout: 30
Instance name prefix: RONNCDB
Candidate servers:
OSDBA group: dba
OSOPER group:
Database instances:
Database is policy managed

Did you spot the OSDBA and OSOPER group mappings in the output? DBCA by default creates the password file and server parameter file into ASM since 12.1.0.1.

You can get a lot more status information in 12.1.0.2 then previously, especially when compared to 11.2:

[oracle@rac12node1 ~]$ srvctl status -h

The SRVCTL status command displays the current state of the object.

Usage: srvctl status database {-db <db_unique_name> [-serverpool <serverpool_name>] | -serverpool <serverpool_name> | -thisversion | -thishome} [-force] [-verbose]
Usage: srvctl status instance -db <db_unique_name> {-node <node_name> | -instance <inst_name_list>} [-force] [-verbose]
Usage: srvctl status service {-db <db_unique_name> [-service  "<service_name_list>"] | -serverpool <serverpool_name> [-db <db_unique_name>]} [-force] [-verbose]
Usage: srvctl status nodeapps [-node <node_name>]
Usage: srvctl status vip {-node <node_name> | -vip <vip_name>} [-verbose]
Usage: srvctl status listener [-listener <lsnr_name>] [-node <node_name>] [-verbose]
Usage: srvctl status asm [-proxy] [-node <node_name>] [-detail] [-verbose]
Usage: srvctl status scan [[-netnum <network_number>] [-scannumber <scan_ordinal_number>] | -all] [-verbose]
Usage: srvctl status scan_listener [[-netnum <network_number>] [-scannumber <scan_ordinal_number>] | -all] [-verbose]
Usage: srvctl status srvpool [-serverpool <pool_name>] [-detail]
Usage: srvctl status server -servers "<server_list>" [-detail]
Usage: srvctl status oc4j [-node <node_name>] [-verbose]
Usage: srvctl status rhpserver
Usage: srvctl status rhpclient
Usage: srvctl status home -oraclehome <oracle_home> -statefile <state_file> -node <node_name>
Usage: srvctl status filesystem [-device <volume_device>] [-verbose]
Usage: srvctl status volume [-device <volume_device>] [-volume <volume_name>] [-diskgroup <group_name>] [-node <node_list> | -all]
Usage: srvctl status diskgroup -diskgroup <dg_name> [-node "<node_list>"] [-detail] [-verbose]
Usage: srvctl status cvu [-node <node_name>]
Usage: srvctl status gns [-node <node_name>] [-verbose]
Usage: srvctl status mgmtdb [-verbose]
Usage: srvctl status mgmtlsnr [-verbose]
Usage: srvctl status exportfs [-name <expfs_name> |-id <havip id>]
Usage: srvctl status havip [-id <id>]
Usage: srvctl status mountfs -name <mountfs_name>
For detailed help on each command and object and its options use:
  srvctl <command> <object> -help [-compatible]

RAC 12.1.0.2 adds a nifty few little flags: thisversion and thishome to srvctl status database. That works really well where you have multiple versions of Oracle on the same machine (think consolidation):

[oracle@rac12node1 ~]$ srvctl status database -thisversion
Database unique name: RONNCDB
Instance RONNCDB_1 is running on node rac12node4
Online relocation: INACTIVE

Database unique name: TWO
Instance TWO_1 is running on node rac12node1
Instance TWO_2 is running on node rac12node2

Verbose!

Some commands are actually more verbose when you specify the -verbose flag:

[oracle@rac12node1 ~]$ srvctl status database -d RONNCDB -verbose
Instance RONNCDB_1 is running on node rac12node4 with online services NCDB. Instance status: Open.
Online relocation: INACTIVE
[oracle@rac12node1 ~]$ srvctl status database -d RONNCDB
Instance RONNCDB_1 is running on node rac12node4
Online relocation: INACTIVE
[oracle@rac12node1 ~]$

But that’s not new in 12.1.0.2 I believe.

Interesting changes for database logging

The database itself will also tell you more about memory allocation:

**********************************************************************
Dump of system resources acquired for SHARED GLOBAL AREA (SGA)
 Per process system memlock (soft) limit = 64K
Thu Jul 31 13:34:58 2014
 Expected per process system memlock (soft) limit to lock
 SHARED GLOBAL AREA (SGA) into memory: 1538M
Thu Jul 31 13:34:58 2014
 Available system pagesizes:
  4K, 2048K
 Supported system pagesize(s):
  PAGESIZE  AVAILABLE_PAGES  EXPECTED_PAGES  ALLOCATED_PAGES  ERROR(s)
        4K       Configured               3          393219        NONE
     2048K                0             769               0        NONE

RECOMMENDATION:
 1. For optimal performance, configure system with expected number
 of pages for every supported system pagesize prior to the next
 instance restart operation.
 2. Increase per process memlock (soft) limit to at least 1538MB
 to lock 100% of SHARED GLOBAL AREA (SGA) pages into physical memory

As you can see I am not using large pages here at all, which I did for demonstration purposes only. I don’t see any reason not to use large pages on a 64bit system these days. I’m curious to see whether the AIX port supports all the AIX page sizes here.

End of part I

This has already turned into a longer post than I expected it to be when I started writing. I think I’ll continue the series in a couple of weeks when I’m finding the time.