Skip to main content

How to setup HugePages on Oracle Linux 64-bit


<<Back to PT Main Page

Steps to Configure HugePages on Linux 64-bit for Oracle Database

Step0: Check if Transparent  HugePages are enabled
cat /sys/kernel/mm/transparent_hugepage/enabled
[always] madvise never


Starting from RHEL6/OL6, Transparent HugePages are implemented and enabled by default. Therefore Oracle recommends disabling Transparent HugePages on all servers running Oracle databases this MOS note (Doc ID 1557478.1)

Step1: Verify the current OS Page Size
#grep Hugepagesize /proc/meminfo


Step2:Configure memlock
Set the memlock user limit in /etc/security/limits.conf file. Set the value (in KB) slightly smaller than installed RAM. e.g. If you have 450GB RAM installed, you may set:
oracle soft memlock 419430400
oracle hard memlock 419430400


NOTE: There is no harm in setting this value larger than your SGA requirements.

Step3: Verify the memlock limit
Re-logon to the Oracle product owner account (e.g. 'oracle') and check the memlock limit
#ulimit -l
419430400


Step4: AMM if already Configured
Disable Automatic Memory Management (AMM) Feature (if already enabled)
SQL>create pfile from spfile;
SQL>alter system set MEMORY_TARGET=0 sope=spfile;
SQL>alter system set MEMORY_MAX_TARGET=0 sope=spfile;


Step5: Calculate Number of HugePages
calculate the recommended value for the vm.nr_hugepageskernel parameter. I calculated the following value for 400GB of SGA.

kernel.shmmax=429496729600
vm.nr_hugepages=204800
vm.hugetlb_shm_group=<gid of dba group>

This is how you calculate the parameter value used in step2 & 5
kernel.shmmax = 429496729600  # 1024 * 1024 * 1024 * 400
kernel.shmall    = 117964800        #  use formula echo "450 * 1024^3 / 4096" | bc
The number of HugePages = shmmax [in KB] / page size [in KB].

Step6:Adopt the parameter value in /etc/sysctl.conf file
Edit the file /etc/sysctl.conf and set the
kernel.shmmax=429496729600
vm.nr_hugepages=204800
vm.hugetlb_shm_group=3500

Step7: Set the desired SGA size


Step8: Adopt kernel.shmall  /etc/sysctl.conf

Set SHMALL equal to the sum of all the SGAs on the system, divided by the page size.

kernel.shmall = 483183820800
If this parameter not sized properly you will encounter following error while starting up the database instance
ORA-27102: out of memory Linux-X86_64 Error: 28: No space left on device

Step9: Restart the Server and Database


  • Stop Database
  • Restart the server
  • Start the database
Step10: Validate the HugePage Configuration

#grep HugePages /proc/meminfo
AnonHugePages:    155648 kB
HugePages_Total:   209920
HugePages_Free:    199991
HugePages_Rsvd:    194872
HugePages_Surp:        0


Database Aletlog will also display the usages of Huge Pages


Starting ORACLE instance (normal)
************************ Large Pages Information *******************
Per process system memlock (soft) limit = 410 GB

Total Shared Global Region in Large Pages = 400 GB (100%)
Large Pages used by this instance: 204801 (400 GB)
Large Pages unused system wide = 5119 (10 GB)
Large Pages configured system wide = 209920 (410 GB)
Large Page size = 2048 KB





Why Use HugePage ?:
By default Linux Kernel manages the memory with 4K Page size, With hugePages you can increase the page size. As you see I configured it 2MB.
You can configure the page size up to 1G if supported by the hardware.
Now Linux kernel maintains the page table to manage the pages. Smaller page size increases the page numbers resulting larger page tables, which requires more memory & pointers to store and manage those pages and hence reducing the overall performance.
With larger page size you got less pages , resulting smaller page table and in tern better overall performance

When Use HugePages?
In general and to my experience if you have SGA>16G you should use HugePages and you will benefit the performance benefits. Oracle suggests using huge pages starting with 8G SGA itself.

Where can we Use HugePages?
All Linux/Unix Platform. Not sure about windows


Comments

Popular posts from this blog

ORA-28374: typed master key not found in wallet

<<Back to Oracle DB Security Main Page ORA-46665: master keys not activated for all PDBs during REKEY SQL> ADMINISTER KEY MANAGEMENT SET KEY FORCE KEYSTORE IDENTIFIED BY xxxx WITH BACKUP CONTAINER = ALL ; ADMINISTER KEY MANAGEMENT SET KEY FORCE KEYSTORE IDENTIFIED BY xxxx WITH BACKUP CONTAINER = ALL * ERROR at line 1: ORA-46665: master keys not activated for all PDBs during REKEY I found following in the trace file REKEY: Create Key in PDB 3 resulted in error 46658 *** 2019-02-06T15:27:04.667485+01:00 (CDB$ROOT(1)) REKEY: Activation of Key AdnU5OzNP08Qv1mIyXhP/64AAAAAAAAAAAAAAAAAAAAAAAAAAAAA in PDB 3 resulted in error 28374 REKEY: Keystore needs to be restored from the REKEY backup.Aborting REKEY! Cause: All this hassle started because I accidently deleted the wallet and all wallet backup files too and also forgot the keystore password. There was no way to restore the wallet back. Fortunately in my case the PDB which had encrypted data was supposed to be deco...

How to Find VIP of an Oracle RAC Cluster

<<Back to Oracle RAC Main Page How to Find Out VIP of an Oracle RAC Cluster Login clusterware owner (oracle) and execute the below command to find out the VIP hostname used in Oracle RAC $ olsnodes -i node1     node1-vip node2     node2-vip OR $ srvctl config nodeapps -viponly Network 1 exists Subnet IPv4: 10.0.0.0/255.255.0.0/bondeth0, static Subnet IPv6: Ping Targets: Network is enabled Network is individually enabled on nodes: Network is individually disabled on nodes: VIP exists: network number 1, hosting node node1 VIP Name: node1-vip VIP IPv4 Address: 10.0.0.1 VIP IPv6 Address: VIP is enabled. VIP is individually enabled on nodes: VIP is individually disabled on nodes: VIP exists: network number 1, hosting node node2 VIP Name: node2-vip VIP IPv4 Address: 10.0.0.2 VIP IPv6 Address: VIP is enabled. VIP is individually enabled on nodes: VIP is individually disabled on nodes:

ORA-65104: operation not allowed on an inactive pluggable database alter pluggable database open

<<Back to DB Administration Main Page ORA-65104: operation not allowed on an inactive pluggable database SQL> alter pluggable database TEST_CLON open; alter pluggable database TEST_CLON open * ERROR at line 1: ORA-65104: operation not allowed on an inactive pluggable database Cause The pluggable database status was UNUSABLE. It was still being created or there was an error during the create operation. A PDB can only be opened if it is successfully created and its status is marked as NEW in cdb_pdbs.status column SQL> select PDB_NAME,STATUS from cdb_pdbs; PDB_NAME             STATUS -------------------- --------------------------- PDB$SEED             NORMAL TEST_CLON            UNUSABLE Solution:  Drop the PDB and create it again. Related Posts How to Clone Oracle PDB (Pluggable Database) with in the Same Container

ORA-46630: keystore cannot be created at the specified location

<<Back to DB Administration Main Page ORA-46630: keystore cannot be created at the specified location CDB011> ADMINISTER KEY MANAGEMENT CREATE KEYSTORE '+DATAC4/CDB01/wallet/' IDENTIFIED BY "xxxxxxx"; ADMINISTER KEY MANAGEMENT CREATE KEYSTORE '+DATAC4/CDB01/wallet/' IDENTIFIED BY "EncTest123" * ERROR at line 1: ORA-46630: keystore cannot be created at the specified location Cause  Creating a keystore at a location where there is already a keystore exists Solution To solve the problem, use a different location to create a keystore (use ENCRYPTION_WALLET_LOCATION in sqlnet.ora file to specify the keystore location), or move this ewallet.p12 file to some other location. Note: Oracle does not recommend deleting keystore file (ewallet.p12) that belongs to a database. If you have multiple keystores, you can choose to merge them rather than deleting either of them.

ORA-16905: The member was not enabled yet

<<Back to Oracle DataGuard Main Page ORA-16905 Physical Standby Database is disabled DGMGRL> show configuration; Configuration - DG_ORCL1P   Protection Mode: MaxPerformance   Members:   ORCL1PP - Primary database     ORCL1PS - Physical standby database (disabled)       ORA-16905: The member was not enabled yet. Fast-Start Failover:  Disabled Configuration Status: SUCCESS   (status updated 58 seconds ago) DGMGRL> DGMGRL> enable database 'ORCL1PS'; Enabled. DGMGRL>  show configuration; Configuration - DG_ORCL1P   Protection Mode: MaxPerformance   Members:   ORCL1PP - Primary database     ORCL1PS - Physical standby database Fast-Start Failover:  Disabled Configuration Status: SUCCESS   (status updated 38 seconds ago)

How to Switch Log File from All Instances in RAC

<<Back to Oracle RAC Main Page Switch The Log File of All Instances in Oracle RAC. In many cases you need to switch the logfile of the database. You can switch logfile using alter system switch logfile command but if you want to switch the logfile from all the instances you need to execute the command on all the instances individually and therefore you must login on all the instances. You can avoid this and switch logfile of all instances by just running the below command from any of the instance in RAC database SQL> ALTER SYSTEM SWITCH ALL LOGFILE;   System altered.

ORA-46655: no valid keys in the file from which keys are to be imported

<<Back to DB Administration Main Page SQL> administer key management import encryption keys with secret "xxxx" from '/tmp/pdb02_tde_key.exp' force keystore identified by "xxxx" with backup; administer key management import encryption keys with secret "xxxxxx" from '/tmp/pdb02_tde_key.exp' force keystore identified by "xxxxxx" with backup * ERROR at line 1: ORA-46655: no valid keys in the file from which keys are to be imported Cause: Either the keys to be imported already present in the target database or correct container (PDB) is not set. Solution: In my case I got the error because I attempted to import the keys for newly plugged database PDB02 from CDB$ROOT container. To Solve the issue just switched to the correct container and re run the import. SQL> show con_name CON_NAME ------------------------------ CDB$ROOT <===Wrong Container selected  SQL> alter session set container=PDB02; Session alt...

Starting RMAN and connecting to Database

  <<Back to Oracle Backup & Recovery Main Page Starting RMAN and connecting to Database Starting RMAN and connecting to Database To start RMAN you need to set the environment and type rman and press enter. You can connect to database either using connect command or using command line option. using command line option localhost:$ export ORACLE_HOME=/ora_app/product/18c/dbd2 localhost:$ export PATH=$ORACLE_HOME/bin:$PATH localhost:$ export ORACLE_SID=ORCL1P localhost:$ rman target / Recovery Manager: Release 18.0.0.0.0 - Production on Sun Apr 4 08:11:01 2021 Version 18.11.0.0.0 Copyright (c) 1982, 2018, Oracle and/or its affiliates.  All rights reserved. connected to target database: ORCL1P (DBID=4215484517) RMAN> using connect option localhost:$ rman RMAN> connect target sys@ORCL1P  target database Password:******** connected to target database: ORCL1P (DBID=4215484517) NOTE: To use connect command you need to ensure that  you have proper TNS sentry...

How to Modify Database Startup Mode (Startoption) in Clusterware Registry

<<Back to Oracle RAC Main Page How to Modify Database Startup Option Using srvctl In this post I will show you, how you can modify the database startup option / Database startup mode in clusterware registry using srvctl. There are some time requirement to change the startup option (from default  OPEN ) to  MOUNT, or "READ ONLY" eg if in case of Physical Standby Configuration Let us Check the Current Configuration $ srvctl config database -d ORCL Database unique name: ORCL Database name: Oracle home: /u01/app/oracle/product/12.1.0.2/dbhome_2 Oracle user: oracle Spfile: +DATA/ORCL/PARAMETERFILE/spfileORCL.ora Password file: +DATA/ORCL/PASSWORD/pwORCL Domain: Start options: open Stop options: immediate Database role: PHYSICAL_STANDBY Management policy: AUTOMATIC Server pools: Disk Groups: DATA,RECO Mount point paths: Services: Type: RAC Start concurrency: Stop concurrency: OSDBA group: oinstall OSOPER group: oinstall Database instances: ORCL1,...

cluvfy Pre Check for RAC Oracle Installation (CVU)

<<Back to Oracle RAC Main Page runcluvfy.sh stage -pre crsinst -n node1,node2 -fixup -verbose Cluster Verification Pre Installation check failed. Verifying Physical Memory ...FAILED (PRVF-7530) $./runcluvfy.sh stage -pre crsinst -n node1,node2 -fixup -verbose Verifying Physical Memory ... Node Name     Available                 Required                  Status   ------------  ------------------------  ------------------------  ----------   node2         7.5443GB (7910784.0KB)    8GB (8388608.0KB)         failed   node1         7.5443GB (7910784.0KB)    8GB (8388608.0KB)      ...