Linux

This section describes how to configure HugePage in Linux. Root permission is required.

Enabling HugePage

The following describes how to enable HugePage.

1. Check the size of HugePage supported in the current OS.

$ grep Hugepagesize /proc/meminfo 
Hugepagesize:	2048 KB

1. Check the user group ID that runs Tibero.

$ id -a
uid=579(tibero) gid=579(tibero) groups=579(tibero)

1. Apply the groups and number that will allocate HugePage in "/etc/sysctl.conf"

The following is an example of applying the kernel parameter when TOTAL_SHM_SIZE is set to 1024 MB and the size of HugePage supported by the current OS is 2 MB (2048 KB).

$ cat /etc/sysctl.conf
......Omitted......
vm.nr_hugepages=512 
vm.hugetlb_shm_group=579

1. Apply the maximum locked memory value in "/etc/security/limits.conf".

The following is an example of configuring the memlock value.

$ cat /etc/security/limits.conf
......Omitted......
tibero	soft	memlock    1048576
tibero	hard	memlock    1048576

1. Restart the operating system.

The modified HugePage value has been applied.

$ egrep -e HugePages_Total /proc/meminfo 
HugePages_Total: 512 KB

1. Configure the Tibero initialization parameters in the configuration file (.tip).

The following is an example of configuring the initialization parameters.

$ cat tibero.tip
......Omitted......
TOTAL_SHM_SIZE=1G 
USE_HUGE_PAGE=Y

1. Restart the Tibero server.

$ tbdown

Tibero instance terminated (NORMAL mode).

$ tbboot
Listener port = 8629
Change core dump dir to /home/tibero/tibero7/instance/tibero.

Tibero 7

TmaxData Corporation Copyright (c) 2008-. All rights reserved. 
Tibero instance started up (NORMAL mode).

Disabling HugePage

To disable HugePage, restore the values that were modified to enable HugePage to their original values. The modified values can be restored by the same process and order used to enable HugePage.

AIX

AIX uses Large Page instead of HugePage. The benefits of using Large Page are similar to those of HugePage.

Enabling HugePage

The following describes how to enable HugePage.

1. Change the Large Page configuration value of the OS.

AIX internally maintains physical memory pool sizes of 4 KB and 16 MB. The size of this pool can be changed to 16 MB by using the vmo command. The remaining space is automatically allocated to the 4 KB pool. From AIX 5.3 on, Large Page pools are dynamically maintained, so the system does not need to be restarted after changing the size.

First, v_pinshm must be separately configured so that the space in which the shared memory is allocated is not swapped to disk. percent_of_real_memory is the amount of memory that TSM possesses as a percentage of the total memory.

$ vmo r o v_pinshm=1
$ vmo r o maxpin%=percent_of_real_memory

Configure the Large Page pool size. num_of_lage_pages is an integer and is equal to TSM / 16 MB.

$ vmo p o lgpg_regions=num_of_lage_pages o lgpg_size=16 MB

2. Configure user permissions.

According to the security policy, all users (except the root user) must have the CAP_BYPASS_RAC_VMM permission to use Large Page. The permission can be configured by using the chuser command.

$ chuser capabilities=CAP_NUMA_ATTACH,CAP_BYPASS_RAC_VMM,CAP_PROPAGATE <user id>

3. Configure the following Tibero initialization parameters in the environment configuration file (.tip).

The following example configures the initialization parameters.

$ cat tibero.tip
......Omitted......
TOTAL_SHM_SIZE=1G 
USE_HUGE_PAGE=Y

4. Restart the Tibero server.

$ tbdown

Tibero instance terminated (NORMAL mode).

$ tbboot
Listener port = 8629
Change core dump dir to /home/tibero/tibero7/instance/tibero. 

Tibero 7
TmaxData Corporation Copyright (c) 2008-. All rights reserved. 
Tibero instance started up (NORMAL mode).

Disabling HugePage

To disable Large Page, restore the values that were modified to enable Large Page to their original values.

Solaris

The functionality of HugePage can be applied by using the ISM (Intimate Shared Memory) function through Large Page.

The advantages of using ISM are as follows:

• ISM shared memory is automatically locked by the OS when it is created. This ensures that the memory segment is not swapped to disk. It also allows the kernel to use a fast locking mechanism for I/O of the shared memory segment.

• The memory structure used to convert the kernel's virtual memory address into a physical memory address is shared between multiple processes. This sharing reduces CPU consumption and conserves kernel memory space.

• Large Page supported in the system's MMU (Memory Management Unit) is automatically allocated in the ISM memory segment. This allocation improves system performance by conserving memory space for managing pages and by simplifying virtual memory address conversion.

Enabling HugePage

To enable HugePage, configure the server initialization parameter as follows.

The parameter is only valid in Solaris and the default value is Y.

USE_ISM=Y

If this function is turned on, shmget is used to create shared memory and the SHM_SHARED_MMU option is added when attaching with shmat. This function is only applied in the server process and the listener process. The client module that attaches the server's TSM does not use this option. However, if the parameter is not set to Y, change the value to Y and restart the Tibero server.

Disabling HugePage

Set the server initialization parameter _USE_ISM to N and then restart the Tibero server.

_USE_ISM=N

Recommended Settings

Tibero does not restrict the size of HugePage, which can be set within the range allowed by different operating systems.

The recommended settings are as follows:

• For small memory (less than 64GB): 4MB

• For large memory (more than 64GB): 16MB

AIX

The method for mounting a CD-ROM in AIX is as follows:

1. Log in as the root user.

2. Create a mount point to which a CD-ROM drive will be mounted.

# mkdir /cdrom

1. Run the mount command.

mount <option><device_name><mount_point>

Linux

The method for mounting a CD-ROM on Linux is as follows:

1. Check whether automounting is working.

$ ps -aux | grep automount

1. Switch to a root user using the root account password.

$ su -
Password : *****

1. Run the mount command

# mount <option><device_name><mount_point>

1. Run the unmount command.

# umount <mount_point>

Solaris

The method for mounting a CD-ROM in Solaris is as follows:

1. Check the volume manager to verify whether a CD is recognized automatically.

$ ps -e | grep vold

1. If there is an existing CD, eject the CD.

$ eject

3. Insert a CD and mount it with the following commands.

# mkdir <mount_point>
# mount <options> <device_name><mount_point>

User Configuration Files

TB_HOME

Issue

TBR-2048 : Data source was not found.

If the property TB_HOME is not set or is set improperly, or if the user configuration file is not applied after Tibero is installed, this message may be displayed when trying to connect to the tbSQL utility.

Solution

Check the TB_HOME property in the user configuration file, change the value if necessary, and then apply the file to the system.

LD_LIBRARY_PATH

Issue

tbsql: error while loading shared libraries: libtbcli.so: 
cannot open shared object file: No such file or directory

If the property LD_LIBRARY_PATH is not set or is set improperly, or if the user configuration file is not applied after Tibero is installed, this message may be displayed when trying to connect to the tbSQL utility.

Solution

Check the LD_LIBRARY_PATH property in the user configuration file, change the value if necessary, and then apply the file to the system.

TB_SID

Issue

ERROR: environment variable $TB_SID is not set

tbdown: environment variable TB_HOME or TB_SID is not set.

Tip file open failure.: No such file or directory

tbdown failed. proc info file is deleted.

If TB_SID is not set, this message may be displayed when the tbboot or tbdown command is executed.

Solution

Check the TB_SID property in the user configuration file, change the value if necessary, and then apply the file to the system.

TAC Installation

The following describes how to resolve issues that can occur after Tibero is installed in a TAC environment.

Adding a Node

Accessing a tip File

Issue

tip file does not exist / reading tip file failed / malformed tip file

If a configuration file (.tip) cannot be found when a node is added, this message may be displayed.

Solution

Check whether the Tibero instance's $TB_SID.tip exists in the corresponding path by referring to TB_HOME and TB_SID.

Accessing TAC

Issue

extracting information from the existing cluster failed

To add a node, an existing node's setting values are necessary. If trying to access an existing node fails, this message may be displayed.

Solution

• Check that TB_HOME and TB_SID are set properly.

• Check that the existing node can be accessed through tbsql.

• If an existing node does not work, start the node with the tbcm, tbboot commands then try to add the node again.

Transferring an scp File

Issue

Remote file transfer failed! / Remote configuration export failed!

If transferring Tibero installation files to a new node fails, this message may be displayed.

Solution

• Check whether the current node can access the node to be added using ping or ssh.

• If the current node cannot access the node to be added, check the settings for account synchronization, networking, and any firewalls.

TPR-Related Error When Using a TAC Raw Device

Issue

The SYSSUB tablespace, which contains TPR-related tables, was added starting from Tibero 5.0 r61295. During the system schema configuration portion of installing and patching, the syssub001.dtf file is automatically created in the directory $TB_ HOME/database/$TB_SID. However, if the property DB_CREATE_FILE_DEST is set to a specific location, the file is created at that location.

However, in an environment using a TAC raw device, the file is created on the local disk of the node processing the system schema. This results in an error that prevents other nodes from accessing the SYSSUB tablespace. The same error occurs in a shared disk environment when the property, DB_CREATE_FILE_DEST, is not set correctly.

Solution

Recreate the SYSSUB tablespace as follows:

1. Delete the existing tablespace.

SQL> DROP TABLESPACE SYSSUB INCLUDING CONTENTS AND DATAFILES;

1. Create a new tablespace.

SQL> CREATE TABLESPACE SYSSUB DATAFILE '<SYSSUB 위치>/syssub001.dtf' ...;

1. Execute the following script.

$TB_HOME/scripts/tpr.sql
$TB_HOME/scripts/pkg/_pkg_dbms_tpr.tbw

Miscellaneous

Port Number

Issue

Listener port = 8629
bind() failed: Address already in use.
Error: Timedout while trying to open port 8629 
Check if there are any Tibero instances running. 
Tibero instance startup failed!

This error occurs when the specified port number is already in use when executing the tbboot command.

Solution Check whether the instance is already in use or modify the port number set in the $TB_SID.tip file in $TB_HOME/client/config.

dbtimezone

Recommendation

The dbtimezone is the reference timezone value used for storing the timestamp with local timezone data to database. Since this value is internally processed in the database, it is not recommended to modify it, because this requires unnecessary operations leading to performance degradation.

The following are the options that can be used with system.sh(vbs).

$ system.sh -h
Usage: system.sh [option] [arg]
-h : dispaly usage
-p1 password : sys password
-p2 password : syscat password
-a1 Y/N : create default system users & roles
-a2 Y/N : create system tables related to profile
-a3 Y/N : register dbms stat job to Job Scheduler
-a4 Y/N : create TPR tables
pkgonly : create psm built-in packages only
-sod Y/N : separation of duties
error : print error whenever sql encounters error

The following describes each option.

Supported Character Sets

The following example checks the available character sets for installing the Tibero binary.

If data is updated frequently in a system, a large volume of I/O happens in the undo tablespace. In such systems, data needs to be distributed by creating the undo tablespace with multiple data files.

The undo tablespace is automatically managed, providing the following advantages.

• Helps design and manage efficiently with rollback capability and segments

• Prevents undo data from being overwritten

• Provides scalability

• Supports Flashback queries

Considerations

Consider the following when calculating the size of an undo tablespace.

Data file size, UNDO_RETENTION, and undo segments

Individual spaces in an undo segment are automatically managed, therefore you only need to set the minimum/maximum number of undo segments and the maximum size of a data file.

• Set the minimum number of undo segments (USGMT_ONLINE_MIN) to 10 (default value). However, if there are too many concurrent transactions, overhead occurs to create additional undo segments. To decrease the overhead, set USGMT_ONLINE_MIN to the number of concurrent transactions.

  If USGMT_ONLINE_MIN is greater than the number of existing undo segments, additional undo segments are created at boot time. USGMT_ONLINE_MIN can be set by node in TAC. For example, USGMT_ONLINE_MIN can be set to 30 and 40 for NODE1 and NODE2 respectively.

• Set the maximum number of undo segments (USGMT_ONLINE_MAX).

  Tibero automatically creates additional undo segments to prevent multiple transactions from sharing a single undo segment. However, the number of undo segments cannot exceed USGMT_ONLINE_MAX.

  In TAC, a total number of Undo segments in all nodes cannot exceed USGMT_ONLINE_MAX. For example, if USGMT_ONLINE_MAX is set to 300 and the numbers of undo segments are 100 and 200 in NODE1 and NODE2 respectively, additional undo segments cannot be created in both nodes. Therefore, it is recommended to set USGMT_ONLINE_MAX to the number of maximum concurrent transactions.

Calculating the Size

The minimum size, expected size (based on the expected concurrent transactions count), and actual size of an undo tablespace can be calculated as follows. The expected size is used when a database is installed for the first time, and then the size can be checked based on the number of Undo blocks per second under the heaviest TPR load.

The following are the formulas for calculating the size of an undo tablespace.

Minimum size

Expected size

Actual size

The following describes each item in the formulas.

The shared memory in Tibero is divided into fixed memory and shared pool memory as follows:

• Fixed Memory Its size is fixed based on a parameter value set at database startup.

  – Buffer cache (database buffer)

  – Log buffer

  – Others: space for global variables and managing working threads and session information.

• Shared Pool Memory It is dynamically allocated during database operation.

  – DD cache

  – PP cache (library cache)

  – Others

Considerations

Consider the following when calculating the size of the shared memory.

• The size of the shared memory cannot be increased dynamically while the database is running.

• The size of the shared memory must be determined properly based on the usage patterns of the buffer cache and shared pool.

- Buffer Cache

• Determine the size according to the Buffer Cache Hit rate in TPR after executing major workloads.

• If the hit rate is lower than 90%, increase the buffer cache.

- Shared Pool Memory

• Determine the size according to the SHARED POOL MEMORY item in v$sga after executing major workloads.

• If the shared pool usage is too high (insufficient free area), increase the shared pool memory.

• At least 1 MB of the shared pool memory is required for each session.

Calculating the Size

The size of each area in the shared memory can be checked with v$sga.

The size of the shared memory can be set and queried with the following initialization parameters.

Shared Memory

The following example queries the size with the TOTAL_SHM_SIZE parameter.

Buffer Cache

The following example queries the size with the DB_CACHE_SIZE parameter.

• Single mode : ⅔ of TOTAL_SHM_SIZE

• TAC mode : ½ of TOTAL_SHM_SIZE

Log Buffer

The following example queries the size with the LOG_BUFFER parameter. (the default value of LOG_BUFFER is 10 MB.)

Shared Pool Memory

The size can be calculated by subtracting the fixed memory size from the total shared memory size as follows:

• Single mode The total shared pool size must be greater than _MIN_SHARED_POOL_SIZE.

• TAC mode Sufficient space for Cluster Cache Control (CCC) and Cluster Wait-lock Service (CWS) is required. About 25% of a total buffer cache size is used by shared pools.

  The shared pool space excluding space for CCC and CWS must be greater than

  _MIN_SHARED_POOL_SIZE.

The free shared pool size for PP and DD caches must be greater than CCC and CWS spaces.