 |
» |
|
|
|
For operation of the distributed lock manager (DLM) software
with Oracle Parallel Server 7.3.x, you set DLM_ENABLED to YES and GMS_ENABLED
to NO, then define two sets of parameters — cluster interface
parameters and DLM lock database parameters. The DLM lock database
parameters relate to the number of Oracle resources and the number
of Oracle processes in the OPS configuration; these values should
be chosen in consultation with the Oracle DBA. Refer to your Oracle
documentation for additional information. Cluster Interface Specific DLM Parameters | |
Cluster-specific DLM parameters are stored along with other
cluster information in the binary cluster configuration file, which
is located on all nodes in the cluster. These parameters can be
entered by using SAM or by editing the cluster configuration template
file created by issuing the cmquerycl command, as described in the chapter "Chapter 5 “Building an OPS Cluster Configuration”." Appropriate values must be identified for the following DLM
cluster interface parameters: - GMS Enabled
Set this parameter to NO when using OPS version 7.3.x.
Default: NO. In the ASCII cluster configuration file, this parameter is
known as GMS_ENABLED. - DLM Enabled
When set to YES, the DLM starts in the cluster when the
cluster starts or reboots. Set to NO if you wish not to start up
OPS/DLM when the cluster is started. Default: NO. In the ASCII cluster configuration file, this parameter is
known as DLM_ENABLED. - DLM Connect Timeout
The upper bound on time available for the DLM to
initialize its shared memory on startup. Default: 30 seconds. In the ASCII cluster configuration file, this parameter is
known as DLM_CONNECT_TIMEOUT. - DLM Halt Timeout
The upper bound on time available to execute the OPS
halt scripts. Default: 180 seconds. In the ASCII cluster configuration file, this parameter is
known as DLM_HALT_TIMEOUT. - Ping Interval
Interval at which the cluster manager sends messages to
the DLM to check the status of its health. Default: 20 seconds. In the ASCII cluster configuration file, this parameter is
known as DLM_PING_INTERVAL. - Ping Timeout
Time after which the cluster manager assumes that
the DLM is not active. Default: 60 seconds. In the ASCII cluster configuration file, this parameter is
known as DLM_PING_TIMEOUT. - Reconfiguration Timeout
The number of seconds that the Cluster Manager should
wait for the DLM to start or reconfigure before assuming the failure
of the DLM. Default: 300 seconds. In the ASCII cluster configuration file, this parameter is
known as DLM_RECONFIG_TIMEOUT. - Communication Fail Timeout
Time after which the cluster manager assumes that
no reconfiguration will take place and takes action on a DLM communication
failure. Default: 270 seconds. In the ASCII cluster configuration file, this parameter is
known as DLM_COMMFAIL_TIMEOUT. - DLM Halt Timeout
The upper bound on time available to execute the OPS
halt scripts. Default: 240 seconds. In the ASCII cluster configuration file, this parameter is
known as DLM_HALT_TIMEOUT.
Distributed lock manager planning can be done using the SAM
high availability options, which let you display defaults or lists
of acceptable values for the parameters listed above. Enter the
appropriate values shown in the SAM display onto your DLM configuration
worksheet. If you are unsure of what value to use for a parameter,
start with the default. DLM Internal Parameters | |
DLM internal parameters are stored in the binary DLM configuration file,
which is located on all nodes in the cluster. These parameters can
be entered by using SAM or by editing the DLM configuration template
file created by issuing the dlmquery command, as described in the chapter "Chapter 5 “Building an OPS Cluster Configuration”." The defaults are sufficient for the
Oracle demo database, but you should adjust these parameters according
to the size of your development or production system. | | | |  | NOTE: Note that values for DLM resources can change between
OPS releases. Consult your Oracle documentation for the version
of OPS you are using to obtain the correct values. | | | | |
Appropriate values must be identified for the following: - Cluster Name
This is the same as the cluster name you use in
cluster manager configuration. In the ASCII DLM configuration file, this parameter is known
as CLUSTER_NAME. - Processes
The maximum number of DLM processes that may run on
the cluster. This is roughly equivalent to the number of Oracle
processes that run concurrently on both nodes. The parameter has
to be changed if the sum of the Oracle PROCESSES parameters in the
two instances (one for each node) exceeds 2200. The DLM default,
which is 2400, will have to be increased to leave some margin for
miscellaneous additional processes. In the ASCII DLM configuration file, this parameter is known
as MAXPROCESSES. - Resources
This is the total number of distributed locks for
which memory must be allocated in an OPS on HP-UX system. The default
is 6000. A list of the actual values of the PROCESSES, MAXRESOURCES,
and MAXLOCKS parameters used by the DLM is found in /var/opt/dlm/logs/dlmstart.log. Consult
the Oracle Parallel Server for HP 9000 Servers Addendum for
the formula for computing MAXRESOURCES and MAXLOCKS. In the ASCII DLM configuration file, this parameter is known
as MAXRESOURCES. - Locks
The size of the DLM lock database. This is a value based
on an Oracle configuration parameter known as DLM_RSRCS. The default
is 2*(Resources) or 12000. A list of the actual values of the PROCESSES, MAXRESOURCES,
and MAXLOCKS parameters used by the DLM is found in /var/opt/dlm/logs/dlmstart.log. Consult
the Oracle Parallel Server for HP 9000 Servers Addendum for
the formula for computing MAXRESOURCES and MAXLOCKS. In the ASCII DLM
configuration file, this parameter is known as MAXLOCKS. - Deadlock Detection Interval
Interval at which the DLM sends messages to determine
whether deadlock has occurred between nodes. Default: 3 seconds. In the ASCII DLM configuration file, this parameter is known
as DEADLOCK_DETECTION_INTERVAL. - DLM Monitor Interval
The interval at which the DLM monitors client processes.
On discovering a dead client process, the DLM carries out lock recovery.
The default interval is 3 seconds. In the ASCII DLM configuration file, this parameter is known
as the PROCESS_MONITORING_INTERVAL. - Subnet Address
The subnet address of the LAN used for inter-DLM messages
passed between OPS instances. By default, this is the same as the
subnet used for the cluster manager heartbeat messages. If you wish
to separate heartbeat traffic from DLM message traffic, select a different
monitored subnet address for DLM than you chose for heartbeats in
the basic cluster configuration. This value is used in configuring the DLM with SAM. - Node Name(s)
These are the same as the node names you use in cluster
manager configuration. In the ASCII DLM configuration file, this parameter is known
as NODE_NAME. Use a separate entry for each node paired with an
IPADDR. - DLM Node IP Addresses
The IP addresses of each node's interface
to the DLM subnet. Must be consistent with the subnet address. By
default, this is the same as the IP address used for cluster heartbeat
on this node, but if you wish to separate DLM message traffic from
cluster heartbeat traffic, you can specify an IP address on a different
monitored subnet than the one you chose for heartbeats in the basic
cluster configuration. This value is used only when configuring
the DLM by editing DLM configuration files. In the ASCII DLM configuration file, this parameter is known
as IPADDR. Use a separate entry for each node immediately following
the NODE_NAME.
Distributed lock manager planning can be done using the SAM
high availability options, which let you display defaults or lists
of acceptable values for the parameters listed above. Enter the
appropriate values shown in the SAM display onto your DLM configuration
worksheet. If you are unsure of what value to use for a parameter,
start with the default. Setting HP-UX Shared Memory Allocation | |
Based on the needs of your system, you may need to adjust
the total amount of memory to provide enough lockable memory for
the DLM, as well as for Oracle programs and the cluster daemon processes.
The default values of the PROCESSES, MAXRESOURCES and MAXLOCKS parameters
(2400, 6000 and 12000 respectively) require the DLM to configure
a shared memory segment, locked in core, of about 5 Mbytes on each
node. For each increment of 5000 in the PROCESSES parameter, an extra
1.6 Mbyte of memory will be required for the DLM shared memory segment.
For each increment by 5000 in the MAXRESOURCES parameter (assuming
MAXLOCKS is increased by 10000 to be twice the increased RESOURCES
value) the corresponding increase in the DLM shared memory segment
is 3.5 Mbytes. A list of the actual values of the PROCESSES, MAXRESOURCES,
and MAXLOCKS parameters used by the DLM is found in /var/opt/dlm/logs/dlmstart.log.
Consult the Oracle Parallel Server for HP 9000 Servers
Addendum for the formula for computing MAXRESOURCES
and MAXLOCKS. DLM Configuration Worksheet (OPS 7.3.x) | |
Fill out this worksheet in cooperation with your HP-UX system administrator
and Oracle database administrator. Figure 4-6 “DLM Configuration Worksheet” is
a sample worksheet filled out. Refer to the appendix, "Appendix E “Blank Planning Worksheets ”" for samples of blank worksheets.
Make as many copies as you need. Fill out the worksheet and keep
it for future reference. |
Printable version
