IBM Tivoli Composite Application Manager for Application Diagnostics, Version 7.1.0.1
Tuning recommendations for large-scale environments
The following steps are tuning recommendations for when the Managing Server is used for supporting large scale environments or high volumes of request traffic. For example, when there are 250 data collectors running at a total of 500 requests per second:
If you are planning to deploy more than 100 data collectors, or if you intend to use high monitoring levels (L2/ L3) frequently, for example in a development or testing environment, it is recommended that you use a split Managing Server installation for better scalability, performance, and reliability. See Split Managing Server installation on Windows for split Managing Server installation.
- Increase the maximum heap size settings of the Kernel, Publish Servers and Global Publish Server. The Global Publish Server change is only necessary if you are using the correlation feature in ITCAM for Application Diagnostics. In MS_home/bin/setenv.sh, modify the following properties:
- HEAP_MAX_SIZE_PS=1024
- HEAP_MAX_SIZE_KERNEL=512
- HEAP_MAX_SIZE_SAM=1024
- Increase the maximum heap size used by the ITCAM for Application Diagnostics Managing Server. Do the following:
- Log into the IBM WAS administrative console.
- Navigate as follows:
- Click Server > Application Servers and select the server_name.
- In the Configuration tab, navigate to Server Infrastructure > Java and Process Management > Process Definition > Additional Properties: Java Virtual Machine.
- Modify the maximum heap size of the server to be 1024.
- Open a Korn shell window. Go to Start > Programs > Windows Services for UNIX > Korn Shell
Do not use an already existing Korn shell. Instead open a new Korn shell window.
- Change to the C directory, then change to the MS_home/bin directory:
cd ..cd ..cd MS_home/binFor example,cd /IBM/itcam/MS/bin"When changing to the MS_home/bin directory, if the file path that you specify contains spaces, place double-quotes around the path. Use forward instead of backward slashes in the path.
- If you have not already created an additional instance of the Kernel, perform the instructions at Add an additional kernel.
- Modify heartbeat settings. In MS_home/etc/kl1.properties, modify the following property:
contract.duration=60- Start one more instance of the Kernel from the MS_home/bin directory:
./amctl.sh wd2 start- If you have not already created two additional instances of the Publish Server, perform the instructions at Add additional Publish Servers.
- Start two more instances of the Publish Server from the MS_home/bin directory:
./amctl.sh ps3 start./amctl.sh ps4 start- If you have not already created two additional instances of the Archive Agent, perform the instructions at Add additional Archive Agents.
- Start two more instances of the Archive Agent from the MS_home/bin directory.
./amctl.sh aa3 start./amctl.sh aa4 start- Increase MAXAPPLS and MAXAGENTS for the database. Perform the following procedure:
- Switch to the database instance user. For example:
su - db2inst1- Choose one of the following options:
- If you do not have a partitioned database, run the following commands:
update dbm cfg using MAXAGENTS 500 db2 update db cfg for octigate using maxappls 300 db2stop db2start- If you have a partitioned database, the DBA needs to run the following commands. For example, if you have a 3 node partitioned database:
db2 terminate export db_node0 db2 update db cfg for octigate using MAXAPPLS 300 db2 update db cfg for octigate using MAXAGENTS 500 db2 terminate export db_node1 db2 update db cfg for octigate using MAXAPPLS 300 db2 update db cfg for octigate using MAXAGENTS 500 db2 terminate export db_node2 db2 update db cfg for octigate using MAXAPPLS 300 db2 update db cfg for octigate using MAXAGENTS 500 db2 terminate
In addition, if you have large amounts of data in the tables affected by the datatrim.sh script, the LOGFILSIZ and LOGSECOND parameters need to be increased across all database partition nodes.
- Set the Remote Method Invocation (RMI) timeout in the data collector. The data collector sockets are initially set for infinite time out. This could make the Kernel's defensive calls hang and could deplete the Kernel if there is a problem with a data collector. Set the following property for all data collectors in DC_home/runtime/appserver_version.node_name.server_name/appserver_version.node_name.server_name.datacollector.properties or DC_home/runtime/app_server_version.node_name.server_name/datacollector.properties:
am.rmisocket.timeout=60000
Parent topic:
Customization for the Managing Server on Windows