Intelligent Management for IHS web servers
In addition to ODRs we can implement Intelligent Management capabilities using Apache or IBM HTTP servers.
- Automatic routing
- The web server plug-in automatically recognizes application server events (create, delete, start, or stop), application events (start, stop, update), virtual host updates, session affinity configuration changes, and other events that affect routing.
- Routing rules
- Routing rules can be used to select specific clusters or servers to route to based on the content of a request.
- Routing information and statistics
- We can retrieve routing information and per-application and per-server statistics. If the Apache mod_status module is enabled, the WebSphere Application Server plug-in with IM enabled adds WAS specific status information to http://your_host/server-status.
If multiple child processes are configured in Apache, the statistics that are presented are only applicable to the single child process that handles the current server-status request.
- Application edition routing rules
- Routing rules associated with an application are automatically enforced by the plug-in to route to the appropriate application edition.
- Edition aware routing
- The plug-in maintains continuous availability during application edition rollouts.
- Node and server maintenance mode routing
- When a node or an application server maintenance mode is set, the plug-in automatically routes the requests that are based on the mode.
- Multi-cell routing
- The plug-in routes to multiple cells.
- Routing to the same application in multiple cells is supported.
- Highly available connection to each cell
- The plug-in connects to a REST service in each cell. The REST service endpoint is available in the deployment manager and in each node agent via the XD_AGENT port. When a connection failure occurs, the plug-in automatically fails over to another REST service endpoint.
- Edition aware caching
- The Edge Side Include (ESI) cache in the plug-in is aware of the edition to which the request is targeted. The edition is used to form the key when both storing and retrieving cache entries.
- Weighted Least Outstanding Requests (WLOR) load balancing
- This load balancing algorithm quickly responds to application server slow down or hang.
- Automatically adapt to dynamically changing server weights
- The plug-in recognizes dynamically changing server weights and automatically uses these weights in load balancing.
- IM health policies
- The plug-in supports the excessive request timeout and excessive response time health policies by alerting the health controller when a configured threshold is exceeded. Health policies that do not involve the on demand router (ODR), such as excessive memory usage, and custom health conditions, are available in the web server plug-in.
- Per-request conditional tracing
- Use the plug-in to enable tracing of specific requests.
All of these capabilities are also available via the ODR. However, the ODR is stabilized.
The following capabilities are not available in the plug-in. Use the ODR for the following scenarios:
- CPU overload protection.
- Memory overload protection.
- Request prioritization.
- Highly available deployment manager.
- Classification based on the user identity in a LPTA token.
- Lazy start for dynamic clusters.
- Workload and storm-drain health policies.
When IM is enabled in the WebSphere plug-in, routing information is not defined in the plugin-cfg.xml file. Instead, the plug-in connects to a REST service to dynamically gather routing information for one or more WebSphere cells. The following diagram is an overview of how IM works for Apache and IHS.
The IM daemon process is a single highly available process, which starts automatically when IM is enabled. Each child worker process connects to the IM daemon process to get routing information. The IM daemon connects to an IM REST service via XD_AGENT port of the deployment manager and each node agent. The IM daemon process connects to only one REST service at a time per cell. If the process that the IM daemon is connected to stops or fails, the IM daemon connects to another process hosting the IM REST service. A child worker process does not connect directly to the IM REST service; only the IM daemon process connects to the IM REST service. When IM is enabled, the plugin-cfg.xml file is populated with endpoint information for all XD_AGENT ports: for the deployment manager and for each node agent. The plug-in requires this data to discover routing information in a highly available manner. After the IM daemon connects to an IM REST service and retrieves the initial routing information for the cell, the daemon immediately issues another request, which blocks until a change that affects routing occurs. In this way, the plug-in is immediately notified of configuration or state changes, which are important for routing.
(ZOS) Restriction
IM requires that all configurations support User Datagram Protocol (UDP) traffic. However, Sysplex Distributor does not support UDP traffic in all configurations. When we use Sysplex Distributor for a mobile deployment manager in a configuration that includes IM, the TCP/IP stack of the deployment manager must own Sysplex Distributor. This setup allows UDP traffic to flow and IM to work properly. We can change the ownership of Sysplex Distributor by issuing a VARY TCPIP,,SYSPLEX,DEACTIVATE command.
Subtopics
- Configure Intelligent Management for web servers
- Configure trace on the WebSphere plug-in
- Configure Intelligent Management properties for web servers
- Intelligent Management: administrative tasks for web servers
- pluginIntelligentManagement.py script
- Set up routing rules for WAS traditional
- Troubleshoot Intelligent Management for web servers
Related: