The reasoning and actual scripts for my approach to manage IBM WebSphere Application server installations, in particular for systemd to start and stop WebSphere.
WebSphere Scripts for managing WebSphere Application Server
This repository contains my scripts to manage IBM WebSphere Application server installations. I set it up to share my start and stop scripts that make sure the infrastructure is started up and shut down correctly on system restart but does not fail when sombody is manually starting or stopping a WebSpere Server or Node Agent.
—› Link to Git Repository: https://github.com/becketalservices/WebSphere_scripts
Why do I not use systemd scripts directly to start or stop WebSphere
In the early days where SysVinit was the standard init-system for Linux, I just used the standard WASService command to creates servcies for all required processes (Dmgr, NodeAgent, Server, Proxy) on my system, set up the right boot order with the priority scheme and everything worked as intended. Today, when systemd replaced the good old SysVinit, it is not that simple anymore as systed tries to keep track of the processes started by the systemd. There are a couple of guides out there on how ot start WebSphere using systemd unit file but none of them worked 100% for me.
Example 1: Unit file using forking
[Unit] Description=IBM WebSphere Application Server [Service] Type=forking ExecStart=/opt/IBM/WebSphere/AppServer/profiles/Dummy01/bin/startServer.sh server01 ExecStop=/opt/IBM/WebSphere/AppServer/profiles/Dummy01/bin/stopServer.sh server01 User=wasuser PIDFile=/opt/IBM/WebSphere/AppServer/profiles/Dummy01/logs/server01/server01.pid SuccessExitStatus=143 0 [Install] WantedBy=default.target
This unit script works quite well when the service is started and stopped usind systemctl command. It detects the running WebSpehre process correctly and shuts it down when requested. (The credentials to stop the WebSphere process are in the soap.client.props so I do not need to add them here in clear text.) This unit script fails, when a different process is starting or stopping the WebSphere process. It looses track of the process and shows it either as not started or failed. Further actions will not be possible as a not started or failed process will never run the stop operation. A start will not work when the WebSphere Server is already running as the start script will exit with a error other than 0 or 143. (usually 255 which also might indicated some other error.)
Example 2: Unit file using oneshot
[Unit] Description=IBM WebSphere Application Server [Service] Type=oneshot ExecStart=/opt/IBM/WebSphere/AppServer/profiles/Dummy01/bin/startServer.sh server01 ExecStop=/opt/IBM/WebSphere/AppServer/profiles/Dummy01/bin/stopServer.sh server01 User=wasuser [Install] WantedBy=default.target
This unit script works better as long as nobody is fiddling arround with the service to start or stop the WebSphere Service. When your WebSphere Server has a different state than systemd thinks, you are in trouble. After system start, the script is started and therefore WebSphere is stared. Now somebody stops WebSphere using the stopServer.sh script. According to sytemd the process is still running. Unsing "systemctl stop websphere" will fail as the stopServer.sh script run by systemd will fail as no WebSphere Server is running anymore. Especially when it comes to the next system shutdown. When somebody stared WebSphere again but systemd still thinks, the service is not running, WebSpehre is not stopped by systemd. The second disadvantage of this direct usage of this scripts is that you have to maintain the boot order inside the unit files. This is not funny to set up and to automate this process is also not that simple.
The solution I choose
I decided to choose a boot system using multiple scritps that are documented in this git repository. The main advantage of these scrips is that they can detect what is installed on the system and can check the status of the WebSpehre process to act accordingly.
- startInfrastructure.sh / stopInfrastructure.sh
This script is the main script that coordinats the work to start / stop the infrastructure.
- startAllNodes.sh / stopAllNodes.sh
This script starts / stopps all nodes on the system as they are registerd in the profileRegistry.xml
- startAllServer.sh / stpoAllServers.sh
This script loops throuh all nodes on the system and starts / stopps all registered servers within this node.
- unit file
The unit file calls the startInfrastructure.sh on startup and the stopInfrastructure.sh on shutdown.
To deploy this scripts execute this steps:
- download all scripts in the tools folder to /opt/IBM/tools (or choose your own. Update in unit file if necessary.)
- update the profile path inside all 6 scripts if necessary.
- make shure this scripts are owned and executable by your websphere user.
- download the websphere.service file from the unit folder into /opt/systemd/system
- adjust the websphere.service file (e.g. adjust the name of your webphere user or change the path to the start scripts)
- enable the service (systemctl enable websphere.service)
- test the scripts by starting / stoping the websphere service (systemctl start webphere)
- reboot your system to test again