Hello, Yesterday I installed from scratch jmeter on some virtual machines in a cloud environment, and had an odd problem with starting jmeter-server. As this might happen to other testers, I'm sharing the workaround that I found
My environment I have 10 JMeter remotes, they are virtual machines (CentOS) that are switched off when no performance test is running,, and can be started with OpenStack. My problem I wanted jmeter-server to start automatically, so I created a .service file for bin/jmeter-service I noticed that the Java process would start on 6 or 7 vm, not on the remaining 3 or 4 It turns out that, with systemd, unless a dependency is explicitly given, the order of startup is not guaranteed. So sometimes, jmeter-server started before the network initialization was finished, and rmi could not bind to any ip address The workaround Simply adding an "After" statement to the .service file (see below), to make sure the network is fully initialized. Important in cloud environments, where the ip address is assigned by a dhcp server. The two other relevant settings in the file are "ExecStart" (the full path to jmeter-server) and "WorkingDirectory" (basically, where jmeter-server.log will be saved) [Unit] Description=Start Jmeter-service After=network-online.target [Service] Type=simple WorkingDirectory=/root/lasttest ExecStart="/opt/lasttest/jmeter/bin/jmeter-server" [Install] WantedBy=multi-user.target ---------------- Your mileage may vary. There might be better/cleaner workarounds then mine Thanks to the JMeter for Jmeter and happy testing, Ivan
