The Docker image is configured with the following HEALTHCHECK to determine if the Elemental server is healthy:
HEALTHCHECK CMD [ "java", \
"org.exist.start.Main", "client", \
"--no-gui", \
"--user", "guest", "--password", "guest", \
"--xpath", "system:get-version()" ]
The Docker HEALTHCHECK documentation says that this by default will be run every 30 seconds. If Docker does not receive a response within a 30 second timeout , it will retry. If there are 3 failures the container will be marked unhealthy.
The above HEALTHCHECK makes an XML-RPC call from the Java Admin Client to the Elemental server.
The issue is that there does not appear to be a timeout on the XML-RPC call, and so during times of high-load multiple XML-RPC calls can be opened by Docker HEALTHCHECK, and remain open to the server.
To prevent these XML-RPC calls staying open, we should add a configurable timeout to the XML-RPC call, and we should perhaps consider adding a dedicated 'health-check' command to the Java Admin Client that takes care of correctly setting the XML-RPC call timeout.
The Docker image is configured with the following HEALTHCHECK to determine if the Elemental server is healthy:
The Docker HEALTHCHECK documentation says that this by default will be run
every 30 seconds. If Docker does not receive a responsewithin a 30 second timeout, it will retry. If there are3 failuresthe container will be marked unhealthy.The above HEALTHCHECK makes an XML-RPC call from the Java Admin Client to the Elemental server.
The issue is that there does not appear to be a timeout on the XML-RPC call, and so during times of high-load multiple XML-RPC calls can be opened by Docker HEALTHCHECK, and remain open to the server.
To prevent these XML-RPC calls staying open, we should add a configurable timeout to the XML-RPC call, and we should perhaps consider adding a dedicated 'health-check' command to the Java Admin Client that takes care of correctly setting the XML-RPC call timeout.