My client has an environment with SDL Web 8.1.1, with the delivery stack on Linux.
We are experiencing problems with the number of connections we make from our web application to the content microservice.
When the server is under heavy load, we make more than 50,000 requests to the content service per minute. On Linux there can be no more than 30,000 (or so) open connections. Unfortunately, Linux keeps connections open for (by default) 5 minutes.
Of course we are looking into ways to change the Linux configuration, but the Linux operators tell me that the root cause is in the application: it seems the CIL (Content Interaction Library) opens and closes a connection on every request to the content service. Instead, it should use a pool of connections.
My question is: is there a way to configure connection pooling in the CIL?
server.tomcat.max-threads
which specifies the maximum number of request processing threads that the connector can handle. The default is 200. This can be modified by creating (or modifying) a file namedapplication.properties
in the config folder, and setting a value, such as:server.tomcat.max-threads=400