Hi, I am currently doing a capacity test on the FlexNet Publisher 2024R1 server On-Premise Linux machine.

I submitted one checkout job every 2 seconds and each job I intentionally put it to sleep for 2 hours (PS: On the client side of code due to business needs, the socket connection to the server is made long-lived. It means that the socket connection are kept open during the entire life cycle of a job. Here, the connection is kept open for 2 hours). The result is that, after submitting ~1000 jobs I started to receive these error messages from the server:

Failed to obtain license for feature ***, version ***, status: -15, reason: Cannot connect to server

Failed to obtain license for feature ***, version ***, status: -16, reason: Cannot read from server

Failed to obtain license for feature ***, version ***, status: -5, reason: No such feature exists (Not expected since checkout successfully below ~1000 jobs)

Failed to obtain license for feature ***, version ***, status: -139, reason: a timeout has occured

Based on the FNP documentation(2024R1, Page 121), it seems that both the lmgrd and the vendor daemon are overwhelmed. However, the documentation also said, the max default connection limit of lmgrd is 1000, but the max connection limit of vendor daemon is 10,000.

I've checked `ulimit -n` is 8192.

So I have 2 questions here:

1. If the server keeps as default configuration, is the capacity limit ~1000 expected due to the long-lived socket connections?

2. Is there a way to adjust the server's configuration in order to reach the 10,000 max concurrent connections as stated in the Documentation (or even as high as 120k showed in recent Webinar slides below)?

Thank you!