(OO) Support Tip: Many http connections towards OO can lead up to feeling of 'stuck' application
Applies to OO 10.x versions.
Problematic behavior: Due to high load, the executing flows might appear as not finishing
When many requests of triggering executions are shot towards OO in a short period of time, the application might look stuck with flows running indefinetly.
When monitoring with JavaMelody ( on OO Central UI page - <OO_url>/oo/monitoring), it appears as a burst of http connections, that never goes down (http connections are never freed). During this time, the system (OO Central) can appear as jammed with many flow executions in status Running.
This high load can be handled with system tuning to have it prepared for this high load:
- workers tuning (see 10.x Performance Tuning guides)
- adding more Central nodes to the system and have them all configured behind a Load Balancer
- increasing OO Tomcat's capacity to process more HTTP requests. (central/tomcat/conf/server.xml -> SSL connector -> increasing "maxThreads" parameter from default 200 , to up to 200 per number of CPU cores the machine has)
(meaning , if a machine has 4 CPU cores, you can increase to up to 800 threads)
- decrease session timeout if you are doing a lot of requests via SOAP towards OO itself that keep the connections busy (HP Solution's "Dynamically Launch Flow" operation up to v1.6.0. was using SOAP reuqests towards OO, when this changed to REST calls)
- Note that if you tune all the other parameters to scale up the Central node, you might also need to increase the db.pool.maxPoolSize from the database.properties file
- starting OO 10.51, a system-wide timeout parameter was introduced to cancel all flows that are running for a longer period of time (established by the customer)