OHDSI Home | Forums | Wiki | Github

Atlas features failing with concurrent users

(Ambuj) #1


I have set up Atlas which is working fine when I try to perform available functions with single user. While doing the load test on it with 18 concurrent users, all the analyses are either failing or stays in starting mode.
I am attaching the screenshot for reference.
RAM : 16 GB
PostgreSQL 10
Data : Synpuf 1K

Please let me know how many concurrent users can access the application to avoid it or if there is any solution to it, I will be happy to implement it.

Thank You

  • Ambuj

(Chris Knoll) #2

The back-end job processing infrastructure allows up to 10 concurrent activities happening at once. Any request past those 10 get queued up and will be handled in order. This is why you may see ‘Starting’ vs. ‘Started’ in the job status: the jobs that are starting haven’t started because of the 10 concurrent limitation.

One thing you should realize is that resources are not infinite and when doing large scale analytics, the platform has to be conscious of those limited resources. So the default is 10 concurrent threads, but I believe that there may be a configuration option to change that value. I’m not sure if that configuration option is user-defined.


(Ambuj) #3

Hello @Chris_Knoll,

I finally figured out the way to have concurrent users. This can be achieved by modifying POM.XML file for the segment <spring.datasource.hikari.maximumPoolSize>. The default value is 5, means not more than 5 concurrent users can attempt to generate Atlas features. I changed the value to 20, rebuild webapi.war and it worked with 18 users like a charm.

Thank You