Indico server issue with high number of concurrent users

Hello,

I’m experiencing an issue with high concurrency users on our Indico server and need some help configuring the application and PostgreSQL database servers. We have almost 4,000 users registered for one event, and we invited 18,000 to this event. Once we open the management page of this events, such as logs page or the list of registered users for checkin, the system crashes and only comes back after we restart the uWSGI service. Does anyone have advice for configuring PostgreSQL and uWSGI to handle high concurrency users?

Thank you!

How much memory does your system have? Do you see oom-killer messages in dmesg?

FWIW, the list of all registered users will be slow with that many users. We plan to make it nicer for such huge lists (e.g. using pagination and/or virtualized rendering) in the future.

The logs page should not be slow, since it only loads a few entries at at ime.


Did you do any tuning on the Postgres config? The defaults are VERY conservative. If you have a large instance, you should have run it on another machine, and generally have SSD storage, a decent amount of RAM and a Postgres config that is tuned to make good use of that RAM.

1 Like

Yes, we tuned the PostgreSQL config, and the issue disappeared. What are the values of the PostgreSQL parameters after tuning?

max_connections = 256
shared_buffers = 4GB
effective_cache_size = 6GB
work_mem = 8MB
maintenance_work_mem = 512MB
min_wal_size = 2GB
max_wal_size = 4GB
checkpoint_completion_target = 0.9
wal_buffers = 16MB