I am making a dashboard on a server that makes postgreSQL requests to a second server. Currently, the clients connect to the dashboard, make the requests to the grafana server, the grafana server makes the requests to the database and provides them to the client.
Since the requests are very very heavy, and there are quite a few clients, I would like the client itself to make the requests DIRECTLY to the database so as not to overload the grafana server.
When you say “the requests are very heavy”, surely this means the requests
place a significant load on the Postgres server - but not on the client making
the request (ie: the Grafan server)?
So, moving the location the requests are made from is not going to make a
difference, I think?
Exactly, you are right, doing that does not decrease the load on the postgreSQL server. But it would decrease the load on the grafana server, wouldn’t it?
It would also help us to solve a problem related to some users not having a good connection to the grafana server (yes to the PostgreSQL server).
How do you measure server overloading? I guess load metric can be high in this case, because there can be processes, which are doing nothing and waiting for results from DB. It can be scary for some admins to have high load (e.g. 400), when they think that “normal” load is <1. But CPU utilization (that’s not CPU load metric, that’s different metric) will be low, so I wouldn’t worry about high load in this case.