I'm performing some load tests on my brand new server Apache/PHP/Mysql (Bitnami LAMP stack ). I'm using AWS with 1 EC2 instance behind a loadBalancer. At the moment I'm using loader.io for the tests. I'm testing a very simple API that perform this query:
select *,sleep(0.5) from debug limit 1
I added the sleep(0.5) because I wanted to see how the server behave with multiple concurrent connections and I found a bottleneck: if I run "SHOW PROCESSLIST" I can see only 5 process even if I have 10 concurrent user. The load test show that the connections are queued because the response time is growing during the test from 500 milliseconds to several seconds (depending on the duration of the test and the number of concurrent users).
I checked
select @@max_connections
and it's 151 (the default). max_user_connections
is 0. What other parameter should I check to increase the number of concurrent connection on my DB?
If I run the test with 5 concurrent users, each one get a response in 500 milliseconds. If I add more concurrent users than the response time slow down.
If I run the load test on an API that does not access the DB there are no issues even with 400 concurrent users.
Monitoring with HTOP I see:
Tasks: 34, 245 thr; 2 running
Could be here the issue?
Thanks a lot