yesterday we released our new server and since then we have had consistently strong, recurring laggs.
Here is a current timings report as link:
As you can see the lag is due to a category named “Connection Handler”. We are not sure what exactly this means and what do do now. We tried help of serveral tools and have now come to the conclusion that an memory leak is causing the lag. At the low point of the laggs the RAM usage is about 3GB, at the high point of the laggs about 12GB. The time between a high and a low point ist currently about 10 seconds. The ram usage increases about 1GB per second to 12GB until it’s get back down to 3GB and the TPS is back to 20.0.
We tried the java garbage collection to tackle this leak but it does not help. That’s why we are not sure if the ram leak is causing the lag or if it’s the other way around. Furthermore the category “Connection Handler (SELF)” doen’t give us a lot of information where to look for an issue.
Tha last thing we did was a heap dump but also there we did’nt find evidence for a leak:
It would be very helpful, if one of you has an idea what the problem could be.
With hopeful regards, Tim