Feedback and Suggestions (Path of Exile 1) - Garbage Collected server side language? - Forum

I'm a software dev and these lag spikes do seem like a Garbage Collector is kicking in. Please GGG, don't tell me that you are using an interpreted GC based language for some code on the server side. If you think you aren't, please verify that you don't. You should rule this out 100% Ever since PoE2 kicked in, this is a constant issue that apparently is very poorly profiled by yourselves. Last bumped on Jun 14, 2025, 6:33:26 PM	Posted by Gr1mscythe_#1926 on Jun 14, 2025, 4:03:09 PM Quote this Post
Judging by the nature of rollbacks that they are inconsistent, as in: use currency on item, crash, rollback, item returns to original state, currency gone. It's most likely that the issue is not with an interpreted language, which is just plain impossible to imagine given the scale and complexity of the simulation and would be simply financially impossible to sustain at this scale, and what is likely happening is that some code that is responsible for communicating with the database either fails to send, or the database rejects/overloads so only some queries are getting executed, and shortly after that the shard crashes because the state can't be properly syncd with DB. It seems this issue is ancient and present in POE1/2 so my guess is that they use the same database for both, and because its the same piece of software they used since original release, it suffers from scalability issues, old codebase and technical debt piles, but to switch would require monumental effort, and there is no guarantee that any other solution would work better for their case. And the lag spikes before the crash are typical of a program stalling due to spam of trace logs due to the same error every server tick. Last edited by testingthings#2158 on Jun 14, 2025, 4:44:34 PM	Posted by testingthings#2158 on Jun 14, 2025, 4:39:11 PM Quote this Post
" testingthings#2158 wrote: Judging by the nature of rollbacks that they are inconsistent, as in: use currency on item, crash, rollback, item returns to original state, currency gone. It's most likely that the issue is not with an interpreted language, which is just plain impossible to imagine given the scale and complexity of the simulation and would be simply financially impossible to sustain at this scale, and what is likely happening is that some code that is responsible for communicating with the database either fails to send, or the database rejects/overloads so only some queries are getting executed, and shortly after that the shard crashes because the state can't be properly syncd with DB. It seems this issue is ancient and present in POE1/2 so my guess is that they use the same database for both, and because its the same piece of software they used since original release, it suffers from scalability issues, old codebase and technical debt piles, but to switch would require monumental effort, and there is no guarantee that any other solution would work better for their case. And the lag spikes before the crash are typical of a program stalling due to spam of trace logs due to the same error every server tick. ChatGPT detected.	Posted by ARealLifeCaribbeanPirate#2605 on Jun 14, 2025, 6:00:59 PM Quote this Post
" testingthings#2158 wrote: Judging by the nature of rollbacks that they are inconsistent, as in: use currency on item, crash, rollback, item returns to original state, currency gone. It's most likely that the issue is not with an interpreted language, which is just plain impossible to imagine given the scale and complexity of the simulation and would be simply financially impossible to sustain at this scale, and what is likely happening is that some code that is responsible for communicating with the database either fails to send, or the database rejects/overloads so only some queries are getting executed, and shortly after that the shard crashes because the state can't be properly syncd with DB. It seems this issue is ancient and present in POE1/2 so my guess is that they use the same database for both, and because its the same piece of software they used since original release, it suffers from scalability issues, old codebase and technical debt piles, but to switch would require monumental effort, and there is no guarantee that any other solution would work better for their case. And the lag spikes before the crash are typical of a program stalling due to spam of trace logs due to the same error every server tick. There's a great deal of assumptions here. Assuming that one would do the right thing is a slippery slope. Believe me, I've seen some odd choices of architecture/framework/language in my time. Let's say that your description of their architecture is correct, then obviously there is a way of profiling the issue, it doesn't seem too obscure. Especially if it is ONLY related to the database, right? Furthermore, I do understand that software developers are expensive but I'm sure that they have several thousands dollars/euros lying around to dedicate it to this issue. It seems SUPER important, no?	Posted by Gr1mscythe_#1926 on Jun 14, 2025, 6:33:26 PM Quote this Post