Atleast for the these recent reports, would there be any benefit to hooking the two archetype registration functions and checking for duplicates (or the unlikely hash collision)? Generating warnings messages à la geometry validation.
It could probably be hidden behind _DEBUG, some flag, or whatever.
Turning Extended Budget off did not make a difference unfortunately.
Tested with AMD 6900XT, AMD 5700XT, Nvidia 3060 OC tested so far.
For hardware context in my case:
AMD Radeon 6900XT 16GB
64GB of physical ram and 160gb Virtual Ram Pagefile.
Samsung 980Pro NVME drives.
W10. 400gb free space.
900mbps down/up connection here as well
The bug happens on my other PC as well with 5700XT 32GB Ram W10 though. Doesn’t seem like hardware makes a difference given how many different people have had it with various setups.
The Nvidia users who tested with me have the same issue. Albeit they crash more often in general than I do for whatever reasons. Various ones. Though all of us are using Extended Texture Budget. Its been the norm for years to use it for us even on other servers because most servers we’ve been on we needed it to not have disappearing roads and buildings.
In this testing where the screenshots were taken, only 15 interiors turned on and some scripts and vehicles.
If there is nothing that can be done to prevent these kinds of problems from happening, anything that helps us identify the archetype and/or the resource that is duplicating it already helps a lot
I sent the required data to @nta. I hope that the solution to this problem will be found soon.
This would at least help with visibility for self-serve crash diagnostics, but risk exists that users will still interpret it as ‘noise’, and it’ll only show up when things actually are already ‘broken’.
Ideally the refcounting functionality would return again but that one was a little difficult to validate, or the server would parse these files and warn/perform some fixup but that’d require supporting RAGE parser formats in server code which would be a lot of new code.
Hello,
I’ve created a way to repro this using even less maps, none escrowed, and created a script to automatically ‘trigger’ this issue.
Repro steps:
Server artifacts: 6418 Windows (happens on 6437 Linux too)
Game build: 2802 (tested 2699, 2189 too)
Tested with Extended Budget both completely disabled and half way through, happens on both.
… don’t say this actually needs to ‘run for 4 hours or so’ per repro, that’d actually completely ruin iteration time 
Well as I said previously I believe this issue is escalated by the amount of ‘maps’ you have running. I was trying to keep it as simple as possible in terms of resources to analyze hence why it took me 4 hours to ensure that those resources were enough to achieve the issue. I don’t know if there are any other factors that may increase/decrease how fast this state is achieved, as I’ve only tested the resource using my computer; could be quicker…could be slower 
For a quick repro, just run any maps you have available and you’ll probably be able to achieve that state in under 5 minutes; I’d send you more of mine but many are escrowed and as we said we don’t believe it’s really related to the maps being used.
Also, the resources we sent initially when we created this thread those contain more ‘maps’ and will be able to achieve that state in 30min-1hour… it’s better than 4h 
Today I’ve also been testing something else, I’ve rolled back my resources to the ones I was running back in January 2022 and this issue also happened using those resources that I know for a fact didn’t have this problem back then. I know this doesn’t mean much in terms of finding the problem itself, but yea.
Thank you!
For me it started to happen I believe after 2612 update but some say 2699. Thats when I really started noticing it. I believe it was fine before 2612. But that would have to be tested. A few people remember it being after 2699 specifically but because I’m not positive thats the time it started I wouldn’t bank on it as it could have been 2612. If I had to put a time frame on it.
Just left my script running for <1 hour, came back to it and started noclipping around normally. This is the state of the map after loading and unloading the already messed up zones, this is how bad it gets.
Completely fresh server, clear server cache, clear client cache running only the vanilla resources & all of gabz maps
The screenshot below is on the exact same session as the above and was taken shortly after - after doing a lap around the map (to ‘force’ other models to be loaded in and ‘push out the corrupted ones’. Once I got back to it, it’s fine again)
Back to normal this one, but the map in general remains completely messed up and never recovers overall, no matter what you do. (Try to SetFocus far away then back to the player, use the natives to unload Los Santos, load the Island, load Los Santos again, nothing recovers it unless restarting the game).
Again I know I’ve mentioned gabz maps several times in this thread but to make it clear that this happens just the exact same without using any single resource from gabz, so it shouldn’t be a problem with his maps. I just use his maps because I guess it’s safer to assume they are ‘well made’ using best practices and there shouldn’t be any duplicates/conflicted archetypes between them.
Test above:
FXServer-master SERVER v1.0.0.6418 win32
set sv_enforceGameBuild 2802
GTA 5 Textures: Very High
Extended Texture budget: 30%, but have also tested with 0% same happens
Have also tried playing around with Distance scaling, but no difference
RTX 3080 TI | i9 12900k | M2 Kingston KC3000 | 32GB RAM @ 6000 Mhz
Might be a long shot, But does the problem occur on build 2545?
Note that I don’t have any random collection of ‘maps’ available. I don’t actually run a server nor am I in direct contact with anyone who does, so I also don’t collect popular map/vehicle addons.
This might be a bit difficult to understand from your perspective as someone who actually does run a server.
Even if not related, it’s still required to simplify this being reproducible. Similarly, given the fact I have administrative access to the platform, I can temporarily grant myself access to whatever content fine for investigation.
Not necessarily the case as also shown earlier in this topic, but even if there are any such concerns these shouldn’t fail in a weird way like that especially if such mistakes are super common.
Either way, ‘it’ll take 4 hours or so to repro’ still isn’t very actionable, so again, uhh, yeah… having an actual set that’ll ‘break with 5 minutes’ would be a lot more helpful.
As above by the way, checking this on 2372/2545 would also help, there were a bunch of R* changes on 2699 and it might be something there became incompatible and needs a platform-side fix/workaround.
Checked this on game builds 2189 and 2545, happened on both as well.
I’ve forwarded via DM the collection of ‘maps’ I have that can reproduce this the quickest, in 5 minutes.
Hopefully will be enough to understand what causes this!
Thank you
Only a handful of things are like that in my cfg. Most things are separate.
I’m wondering if that has anything to do with it…
And here’s why I think so: Once upon a time I was installed the MHC from the ambitioneers. I followed their installed directions to a T. No matter what I did, their building wouldn’t render, only the lobby.
Eventually, I found out that their resource DID NOT like being started this way and needed to be ensure specifically towards the top of our server’s directories server-data/resource/mhc_directory
I wonder if there is something with this ensuring method that could prevent proper mounting of files…
Edit: There is also a chance what is causing this are maps with outdated files. As game builds change, the way R* updates things can break older resources… The reason this issue could be happening for servers with a lot of maps, and servers with 1 map installed is that there are specific maps affecting/breaking the data.
I have a post pending approval and I am able to remove these lod issues by removing some of the “older” or “conflicting” mapping from the server. I would stress making sure if we are to continue looking at this issue, we are testing with maps that are up to date and support the latest gamebuild.
Are using a resource called “maploader” by MorovianLion? or are you letting them standalone and the vanilla stuff is handled by bob74_ipl?
EDIT: Ok, so, I can’t speak for yall, but I figured out the issue on our end. We use a resource by MorovionLion called ‘maploader’. It checks your distance from interiors/ymaps and loads and unloads ymaps defined in its config.lua.
The issue we were having specifically has to do with children ymaps being unloaded messing up the parent lod ymaps.
So, I’m essentially not saying that maploader was the issue, however, if there is an instance where you have a ymap with a parent and the child ymap is being loaded/unloaded improperly without consideration for the parent ymap, this issue will manifest.
Now, is this the only cause/explanation for this issue? No, but it is definitely worth considering.
This is a big deal with the Gabz maps because so many of his maps are replacement maps that he has to remake both the parent and child ymaps and replace them with his own. So, yeah, there’s that. @ me for any questions please.
Resource in question: Maploader [standalone][paid]
No we don’t use that script. Just bob74. We had another script loading some ipls too. But I turned off all of these when testing. Its one of the first things I did to make sure no scripts were causing it for us.
Eh Does that script even work to fix Rockstar Editor mode crashing in FiveM? Doesn’t seem worth it even if it does if its causing those issues for you.
I use it for optimization purposes. I assume since the interiors and everything attached to that ymap aren’t loaded in, that’s less strain on player’s with under performing pc’s.
I’m sure this has absolutely nothing to do with the problem.
