引用自 N00B3NATOR：

Need any help with beta testing it? I'm sure some of the regulars wouldnt mind helping test things out.

I'm going to just roll it out and everyone will test, I've been busy at work and personal obligations but have been working on it in the little free time. Hopefully I can release very soon, maybe first look tonight.

Hi Dragon, have you tried this fix: http://psteamcommunity.yuanyoumao.com/groups/SernixINS/discussions/0/133260492062963221/#c133260909498400152

Yeah, we're doing a minor rehaul from a combination of previous planned work, classic discussion thread, and recent ideas from INS server community with help from their admins and hodor/blueshy.

Im thinking of double jump more of a utility for bots, currently if they get stuck, they die. Lots of variables to consider + immersion too. Ill hold off and maybe just throw it in once and see if previously bad maps play better.

Im trying to minimize what I need to modify so I'm going to try to implement the points in a way that is minimal and not a huge balancing act, I almost burned myself out with just updating models but luckily things started to work and it moved forward.

Priority now is simple point system, vip mechanic, more classic spawns, and rehaul bot spawn system so we can optimize performance. Perks should be easy to add.

EDIT:

Yes, m249 is still modded but can be overwritten now when new update rolls out.

Great analysis Hodor. I'll take into consideration when making code changes. I finished new models and they are amazing, took too much time though but slightly justified with reduction of 150+ MB of asset space and more customizability for players. Blueshy is finishing up new carrier weights which look great. I'm going to add our own point system that makes sense. What perks do you like from INS? Double jump help bots get stuck less?

Thanks Hodor for the amazing feedback and effort, you are one of the main driving forces for these changes to be implemented. Blueshy was mentioning some of the negatives as well as positives. INS admins were telling me about some of the woes they want to correct too.

Out of talking with everyone, there are some great ideas that'll take less effort to implement with big changes but ultimately getting to a more classic feel. I already started working on some of the theater side yesterday but ran into a few headaches. I really want to show ammo/armor aesthetics for players but a huge part of me knows that aesthetics doesn't translate as much in gameplay though I think this will help in long run with new armor types like bomber armor or security jug armor.

So after talking with INS OP Bots admins, I got a ton of insight and have an idea of what I want to tweak and add.

I'm going to be rolling out in 3 phases this week if I have the time which I should.

Phase 1 will be just modifying all the sourcemod settings to feel more classic, optimizations, armors for bots, reduce max players from 18 to 17 and maxbots from 20 to 18 alive at once (2 jugs, 2 bombers with long delay spawn and reduced armor).

Phase 2 will be new spawn system, this may allow us to increase player count back to 18, though not sure (this might get moved to phase 3, too many unknowns).

Phase 3 will be added point system to key items, maybe OP items, perks, and VIP which I have some ideas for.

引用自 🎄Hodor🎄：

Hm, the choke on sernix usually happens when the majority of our players are not touching map brushes. Like everyone walking on the bridge, or holed up in a 2 story house like estates. It's consistent and begins when the bots mass spawn and no friendly has died yet.

There are other rare cases where players are on proper brushes but the bots are stuck jumping behind some object until they die.

I'm assuming it's pathing since it happens at the start of a new objective and the bots just spawned far away. From what I read on 60op server discord, nothing much is running on their virtual machine. I don't know if they also have a huge pathing load but simply overcome the choke with sheer cpu power.

How about running their theatre as a test? Seems like the easiest way to guess if we have a culprit mod and eliminate this theory.


Over on bridgeatremagen on their side, I've gone into spectator and flew to a secluded part of the map where the game only needs to update me at 15-20k/s. The tick time variance is truely 2ms at most.

While on sernix it reaches 30ms when the "pathing lag" is present and bandwidth used is well under our limit (client net_graph 4 reports 0 loss 0 choke too)



Funny thing is this choke has gotten worse on bridge and now starting to creep into maps previously unaffected in older iterations of sernix threatre. I used to compensate my high ping by pumping more bullets into bots than the average player, 3 bullets more on average. But I have to pump around 7 more shots now.



How about tesing their theatre before putting effort into design changes? They have bandaid balancing like endurance perk to overcome weight limits, speed perks to boost speed. Stuff that seem lazy on paper but worked well when paired with the vip grown gear pool.

Playing the same map with a crappy team results in a poor kit army. High recoil, low tier guns, slow speed, no sunglasses, no gas mask, etc. But when a good vip and signaller show up, the growing gear pool changes win-loss odds over time. Favourite part is the gear dynamics that follows if the team is smart enough to protect those two.

I just updated our betterbots.cfg and removed any type of pathing cvars (was only a couple) and any bot settings that supposedly increased bot listen range etc. so these are now vanilla settings, reaction time etc. should be the same. This is currently live. I just opened a dialog with Mutex on discord to see if he can give any feedback.

As for switching to their theater for testing.. This might give us false results and also make things unplayable. They are using their own sourcemod scripts which I do not believe are shared publicly and probably work hand and hand with their theater (VIP tracking, Signaler allow respawns among other class specific perks).

The best method would be to modify or disable some of our sourcemods while the theater would be left alone. There might be a loop in sourcemod that is unoptimized or causes unnecessary stress on server. I am going to work on this tonight if I have time.

Now lets talk about OP Bots vs Sernix features.
They have some great ideas and I was hoping Axin would help develop new theater features, I would love to emulate some of their ideas on our server while getting rid of bandaid type of balancing like automated respawning and reinforce saves, slow the pace down a bit and optimize server. The issue now is I just dont have the time and energy to invest in large rehauls but I will make time if there is potential.

Now when I say emulate ideas from their server, I do mean the general things that make it fun but also keep the spirit of the original Sernix which has always been team focused, more focused on immersion, and less arcadey, though not milsim.

I think change is good but Id like to try and make big changes now with as little effort as possible (so mostly sourcemod stuff). Theater changes do take effort and testing, though it just depends on what we are changing. Rolling back things is a little easier. Ive been thinking of going back to vanilla insurgent models and possibly adding in ISIS models ones later on which are less demanding on asset size/memory cap.

Ill reach out to Mutex but if you wanted to carry over theater changes, what are some weapons we can carry over easily (mainly unique ones like gas etc.)

引用自 🎄Hodor🎄：

My other theory about your cpu choke creep is inertial ragdolls turned your server into a stealth bitcoin miner.... bad of me to stereo-type russian originated mods, but it did happen early last year as did your increased choke. =P


The news is about trojan for clients, but possible for servers too.

https://www.pcgamer.com/hackers-hijacked-pcs-using-source-engine-kill-animation-exploit/

Thats pretty crazy and may explain how our previous servers got hijacked (though those servers were not updating windows either) aside from NFO. Our servers have very high security and valve fixed that exploit so I think this is just a persistent issue in general.

Hodor, that is interesting.. I'm really curious what can be causing the majority of the stress on our side. Is our bandwidth always 100k/s or just certain maps, moments? Maybe if you see server owner on 60 OP bots you can ask for bandwidth settings and or if he knows what may be causing stress in his testing. I doubt its just bot pathing, I dont think I even modify cvars for that anymore. Server side ragdolls do contribute though the stutter usually happens on bridge before anyone dies and bot rags are client side.. It can be the amount of sourcemods in general communicating tons of data and maybe the continuous flow of bots respawning.

I've been thinking of doing a bit of a pull back rehaul for Sernix to get a more classic feel on the main build instead of splitting builds.

What are some main gameplay elements you think we can emulate from other servers and or what those servers do well/better in general (not so much theater content, but point to point capping, counter-attacks, checkpoint flow).

Some main ideas for a more classic rehaul:

• Lower intensity overall of Sernix
  
• Bots per point would have a much smaller pool of lives
  
• Reinforce timers increased
  
• Bot armors reduced including jugs. Armor would be lower than security (first time ever).
  
• Security armor would be slightly reduced as well, but not as much as bots
  
• Ambush timer would be almost non-existent or off completely.
  
• Wave spawning would be increased significantly or completely turned off depending on initial bot pool.
  
• Counter attacks shortened.. (need advice here given above possible changes), do we respawn everyone? How do we make them challenging enough without making them really long?
  
• Possible remove extended mags for secondaries that already have 30 round mags.
  
• Increase jug spawns (though with further armor nerf, they may not need it).

If you can track when people die and spawn rag to see if that changes things massively in stress/bandwidth or other elements like when multiple players are dead, bots reinforcements deplete. We can try disabling some portions of the mods

Thanks Hodor! You've been a great help.

引用自 dangus：

I would say the jammer ring could totally be removed and I don't think anyone would care. I do like some indication of which bodies are workable since there's so many bodies that stick around after they should have vanished.

Speaking of vanishing bodies.... why do bodies persist longer than weapons? Fallen weapons are useful and bodies are not..... it would be great if that timeout was reversed

I reduced it because I think it helps with client crashing though I'm not sure. Too many Edicts bug I believe has to do with too many models etc. from what I remember from devs.