{HoF} Evil's Sniper Deathmatch

Image
Image
Image
Image
Image
Post Reply
Nelsona
Posts: 1117
Joined: Sat Sep 30, 2017 5:03 am

Re: {HoF} Evil's Sniper Deathmatch

Post by Nelsona » Sun Feb 16, 2020 12:56 pm

Heading to the next Administration Problem. A map with a name self-explanatory but used in playground repositories being out of UnrealEngine1 and causing issues from now on won't be my problem.
If purpose of this #1 server is running all crapped up junks, with all respect, I'm inviting you to do the setup if you think that your server has another Engine more capable than another the same used everywhere UE1.
Required_Other_Management.PNG
Required_Other_Management.PNG (23.19 KiB) Viewed 2407 times
You should have a clue that CPU power it's not the same thing as Engine's power. Engine is reaching at boundaries before to see CPU loaded at 40%. UE1 it's not done for running 2k4 maps.

With regard to some tweaks and my notes. I see that server has strange sudden errors since... Was it moved ? The most of interesting finds are somehow linked:
- in account messages by NfoServers they told me about some DDoS ing which they blocked in M.G.S. Address was coming from some Amazon nearby names;
- in account that around error time another IP was pinging or whatever ICMP packet coming from the same Amazon related range deal;
- in account that some secure query was around the same error time by another Amazon related range.
Okay, okay, I understand what they want. They want to get into firewall policies for good... What's their business here ? Are they hacked with some rootkits and they don't even know what's going on ? Is this just an network echo ? Is this IP spoofing ? Light me, please.

At next reboot update:
Server will print memory status each 2 minutes - real-time minutes not game-speed minutes. Perhaps we need to know memory load around time when things are going down-hill.
Not often maintained
My UT Mapping works...
Learn the rules like a pro, so you can break them like an artist.
- Pablo Picasso -

Nelsona
Posts: 1117
Joined: Sat Sep 30, 2017 5:03 am

Re: {HoF} Evil's Sniper Deathmatch

Post by Nelsona » Mon Feb 17, 2020 10:00 am

Continuing.
When I'm writing these server banner shows RED - server in outage. I quickly went to see what's going on and I captured this. You might want to fix this UTES thing unless server will need a guardian or a twice restart per day.
Will_be daily.PNG
Will_be daily.PNG (57.17 KiB) Viewed 2384 times
It's all about not wrapping array elements count and data goes out of boundaries. These lines are not saved because application is closed without to throw all write-buffer content on disk... If you look at error it is about a failure at attempting to allocate space for an array and in top - before player join, we do have an access out of bounds. Probably some data from that 20000 elements array will never fit into an array having 5000 elements... :?

Now you see why I'm getting mad ?
It reads: Windows can check on-line for a solution to the problem.
Really ? I did not know that this Window guy it's so addicted to UT and he wants to solve the problem, perhaps it would be priceless to stop being retarded since forever - because he won't solve anything on-line - and to close f...ing program allowing server to reboot correctly.
I read about auto-closing programs and some key it's required to registry but here I don't see if that type of registry is concerning a Window Server because as I see here we have other sort of registry wrapping - the said key it's not there and I think it should be created in hoping to not see it useless. I might ask others if exist a solution that can be implemented - and I want to see a FIX for UTES too.

Update: At next reboot, in hoping to not have a server lock-down a la Windoze crap, UTES.ini will have a replacement with all arrays cleaned starting with position 2001. Update will be reinforced until problem will get solved. What is happening with that array I think it triggers a memory allocation error due to the wrong script. In such a script which I did in a MapGarbage flawed lab version, I think I corrupted memory and I had to reboot machine, lol.
Not often maintained
My UT Mapping works...
Learn the rules like a pro, so you can break them like an artist.
- Pablo Picasso -

User avatar
SC]-[WARTZ_{HoF}
Site Admin
Posts: 394
Joined: Wed May 10, 2017 7:08 am

Re: {HoF} Evil's Sniper Deathmatch

Post by SC]-[WARTZ_{HoF} » Mon Feb 17, 2020 4:21 pm

I've been meaning to combine the AKA playerlist script with UTES so there would be just the one array. Just haven't had time to work on it.

Nelsona
Posts: 1117
Joined: Sat Sep 30, 2017 5:03 am

Re: {HoF} Evil's Sniper Deathmatch

Post by Nelsona » Mon Feb 17, 2020 5:20 pm

You can take your time because I've generated an INI with empty slots which takes another time to fill up. From time to time I will save top stats - IN HOPING to not get over 32 Bits numbering system. Score hunting was never good in my opinion. It's easy to hit boundaries now days...

I forgot a mention:
Main ini is reduced to minimal load for a server according to EUT ini recommendations, more exactly assets which are addressing Player are now removed from server: Sound drivers, render, input, etc.

Code: Select all

[Engine.Engine]
;GameRenderDevice=SoftDrv.SoftwareRenderDevice
;AudioDevice=Galaxy.GalaxyAudioSubsystem
NetworkDevice=IpDrv.TcpNetDriver
;NetworkDevice=XC_IpDrv.XC_TcpNetDriver
DemoRecordingDevice=Engine.DemoRecDriver
Console=UTMenu.UTConsole
Language=int
;GameEngine=Engine.GameEngine
GameEngine=XC_Engine.XC_GameEngine
EditorEngine=Editor.EditorEngine
;WindowedRenderDevice=SoftDrv.SoftwareRenderDevice
;RenderDevice=GlideDrv.GlideRenderDevice
DefaultGame=Botpack.DeathMatchPlus
DefaultServerGame=Botpack.DeathMatchPlus
;ViewportManager=WinDrv.WindowsClient
;Render=Render.Render
;Input=Engine.Input
;Canvas=Engine.Canvas
Probably even DemoRecordingDevice driver it's useless in this environment...
Goooood... The next appearance - I've installed that memory/tick reporter saying what is there each 2 minutes... We do have problems or we are under-resourced and then I'm not surprised about that tick-rate vs ping response at players... I think we have to test XC_Engine v24 in the same environment but other server instance on another port.
Take a look at this:
Tck_15_of_40.png
Tck_15_of_40.png (513.22 KiB) Viewed 2369 times
Report is a truncated string granted by Core.dll natives...
Doing some math: 1000 ms ÷ 15 ticks = around 66+ ms. These 66 milliseconds are added to networking original ping response growing player with 80 ms at 146+ ms and then this is visible on player's end. This can be seen even in browser without to join in game.
Not often maintained
My UT Mapping works...
Learn the rules like a pro, so you can break them like an artist.
- Pablo Picasso -

User avatar
SC]-[WARTZ_{HoF}
Site Admin
Posts: 394
Joined: Wed May 10, 2017 7:08 am

Re: {HoF} Evil's Sniper Deathmatch

Post by SC]-[WARTZ_{HoF} » Mon Feb 17, 2020 6:17 pm

Whatever you think is best regarding XCGE version feel free to install. I have saved the UTES stats ini this morning containing for all clients. I will be upload new scoreboard minus UTES stats so is for future development optimization. I've explained to most clients that I will retain their individual stats from prior version for when time comes to re-integrate them in the future. I need to add code that auto prunes inactive clients per so many days.

Nelsona
Posts: 1117
Joined: Sat Sep 30, 2017 5:03 am

Re: {HoF} Evil's Sniper Deathmatch

Post by Nelsona » Mon Feb 17, 2020 7:43 pm

Keep in mind that Players based on names records were around 9000 if not even more, I'm not looking there right now, (duplicates included) and those stats are fitting into 5000, 4000+ are gone for good.
I think here we want to talk about a "reset" with a clean UTES tested in lab for 10 slots and figuring how to remove old records, and then setting up 5000 or less - I assure you that we do not have so many players, they are duplicated for their reasons, more or less dumb or wise. I have to look in that Server about physical capabilities because at first check it do looks stronger but here I don't see how strong might be, or this is a networking issue and then I have to see which version of IpDrv.dll it's used. Next move, setting up update batch for the rest of files: DLL, EXE, INT in order to do an "instant" self-update when server is rebooting - so far it's dedicated only for SniperServer (I removed it from Redeemer because that's another story) and it's operating main UT files INI U UNR UAX UMX UTX and nothing executable.
Whatever will be installed this configuration must have a backup placed in a safe spot. I would be curious to see if UT.exe is acting different than UCC here, that thing is granting access to configuration in run-time to several mutators.

Another stage it's my home server strategy, server starts with UCC but ignited by Start command which is able to setup a higher priority to server compared with the rest of applications, this can be "feel" by player in certain moments, it's not the same as default priority.
Not often maintained
My UT Mapping works...
Learn the rules like a pro, so you can break them like an artist.
- Pablo Picasso -

User avatar
SC]-[WARTZ_{HoF}
Site Admin
Posts: 394
Joined: Wed May 10, 2017 7:08 am

Re: {HoF} Evil's Sniper Deathmatch

Post by SC]-[WARTZ_{HoF} » Mon Feb 17, 2020 8:07 pm

Nelsona wrote:
Mon Feb 17, 2020 7:43 pm
I think here we want to chat about a "reset" with a clean UTES tested in lab for 10 slots and figuring how to remove old records, and then setting up 5000 or less - I assure you that we do not have so many players, they are duplicated for their reasons, more or less dumb or wise.
I think the "Evil Weapon Test Server" on my local VDS would be good for UTES Testing. If you can figure how to remove old records in that array great because I haven't found a solution yet. Curious to see what method you come up with.

Nelsona
Posts: 1117
Joined: Sat Sep 30, 2017 5:03 am

Re: {HoF} Evil's Sniper Deathmatch

Post by Nelsona » Mon Feb 17, 2020 8:12 pm

Then I need some time to figure how goes the mod, perhaps weak records in points in front of new records are a priority for removal - zero things more exactly. Player entered and quitting, doing nothing. This new comer has nothing and then it should be wiped.
For me Time related things are not my best point, when I'm thinking at OLD records I need to establish a "days-counter" from last seen time in order to purge data if said player is gone for 1 month or such - this time deal it's interesting to do. You cannot keep records dated from 2nd world war... The solution perhaps is already in Nexgen - heavy task...
A more cheap solution: Updating "new scorers" comers on top of database. In time records from bottom are removed - even if are champions - if they are not coming back so soon - this is not a museum. No time deal no other funky processing, just based on score vs presence. Some score ? Move to database. Score but not seen that much, going down and down, deleted. All cycle it's probably based on 3000 records. At end of each match are sorted the best ones, and database works more dynamically instead of keeping records for ages.

Another solution:
Wipe everything each month - that simple. Are you the best ? Okay that was in November, perhaps in December we have other champions. Huh ?

Edit: In "idle" time, run-line will get turned into:

Code: Select all

start "Evil_Sniper_Server" /WAIT /ABOVENORMAL /MIN exec_hof XC_Core.XC_ServerCommandlet UT-Logo-Map.unr?Game=BotPack.DeathMatchPlus?Mutator=MV..........
I'm not sure if latest UCC provided by Higor is helpful for nasty errors, as I recall I dropped it away because logging was wrong, funky things have been written in log file...
The result will be probably two windows: Master Window launching another Second Window "Titled" with game-server name - I hope to see that closed for good when a game-crash occurs...
I gotta study what's the deal with registry here and also how to compute affinity - which CPU/s is/are running UT because several natives are working in multi-cores ignoring myths - I could see that when I set affinity manually from TaskManager in two cores, it's visible well which two cores from four have more stress. In INI at XCGE section I defined that thread = 2. We might have too many servers running in one core... We need to take in account how do we share CPU resources between servers instead of forcing only one core for everything.
And... I think I got exactly what I wanted, exactly for game-servers or such...
https://www.raymond.cc/blog/disable-pro ... rver-2008/
Respect for all tech dudes from this Planet !
Not often maintained
My UT Mapping works...
Learn the rules like a pro, so you can break them like an artist.
- Pablo Picasso -

Nelsona
Posts: 1117
Joined: Sat Sep 30, 2017 5:03 am

Re: {HoF} Evil's Sniper Deathmatch

Post by Nelsona » Tue Feb 18, 2020 7:21 am

And... we have updates.
RedeemerSniper is running in priority "AboveNormal" for allowing the rest of devs to work (HighPriority it's too much) and running in the first two cores:

Code: Select all

start "Redeemer_Sniper" /WAIT /AFFINITY 0x3 /ABOVENORMAL /MIN exec_hof XC_Core.XC_ServerCommandlet UT-Logo-Map.unr?Game=BotPack....
Evil Sniper Server is running as well in priority "AboveNormal" and using next two cores:

Code: Select all

start "Evil_Sniper_Server" /WAIT /AFFINITY 0xC /ABOVENORMAL /MIN exec_hof XC_Core.XC_ServerCommandlet UT-Logo-Map.unr?Game=BotPack....
All of them are minimized by default for preventing wasting CPU resource with rendering text interface - it looks like this is even affecting tick-rate.
Both of them will use a Main Start Window and the second for said managed server task.
Probably resources are now shared properly as long as I could see these, of course this is not a heavy load but it explains a lot:
4 Players:
4PL_39_of_40.png
4PL_39_of_40.png (130.53 KiB) Viewed 2337 times
And for 5 players - one more it's not even in account - 4 USA + 1 Australia
5PL_39_of_40.png
5PL_39_of_40.png (484.22 KiB) Viewed 2337 times
I think now we have some setup and for the rest it's about CPU power.
Advice, let servers minimized all time after inspecting them.

What else I could see ? Server was configured for granting a higher priority to background processes and not for foreground applications, which is not that good because we do have applications launched, for having logs, not processes blindly closed on errors. By increasing priority for servers it looks like I got the resources needed for a better tick-rate, we do need this priority in order to run them smoother.
Not often maintained
My UT Mapping works...
Learn the rules like a pro, so you can break them like an artist.
- Pablo Picasso -

Nelsona
Posts: 1117
Joined: Sat Sep 30, 2017 5:03 am

Re: {HoF} Evil's Sniper Deathmatch

Post by Nelsona » Tue Feb 18, 2020 7:20 pm

And results noticed:
- USA players have ping under 100 ms even if are 12 Players connected;
- at 13 players charge we still have 34-35 tick-rate I did not see values going under 30.
Perhaps we have a better setup right now, I think images are no longer needed.
Not often maintained
My UT Mapping works...
Learn the rules like a pro, so you can break them like an artist.
- Pablo Picasso -

Post Reply