The Unrealism Zone Server is Back!

[Arkon]

ScriptWarning: Pupaeheld DM-Turbine.Pupaeheld0 (Function Botpack.TournamentWeapon.TweenDown:0054) PlayAnim: Sequence 'Down' not found in Mesh 'Pupae1'

That's an example of a really dumb bug proving how much skill was in coding this package, it's a waste of time with WeaponLord. Suddenly default UT stuff must have weapons meshes/animations because of brain-farts - these are the laggers - more than expected; regarding to Suppressing ScriptWarning these being visual things blabbering in client too... is not a solution.

Yeah, that was when I noticed the CPU usage jumping too; the log was copied directly from UnrealServer Console live while observing Task Manager. so the ScriptWarning issue causing the lag was a given at that point.

A shame, too; it's a really addictive mod, which plays much better offline. Oh, well. I guess I'll bring back MH, and/or UT Survival through different listen ports, but they really needed a lot of work before dropping the server, and backing it up nearly three years ago. Or prepare a WeaponLord-esque setup by using StuffSwapper, which is an awesome mod in its own right.

[Dr.Flay]

Is there any chance that the UT2U1 Mutator by @MrLoathsome helps with this problem ?
https://www.ut99.org/viewtopic.php?f=7&t=4872

[Arkon]

^^Hey, Doc!

Server Update - Server should be live again no later than this Thursday evening.

[sektor2111]

Is there any chance that the UT2U1 Mutator by @MrLoathsome helps with this problem ?

This have nothing to do with that mutator.
Pupae doesn't have animation "Down", Barrell doesn't have animation "Down", codes have to be rewritten using an animation available doing "tweendown" from mesh else we can speak about the same IntroDude insanity, but at more intensive level. That mutator is for firing old weapons not for bad meshes deal.

[Arkon]

Another Update - The new unit has arrived, and with a nice surprise; not only did it come with a 64GB SSD, it came with 4GB of RAM instead of 2GB.

Will have it up by the end of the night!

Yet, Another Update - Server setup is not completed because I felt like s*** earlier.

No worries, though; so far, I'm feeling much better.

...And Another Update - Server is finally back up! Please note - there may be changes in the near future.



The Server So Far - Update

After some more testing, this Optiplex 160 with the Atom 330 is a beast compared to the Atom 230 I had before---especially seeing that it's a quad-threaded unit with an SSD.

The "Hatchet Gun", which was the weapon in the WeaponLord mod that caused the extreme lagging issues was tested again on this Atom 330, and it didn't have this issue. I even added "Arka" on the server, and had both firing the Hatchet Gun automatically while there were other bots running around with WeaponLord weapons, and there was still no extreme lag with either client. Tried it on a completely different network with the Lenovo; still no extreme lag. The ping did go up significantly with all clients, but the server was still completely responsive. :

The total CPU usage capped around 23%, and only one core was pushed closer to it's limit. If I did the same thing on the Atom 230 with only one player, the server probably would've crashed. And the icing on the cake was it was still running under 30 centigrade, which is even cooler than the Atom 230, ironically. On the other hand, I'll have to see how it will work with other players, if they decide to vote/play the WeaponLord mod via MapVote, but I highly doubt everyone will be running around with Hatchet Guns at the same time---especially seeing that there are, like 110 weapons in this mod to get lost in!

This thing is a keeper!
One more thing - If you see Arka, that's just me me on the Lenovo.

[Barbie]

I think that more CPU power helps in this case but I am not sure.

FYI the CPU usage over a week of a server running one instance of UT (MH) only; up to 5 players simultaneously:

proc0.1week.png (14.64 KiB) Viewed 1759 times

(CPU1 looks identical.) Of course peaks may not be recorded.

cat proc cpuinfo

processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 62
model name : Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz
stepping : 4
microcode : 1064
cpu MHz : 1998.828
cache size : 25600 KB
physical id : 0
siblings : 10
core id : 0
cpu cores : 10
apicid : 0
initial apicid : 0
fpu : yes
fpu_exception : yes
cpuid level : 13
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf cpuid_faulting pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm ida arat epb pln pts dtherm tpr_shadow vnmi flexpriority ept vpid fsgsbase smep erms xsaveopt
bogomips : 5600.16
clflush size : 64
cache_alignment : 64
address sizes : 46 bits physical, 48 bits virtual
power management:

Memory is also no candidate for a bottleneck:

system3.1week.png (8.52 KiB) Viewed 1759 times

I think UT99 uses only one core (thread) of a CPU even though there are more available?

Correct, it's a single threaded application, AFAIK.

[Arkon]

I think that more CPU power helps in this case but I am not sure.

FYI the CPU usage over a week of a server running once instance of UT (MH) only; up to 5 players simultaneously:

I think UT99 uses only one core (thread) of a CPU even though there are more available?

Correct, it's a single threaded application, AFAIK.

Holy Moses! You got a UT99 server running on an E5...anything?!? I'm assuming you're using it for much more than UT99, though, needless to say. Yeah, in your case, ANY bottleneck would surely not be the result of the hardware---that's a given (unless something completely unearthly, and oracular were to happen, which would be virtually impossible on a Xeon, I'm sure).

Yes; I also believe UT99 is single-threaded, as stated in your response to Aldebaran. However, I believe the "hyperthreads" would be utilized on a system like mine---not directly utilized by UT99, but only with how "Hyperthreading" works, itself. Forgot to mention this in my previous post; While attempting to kill the server with WeaponLord, the other two threads were sitting around 30-50% usage, so I believe the Hyperthreading function was offsetting that second physical core's load onto the threads, which would probably explain the massive increase in stability. The first physical core was mostly sitting below 30% usage, but was not idle. There is probably a lot more I wanted to explain about this, but I really need to pass out, for a while.

...I should REALLY remember to take screenshots of these things, too. Lol if I'm not feeling too lazy later on, I'll probably run this "stress-test" all over again, and take screenshots, but aside from being half-sleep, my Lenovo is in pieces at the moment. I also forgot to upload that video with the old Atom 230 server running with WeaponLord, but I'll get to that, too, I guess.

[sektor2111]

UScript is executed in a single thread but the rest of "access" using Win API and the rest of OS resources are using what kernel uses. I tested an UT server in 1 core and 2 cores - IT IS different, small parts of C++ calling WinOS libs are taking place in how many cores you have, the rest is running in a single core - if I'm not mistaking after my observations. I prefer 2 Cores (exactly what I have) and priority 1 step above normal, In this way I see timer values having termination "000" more exactly fixed values and not "723", "874" or such.
I'll show you something a bit later... I'm warming up machine first... Else fixing timer seems to have effects - I'm not sure if are a very need but results are as follows:
- 2 Cores higher priority without Tickfix:

Code: Select all

XC_Engine: Set XC_ServerActor UT-Logo-Map.XC_ServerActor0 as AdminLoginHook
ScriptLog: Addon0 is looking for XC_Engine if available...
ScriptLog: Addon0 has detected XC_Engine Version 19
ScriptLog: Addon0 will use native iterators...
ScriptLog: Cilia has new score.
ScriptLog: Game ended at 7.404762
ScriptLog: !!!!!!!!!!!!!!! CALC END STATS

See value for game-ended ? Now the same config but with TickFix bellow:

Code: Select all

XC_Engine: Set XC_ServerActor UT-Logo-Map.XC_ServerActor0 as AdminLoginHook
ScriptLog: Addon0 is looking for XC_Engine if available...
ScriptLog: Addon0 has detected XC_Engine Version 19
ScriptLog: Addon0 will use native iterators...
ScriptLog: Kara has new score.
ScriptLog: Game ended at 7.426400
ScriptLog: !!!!!!!!!!!!!!! CALC END STATS

The fact is that if I start server in normal priority values are not ending with "000" not even with tickfix - so I'm using 2 cores above normal priority and tickfix - now I have a visual confirmation about what it does. In this configuration MH2 keeps tickrate around 29-30 (small deviations), else it might develop values between 21 and 32 - false or true they do look different - else that 32 seems strange as long as my value configured is 30.
In times where I was testing using 1 core I could see some lags... randomly which were not supposed to happen. System processing something has been exhausted and UT server developed red flag. Probably a very powerful machine might run a server in each core but I really don't know if won't be skating at a moment.

[Aldebaran]

cat proc cpuinfo

The main difference to me seems to be the amount of cores available and memory.

[sektor2111]

Playing with cards open.
If you open some of those maps lagging in Editor and you'll ask MAPPERS like Iacobus_Imrlz (if well recall ID) or Frag'n'Brag or other experimented people, they are going to have a nightmare in how do they have been build. Majority of "sudden mappers" were dropping stuff (shit) in years when MH was poor in maps and everyone getting tired of default maps and "portations" with no end, and were making "Great Levels". To summarize my MH/MH2 runs smoother - yes, a dual server + MA. Why ? Sure, I think I have around 127 maps - global number in Mapvote MH+MH2+MA - good ones and a few fixes as demo prepared for curious people - those which nobody was thinking at Bot Support or supposed impossible to do - I did those modifications for my bots and me as main purpose - to not mention fixing packages directly without writing X+Y pages with "ReplaceFunction" - that stuff was not done in that time and I don't have originals any more. So imagine that I dropped in "trash-storage" the rest which are not suitable for my needs and which are causing troubles or BT-MH craps because my game-play is based on hunting not jumping or spamming. To not forgot mentioning stuff implemented in MyLevel dropping trash in console based on "development". So to speak Sorting stuff for server is a JOB Must-Have. I'm trying harder to stay on 1024 channels, the rest is NO GAME.
At this moment I'm using a process manager which is able to force CORES how I want but I don't make any modification. The process manager I'm using for Byte-Stuck (if will happen - because it looks vanished). Process-Manager is able to track applications and to figure if a process takes 90% CPU for some time and it will force an end-task operation automatically, actually I don't need to guard server any more in this way, sooner or later will get a restart if something goes wrong. Another thing, I'm using a memory optimizer tool for releasing memory when OS is bullshitting with cache eating resources. More problems which I had in the past now are history and that's why I can be relaxed in playing UT, the rest of problems might be future hardware crashes... or major changes in Internet structures needing other hardware.

[Aldebaran]

So to speak Sorting stuff for server is a JOB Must-Have.

I agree with you. The thing is some maps work on some servers well and on some servers not so good. And sadly these are maps that makes much fun. So I only want to find out if I can get them to work on my machine. But I agree that they should not be offered in mapvote when they don't work fluently.

[Barbie]

You got a UT99 server running on an E5...anything?!?

Ups, I forgot to mention that its is not a dedicated physical but a virtual (rented) machine. I guess the CPU info comes from host, and its resources are shared with some other VMs of other customers.
And from now on I'm not sure what my CPU graph shows: the usage of the physical or virtual CPU.

[Aldebaran]

Do you have rented one core for your ut installation or more?
And how much RAM is yours?

[Barbie]

Do you have rented one core for your ut installation or more?

No, 2 virtual CPUs are included. You can have a look at Strato - my rented machine is similar to that "VPS 200 Linux" (the smallest offer they have).

And how much RAM is yours?

As I understood the terms 2 GiB are fix and up to another 2 GiB are added automatically if the OS requires it ("Balloon"-memory).

[Aldebaran]

Thank you for the info, Barbie.

I found something. When starting the MH-NodeInvaders map the value of channels shown after pressing F6 increases to 1023 within few seconds. So the map perhaps need more channels and 1023 is a maximum? I don't know what the channels value mean. When looking after "channel" in server.ini I found nothing editable.
I tested another map: MH-HellGauntletV3. I had same issues right from start. When I move into the big corridor the channels value increases immediately to 1023, when leaving the corridor the values go back to about 730. When the value is 1023 it lags hard.