Server Issues? Questions? Tips for 4.33.1?

[Sinu]

Still got some interesting AI behavior - my lead and I were taxiing out of Seosan when we observed the usual AI spinning and teleporting.

Suddenly, lead went "poof" in a fireball and exploded on the taxiway with no warning. Apparently, he collided with an invisible F16.

Very strange.

[=VG= SemlerPDX]

Ouch! What's always frustrating to me is turning off HQ fragging at Seosan before any flights have fragged, time stopped - then getting into a mission in the first hour of the war only to have to compete for runways at Seosan with AI Vipers!! Not like there's even time for them to have been diverted from another field or any such reasons.

I wonder if other servers have the warping AI issue...

[Sinu]

I did run a test today - my friend and I took the current VG server save .cam file and he hosted it off his PC.
He's in the UK and I am in the US. We didn't experience any AI lag or warping problems. Of course, that is with only two clients and is not really a good test of the loads the dedicated would experience.

That indicates to me that the problem doesn't lie with the save file. I have also noticed that it is more of an intermittent problem on the dedicated server. I wonder if the server install of BMS has some kind of strange problem? Certainly unlikely but it is weird behavior nonetheless.

On the bright side, we were able to get a few good flights in: http://images.akamai.steamusercontent.com/ugc/499147520067881016/6CCC938B9F830F1D9F0BB8681319F5F1DB3C94F6/

[CobaltUK]

The "AI spinning and teleporting" can be seen with the minimal usage of 1 player so do no think its load related. I cant help but think that the server is being strangled now and then for some reason. I also wonder if the pc spec as recently described is capable of doing the math and dealing with the in/out data for a busy server. But that would not apply to the minimal 1 user situation.

I have not seen on bms forum any guide to minimal hardware for a `busy` campaign server.

[Sinu]

If the server has a CPU bottleneck then we are SOL.

I did see on the BMS forums that there is a program that can load the server client into the 3D world but does not draw any 3D textures. This might be something to try to relieve the load on the server CPU/GPU.


Note: I literally no nothing about how this thing works, I just took a quick read through its release thread.

Link: http://www.bmsforum.org/forum/showthread.php?24564-REL-dedicated-server-for-Windows-with-no-GPU

[Sinu]

I wouldn't mind the lag if it only effected AI on the ground but it is game breakingly bad when in A-A combat...

Reference this ACMI file:
http://puu.sh/oBlm0/cfe03f7a51.acmi

Note the spinning behavior of the MiG23s, SU-25s and the Patriot missles.

[Brain]

I did run a test today - my friend and I took the current VG server save .cam file and he hosted it off his PC.
He's in the UK and I am in the US. We didn't experience any AI lag or warping problems. Of course, that is with only two clients and is not really a good test of the loads the dedicated would experience.

Warping happens even with one player on the server, even when the server PC has just been restarted.
We got 18000+ ground units across several hundred battalions and hundreds of planes airborne at the same time. Just because they are not "in a player bubble" doesn't mean they are idle. The war goes on 24/7 and all of the battles still need to be calculated in some way.

If the server has a CPU bottleneck then we are SOL.

Since you had no issues on your testflight I would assume connection/distance is not the issue but processing power. BMS might not be able to fully utilize a quad core CPU but having the operating system in the background on a 2.5 ghz dual core doesn't help. So, yeah, we might be SOL for the moment in that regard.

I downloaded the campaign and let it run for ~48 hours. Spoiler alert! https://www.dropbox.com/s/gmy5q8za4h0po2w/VG-Camp-29-4-48h.cam?dl=0
Well, it's not as bad as previously but DPRK is still being clubbed pretty hard. Most radars you'll see on the map are actually just search radars. It will be well sufficient for the weekend but after that it's going to be pretty uninteresting again (read: little to no opposition). And remember: that's without a single human flight, just me looking at the map (and going to the grocery store).

Looking at the squadrons I noticed all stock Viper squadrons are still set to HQ. For months we had 3 of them disabled (Kunsan, Osan, Seosan) and still the campaign won itself. What exaclty is the thought behind it? How is it supposed to help when BLUFOR is already overpowered?

... It may take a few days, but no matter what changes we make AI DEAD/SEAD flights with HARMs will render those sites inoperative after some time period. It's just a matter of how long that takes.

No, that's wrong!
You can edit the AI priorities so they won't fly that many SEAD missions. I brought that up twice now, even with a guide on where to find these settings in Mission Commander!

One well executed human flight against a very active DPRK airbase can change the entire "balance" that we may think we have setup.

And that's what it should be all about: player action and it's influence on the campaign.
Take out airfields for air superiority and limiting DPRK ability to strike back, take out engineer battalions to delay repair of said airfields, fly SEAD to open corridors for follow up missions, destroy war production to limit supplies .... with the default prio settings the AI will do all those things for you and it's just a matter of time until there are no valuable missions left to do. Just changing some planes around only changes how fast this happens. Everybody playing on the server has been able to observe this.

I can't deny that a well executed OCA strike will tip the balance but if it was a hard fight it feels rewarding and it's only a temporary victory. You'd have to destroy the engineers tasked with repairing as well, which is suicidal with all the air defenses in the area. So, a 'permanent victory' would require multiple missions or a larger flight/package (haven't seen many people online, unless that changed during the last week).

If you guys are meeting up in TS these days, drop me a line so we can go over some scenarios. I'm starting to feel like I'm unable to clearly express my concerns and explain my reasoning, so a direct dialogue might help answering all questions.

[Kaos]

I totally agree with Brain last post. I hope you will move in that direction and make the campaign more interesting and challenging.

[=VG= SemlerPDX]

Brain LAN_WROTE ...

Looking at the squadrons I noticed all stock Viper squadrons are still set to HQ. For months we had 3 of them disabled (Kunsan, Osan, Seosan) and still the campaign won itself. What exaclty is the thought behind it? How is it supposed to help when BLUFOR is already overpowered?

Oh, holy flipping face palm!!! I messed up big time there!! I cannot find the emote to express my embarrassment. I will check for online players and fix this immediately. This was a big mistake and I never meant for those squadrons to be active again - in fact, I intended to disable the A-10's at Osan or Kunsan wherever they are to see if that gives the DPRK that slight edge that is not overpowered. Big mistake.
!hang

[Brain]

Well, we won't run out of playable F-16s anytime soon :D

After constructing some kind of "CPU Cooler of Dr. Franken" my (formerly) passive cooler now seems to do the job. LOL

Encountered a pack of MiG-23s followed by Frogfoots. TGP showed no signs of the MiGs spinning. Frogfoots held formation nicely, popping flares, going evasive.
...there's more but you get the idea. It ran perfectly.

The catch is I saw VG was looking at the map again ;) Semler, did you adjust the "player bump time"? We found this thing waaaay back and it fixed the server getting kicked periodically. Hidden in some obscure config file.

Airspace was pretty crowded at the time, DPRK flying sweeps over Osan, multiple escorted Frogfoot packages.
That doesn't sound like particularly low server loads, especially with 3 active pilots at that time. I don't know if anybody waved their magic wand or what.
I remember something along the lines of hosting in 2d is not supported. Can anybody confirm that?
No graphics, flight physics or plane avionics to worry about, no "bubble" around the host's plane...and all "bubbles" are guaranteed hosted on a client PC. Sounds like pretty good hosting conditions if you ask me..?

Going up again in a bit, choosing some crowded airfield (if any). Let's see how AI handles something simple like taxi.

EDIT: Wingmen seem to work. "Duck" took an AI wingman and he seemed to be fine. Previously they would crash on the airfield or just vanish. 6 Players flying at the same time, smooth AI movements, excellent.

[Brain]

Another mission done:
Gave some support to a human 2-ship out of Kunsan. They were engaging some MiG-17s which behaved perfectly fine. When both left (after 1 got shot down) I could see the MiGs jumping once, then everything went back to normal. That's the typical host handoff you'd also see from tankers. In case of a smoking MiG-17 it looks just like the "lag" we usually see from AI.

On my way RTB I picked up the remaining plane from the other package, now AI controlled. Flying close formation was no problem at all, formation landing possible (which I screwed up), perfect taxi back to parking position.

I'd say now client's properly "host AI in their bubble". Whatever interfered with it seems to be gone since the server is not in 3D anymore. Average server population seems to be increasing, so there should be some people able to comment on that.

[=VG= SemlerPDX]

Server has been running at the 2D map for days now. It actually seems to have made things more stable. Based on what others have written above, this concept that the "player bubble host" is taking up some of the workload of AI units close to said player may be the most beneficial route as we cannot afford to upgrade the BMS server hardware at this time, and every one of our PC's is definitely a few generations ahead of the server CPU (whether AMD or Intel) and far more capable.

I'll say this for the old workhorse of a server: I've tweaked and fine tuned it's copy of Windows to be as slick as possible, there's only like 41 processes running at any given time including IVC server and the BMS server, and it has mostly modern cable internet with a stable ~35Mbps down and ~12Mbps up.

Let me know if it crashes, of course - it's harder for me to verify server health when it's at the 2D map, so I'm gonna rely on word of mouth from one of you for the most part when it is not obvious that the server crashed. When it doesn't display a crash popup from Windows or the like, I just cannot tell.

We'll see if this works like this.

[Sinu]

I can say AI performance has improved hugely ever since moving the server to the 2D screen. I haven't observed flipping or spinning AI and ground vehicles are still behaving fine.

Seems like all of the things relating to hosting BMS servers is just done by tradition or rumor. Nobody seems to have documentation on how different actions (like putting the server in 2D) really effects anything.

It's all a guessing game, really.

[Sinu]

Just played on the server today --

Ground vehicles were warping
Every aircraft in the air was warping, lagging
Flares and weapons were lagging

Not a good day.

No idea what changed. There wasn't any lag at all the last time I flew.

[CobaltUK]

Mission 30 mins ago was badly lagging/warping. Server in 2D but not sure if its been like that a few days now or freshly into 2D.

[=VG= SemlerPDX]

This morning it was crashed. Restarted to the last autosave, but I didn't do anything to force it back to the 2D map. Gonna wait until one of you is online later so we can try to do so and see if it continues to run well as it has this past week. If it works, I will add that step to the restart procedure (launching then returning to the 2D map).

What we did this last time was simply ignore the anomalous event of the server forcing itself back to the 2D around last weekend. Time to see if we can make it happen at will with same results.
!hi

[Sinu]

Flew today and lag seemed okay (there was just 1 or 2 aircraft spinning every now and then, nothing major)

Certainly seems like the difficulty settings we are running have a significant trade off in realistic weapon performance on ground targets.

I did a visual BDA of an armored battalion (18 T-62s with support vehicles) after 4 JSOWs and 4 CBU-87s and saw all vehicles on fire with the TGP.

After exiting the mission there were 4 "confirmed" A-G kills. Many, many vehicles "damaged." I guess this is just how the sim increases vehicle density -- by making them way, way harder to kill?

[=VG= SemlerPDX]

There is no part of the difficulty sliders that allow for damage modifiers, it's not part of that aspect that we can change - all that is hard coded. Unless this is part of that "morale" thing that ground troops get that you mentioned and unless those OPFOR troops were buffed somehow from said "morale", I cannot imagine why you experienced such lackluster performance with those two cluster bombs...
!dntknw

[Brain]

Has been a reoccurring problem for months. A GBU-12 directly hitting a soft target MIGHT result in 1 kill but everything hit by the splash radius will show up as "damaged" in the debriefing. Doesn't seem to happen in offline flights.

If the server has a 12mbit connection, shouldn't the BW be more around 11000 (1024 per player + 10% overhead)? I know it's just a maximum cap and not necessarily used fully, but if the server tries to send 25mbit worth of data and only 12mbit make it through ... you get the idea.

Have noted popping in ground units again. When switching A2G weapons, looking at the TGP all targets vanish and reappear a few seconds later. Seems to be worse with multiple battalions in line of sight.
Is BMS resetting the bubble in this event, as stupid as it might seem? Is the delayed popup a sign of network data not making it through when refreshing the bubble?

Don't think we've made it past 8 players at once so we could give lowering the host BW a shot. Since the server is physical unable to go beyond 12mbit it shouln't hurt. If it does ... BMS is strange.


Campaign is going well ... for BLUFOR. The ground war was pretty good, took several days to push DPRK back over the DMZ. That's what I expected from removing the Osan Hogs.
Air war is a game of chance. There are some OPFOR flights going on each day but according to the force levels there aren't that many planes left. It's certainly not enough to fly CAP over the FLOT, allowing BLUFOR to conduct missions unopposed.

[Sinu]

Flew today. AI lag was nonexistetent. Ground units appeared fine and there was no disapearing airbases or anything like that.

The difficulty sliders we have selected seem to give the DRPK nigh-infinite ground unit density. I am still experiencing difficulty in getting actual "kills" versus ground units after missions where I obviously had good effect on target.

NATO forces are unable to penetrate the FLOT for whatever reason - presumably due to the ridiculous amount of DPRK forces.

I will do more investigation as to what that Ground Forces slider actually does - aside from adding more GUs to DPRK.

[Brain]

I second that.
Disappearing ground units on TGP is a rare case when switching a2g weapons. If you take your time when doing attack runs it shouldn't be a big deal.

BLUFOR is probably stuck behind all those destroyed bridges. It is a low priority target for AI but since they took care of everything else it's what gets fragged.
Clearing the area of OPFOR troops might allow capturing and repairing certain bridges. Question is if AI commander will give the order to an engineer or if they are all busy elsewhere.

Found the campaign events, they are stored in the saveX.tri files. Have to dig further as this doesn't appear to be the place handling reinforcements. How did you spawn the Vinson on day 1 again? Now, back to my shovel, lol

[Sinu]

I spawned the Vinson using Mission Commander and via this man's tutorial: https://www.youtube.com/watch?v=l302aJ7TbsM

[Sinu]

Tried to connect to the server today and it seems like it is crashed out.

[=VG= SemlerPDX]

Restarted to last autosave. Sorry for the slow response time.

[Brain]

My ISP went full retard, so instead of wasting my time on the net I spent some of it on this save. It's nothing special, just some proof of concept crap I've thrown together. Included 2 additional saves to spare you waiting for things to happen at 64x. Main focus in this 1st iteration was seeing how things influence the ATO, so I didn't advance too far yet. It's based on the day1 save VG is currently using.https://www.dropbox.com/s/y20ndwyld84cdhq/VG%20prio%20POC.7z?dl=0


For the most part I adjusted BLUFOR prios willy-nilly in mission commander just to see how things turn out. Here a quick summary of changes:
reduced MISSION prio for Strike, SEAD, OCA, Sweep to make BLUFOR less aggressive against high value targets. UNIT and OBJECTIVE prio had to be adjusted accordingly or the AI would still focus on these (Setting UNIT prio for aircraft was done in game, since MC doesn't appear to have an option for that, which seems odd). Maxed out MISSION prio for Escort & all kinds of CAP to defend friendly air assets and airspace.
All Vipers had their speciality set to A2A, Strike Eagles set to General. Reactivated Hogs to help dealing with DPRK ground forces.
All already scheduled SEAD/DEAD/OCA packages have been cancelled.

OPFOR has been adjusted in a similar manner but are slightly more aggressive to make human DCA viable.

I'll just let the save speak for itself for those interested (pay attention to the kinds of mission in the ATO!!!). It's very flawed but the whole point is showing how the fragging AI can be influenced to make players have more of an impact. Again: I think that's what it should be about, not cleaning up behind the AI.


Another thing:
"IF_BORDOM_HOURS" in the campaign's .tri file will end the campaign early based on a timer. Basically if no other event happens (mostly capturing objectives from what I understand) in the given time frame the campaign just ends.