Jump to content

SAS_Storebror

Members
  • Content Count

    1196
  • Joined

  • Last visited

Everything posted by SAS_Storebror

  1. Thanks for confirming the issue Volker. I don't want to push the devs, just had the feeling that this issue might go un-noticed because the only reply to it before was along the lines of "oh, cars and tanks work just fine" so if the devs weren't aware of it, they might have thought that it was just me. Cheers! Mike
  2. Brief description: Dedicated Server CPU usage goes through the roof, especially with AI planes on a mission Detailed description, conditions: On Version 2.012d, for instance a mission which used 10-15 AI fighters and 4 AI bombers in formation on each side (18-23 AI planes in total), together with lots of ground action, caused a CPU load of 10-15% (measured on single core). Resulting SPS was 50, Tick Delay close to 0. On Version 3.001, the same mission causes 15% CPU before planes spawn already, and as soon as the first AI planes spawn, the CPU usage reaches 75-80%, resulting SPS is still 50, but Tick Delay is 15 and the server is at the edge of stalling. Using a test mission (see attachment) with just 8 AI bombers (4 on each side, in formation), we get these results: Version 2.012d CPU load 1-3%, Tick Delay close to 0. Version 3.001 CPU load 25-30%, Tick Delay 7-8. That's just 8 AI planes and the server is under heavy stress already! The test mission can be found attached to this post. Additional assets (videos, screenshots, logs): Test mission can be found attached to this post. Your PC config data (OS, drivers, specific software): Windows Server 2016, Intel Core i7-4770 @3.8GHz, 32 GB RAM, 2x2TB HDD (Raid 1), 1GBit/s non-clocked. CPU Load Test 3.001 v1.zip
  3. I take the liberty to *bump* this thread because the issue is definitely there and it's a severe issue to Multiplayer Server hosting. Unfortunately I have the feeling that no one cares. Cheers! Mike
  4. Will do and report back, takes a few hours - RL stuff you know If it's definitely sure that the "OnKilled" Event will always get fired when a plane dies, then this would basically be good enough. I'm afraid though that situations may exist where e.g. a pilot gets shot, but the plane keeps flying for hours and no "OnKilled" event gets fired. Or where a plane gets shot to pieces, so that the game counts it as "destroyed" and the pilot doesn't engage anymore, yet the plane keeps flying for hours again and no "OnKilled" event gets fired. That'd be a worst case scenario. Disabling Spawn at "OnCriticalDamage" ("pilot ejected" is no availabled event) seems unnecessary cause in that case I could just remove those event links completely to the same effect. I see the intention. I did use a counter already, even though as I'm gonna respawn the planes individually too, this might not be necessary. In your scheme you are monitoring the kill and destroyed events. What happens in case the pilot gets killed and the plane keeps flying? What happens when the plane belly lands and keeps lying on the ground? Does any of the two events get fired in those cases? And furthermore, what happens when on one plane both the "kill" and "destroyed" event get fired, e.g. 3 seconds after each other? In my book, this would trigger the counter two times, for one plane. That's exactly what I'd like to avoid. My scheme currently is like this: Plane --> Event OnPilotKilled/OnPlaneCrashed/OnPlaneDestroyed/OnKilled --> [Filter out multiple Events] -> wait 30 seconds -> Delete -> wait 2 seconds -> Respawn Plane The critical part is filtering out multiple events. For instance I've tried using a self-resetting "1" counter which I've deactivated right after firing it, and tried to reactivate the counter from an "OnPlaneSpawned" event of the plane. Doesn't work. Once the counter got deactivated, the reactivation doesn't seem to work and the counter will never fire again. And without that counter deactivation, I get the "multiple respawn issue" as expected. A plane gets the pilot killed, crashes and finally gets destroyed and therefore killed -> All 4 events fire and as a result, for that single plane downed, 4 new planes spawn. That's illogical in itself. How can a deactivated Spawner spawn a plane that fires an "OnSpawned" event to reactivate itself? Cheers! Mike
  5. Hi folks, I'd like to spawn an AI plane on my mission and respawn it once it has been "killed", which means it's "out of business" in whatever way. For instance the pilot could be dead, have bailed out, or the plane could have crash landed or could be completely destroyed. The problem I have is: I can point all these events like "OnPlaneDestroyed" to an MCU, and let that MCU trigger a respawn, however by doing so I sometimes seem to get multiple respawns, i.e. first the AI pilot gets shot, then the plane crash lands and finally it hits a tree - that seems to trigger all three events, hence 3 times respawn. So I thought I'd somehow "buffer" these events, but so far I'm unsuccessful. If I put a counter inbetween and disable it for a certain amount of time, it seems not to be able to re-enable it in a way that it would trigger again later on. If there'd be one singe event that will always fire on a dead plane, and that will reliably fire only once, then this would be great, but so far I fail to find one. On the other hand, if there'd be a possibility to somehow filter out these multiple events, that'd be cool as well, but I'm running out of ideas how to achieve that. Any clue from you ME experts? Cheers! Mike
  6. If it's really wanted to see mods (wanted by the devs I mean), it would surely help if there'd be some kind of documentation, or if that's too much, some hints regarding the points modders can hook in on the game in order to change things. A general outline of which parts of the game are supposed to be moddable would be welcome too. Cheers! Mike
  7. Thanks for the updates @=FB=Vaal, all is looking good now. @WW_SittingDuck: You only need the last objects.zip or, alternatively, download the full installation file again and run the updater. Cheers! Mike
  8. Since this takes place on file system level, it will definitely work. Cheers! Mike
  9. I bet that this is strongly related to the excessive CPU load caused by the 3.001 update as reported here... Cheers! Mike
  10. Has been reported here... ...already. Not aimed at you, but just when I tried to link that thread/post, I noticed to what extend this board section is flooded with things that don't belong here, things that have been reported a hundred times before, things that have been solved in dozens of other similar threads already etc. pp. Would be great if there could be moderators who keep this board section clean, because as I understand, it should be some kind of support for the Devs in order to keep track of real existing bugs, instead of being just another lounge where people chatter all across. Cheers! Mike
  11. Does this statement cover DServer performance issues as well? Cheers! Mike
  12. Eww... Just reinstalled the whole game from scratch, 10.24GB of 8.09GB (sic!), just to see that the empty .list file would be created again. After some experimenting with the mission, this is what got me back the .list file again: Use "Save As..." to save the mission to a different folder with different name. Use "Save As..." to save it back to original folder/name. It's weird, but that's what did the trick this time. Cheers! Mike
  13. I must say that up to 2.012 the CPU consumption of AI bombers was significant compared to any other AI plane, but nothing to be worried about too much as long as you own a reasonably powerful server to run the missions on (which we do). With 3.001 though, AI bombers are a no-go and AI fighter planes basically too, only very small numbers will work and only if there's not too many human players around at the same time too. Definitely something to look into for the devs. If I understand correctly, other servers are suffering from overloads since this update too, so this is a rather important issue. Cheers! Mike
  14. No sir. As soon as locked loadouts and/or weapon mods are activated on the server, we see fancy issues with locked loadouts all over the place, for both "old" and "new" planes in 3.001, where everything was working fine and dandy in 2.012 before. This isn't even consistent among subsequent sorties. I haven't reported it so far 'cause I've yet to figure the precise pattern of this issue, but it's definitely not related just to the "new" planes. Cheers! Mike
  15. I've been talking about AI planes. The thread I've linked contains a very simple test mission with just two flights of 4 bombers each on it. If you see something that's "not correctly set up" on it, please be my guest and let me know. Once again: Prior to 3.001 update, on version 2.012, this test mission and it's much bigger "real" brother were running perfectly fine with no CPU load issues at all. Cheers! Mike
  16. Working - yes. Performing well - unfortunately no. As reported here... ...we see a massive increase in CPU usage with DServer which barely makes it possible to serve any mission with any AI planes on it at all. Mind you, these missions were performing extraordinary well with 2.012 before, causing less than 20% CPU usage (peak value). Cheers! Mike
  17. Yes of course, you must not click "save" or you will have to edit back the difficulty settings again. Cheers! Mike
  18. Dear Devs, Could you please have a look at the CPU load caused on dedicated servers after updating to 3.001? On our test server we've been running a "work in progress" mission for months. There's quite some AI on it, like 20-30 cars and tanks, about 10-15 fighters in the air and lately we have added two flights of 4 bombers on each side. All of this together, on version 2.012 caused a CPU load of 10-15% on our dedicated server (measured on a single core, i7-4770 @3.8 GHz, 32GB RAM, Windows Server 2016). After updating to 3.001 we have found that with nothing but AAA active, the CPU load is at ~15% already. When the first fighters and cars/tanks spawn, it goes up to 35-50% and once the bombers spawn we are at 75-80% CPU load, which means we still have an SPS of 50.0 +/- 0.3, but with a tick delay in the range of 15 the game is right on the edge of not being playable anymore. Today I've made a very simple test mission where there is nothing but airfields to spawn on, and one of our 4-bomber-flights on each side. The bombers will get activated 1 minute after first player spawns. You will see that exactly at that time, the CPU load will increase significantly, beyond anything we knew with the old version 2.012. In our case, CPU load of this very simple mission with just 8 AI planes prowling the skies will go up from 0-2% to 25-30% right at the moment when the AI planes are activated, and it stays like that. Tick delay will be ~7-8. For comparison, a similar testing on version 2.012 "back in the days" showed something like 1-3% CPU usage and no change in tick delay when 8 AI bombers spawned. This issue makes it more or less impossible to serve AI crowded missions on dedicated servers at the moment, and I have a feeling like it could be related to the FPS issues reported for the campaign system too (where AI has to be handled on client side then). Test mission can be found attached to this post. Cheers! Mike CPU Load Test 3.001 v1.zip
  19. Done, please see attachment for the logs of the mission where stats.cmd bails out upon. Cheers! Mike stats_issue_2018-03-14.zip
  20. Yes it's basically "expert" setting, except for the Nav Icons. As it stands, you have to open the SDS file in an editor for now and apply your settings manually. Cheers! Mike
  21. Not only that. Once you choose "custom" difficulty settings, no matter what you choose or set, the setting will always be: objectIcons = false navigationIcons = true aimingHelp = false courseWeaponsAimingHelp = false padlock = false simpleDevices = false allowSpectator = false easyFlight = false autoCoordination = false autoThrottle = false autoPilot = false autoThrottleLimit = false autoMix = false autoRadiator = false noMoment = false noWind = false noMisfire = false noBreak = false invulnerability = false unlimitFuel = false unlimitAmmo = false engineNoStop = false hotEngine = true Cheers! Mike
  22. Stats parser (stats.cmd) stopped working on first 3.001 mission to parse. Log: [2018.03.14 18:59:02] unexpected error Traceback (most recent call last): File "C:\Python3\virtualenvs\il2stat\il2_stats\src\stats\management\commands\stats_whore.py", line 16, in handle stats_whore.main() File "C:\Python3\virtualenvs\il2stat\il2_stats\src\stats\stats_whore.py", line 69, in main online_timestamp = update_online(m_report_files=m_report_files, online_timestamp=online_timestamp) File "C:\Python3\virtualenvs\il2stat\il2_stats\src\stats\online.py", line 24, in update_online data = parse_mission_log_line.parse(line) File "C:\Python3\virtualenvs\il2stat\il2_stats\src\mission_report\parse_mission_log_line.py", line 231, in parse data = atype_handlers[atype_id].match(line.strip()).groupdict() AttributeError: 'NoneType' object has no attribute 'groupdict' [2018.03.14 18:59:02] Lock 3152032611184 released on C:\Python3\virtualenvs\il2stat\il2_stats\file.lock Traceback (most recent call last): File "manage.py", line 7, in <module> execute_from_command_line(sys.argv) File "C:\Python3\virtualenvs\il2stat\il2_stats\.venv\lib\site-packages\django\core\management\__init__.py", line 364, in execute_from_command_line utility.execute() File "C:\Python3\virtualenvs\il2stat\il2_stats\.venv\lib\site-packages\django\core\management\__init__.py", line 356, in execute self.fetch_command(subcommand).run_from_argv(self.argv) File "C:\Python3\virtualenvs\il2stat\il2_stats\.venv\lib\site-packages\django\core\management\base.py", line 283, in run_from_argv self.execute(*args, **cmd_options) File "C:\Python3\virtualenvs\il2stat\il2_stats\.venv\lib\site-packages\django\core\management\base.py", line 330, in execute output = self.handle(*args, **options) File "C:\Python3\virtualenvs\il2stat\il2_stats\src\stats\management\commands\stats_whore.py", line 16, in handle stats_whore.main() File "C:\Python3\virtualenvs\il2stat\il2_stats\src\stats\stats_whore.py", line 69, in main online_timestamp = update_online(m_report_files=m_report_files, online_timestamp=online_timestamp) File "C:\Python3\virtualenvs\il2stat\il2_stats\src\stats\online.py", line 24, in update_online data = parse_mission_log_line.parse(line) File "C:\Python3\virtualenvs\il2stat\il2_stats\src\mission_report\parse_mission_log_line.py", line 231, in parse data = atype_handlers[atype_id].match(line.strip()).groupdict() AttributeError: 'NoneType' object has no attribute 'groupdict' Cheers! Mike
  23. Actually if anything has trees at all anymore, than most likely it is a road. Forests seem to be almost gone completely. No idea whether this is historically correct, so I'm just reporting the difference I see. I know this might not be the right place, but any other threads seem to remain unnoticed at the moment too. The mission's CPU load is really something that puzzles me as well. We've seen something like "exponentially raising CPU usage" with the number of AI flights before, e.g. with 2x4 bombers in the air, the CPU load was ~15%, but when 6x4 bombers were there, it was 100%. But now, with 2x4 bombers on the very same mission, the CPU constantly scores 75-85% load, and that's right on the edge of becoming unplayable - on a rather "empty" mission. For today I think I'll leave it "as is" and wait for the more experienced guys to chip in. Maybe tomorrow we know more. This patch, as it looks to me at the moment, has lots of nice new stuff and eye candy for single play, but MP... well... might need some careful double-check. Cheers! Mike
×
×
  • Create New...