Anonymous | Login | Signup for a new account | 2021-01-15 21:27 PST | ![]() |
Main | My View | View Issues | Change Log | Roadmap | Summary | My Account |
View Issue Details [ Jump to Notes ] | [ Issue History ] [ Print ] | |||||||||
ID | Project | Category | View Status | Date Submitted | Last Update | |||||
0008177 | opensim | [REGION] OpenSim Core | public | 2017-05-23 13:11 | 2017-07-01 17:38 | |||||
Reporter | BillBlight | |||||||||
Assigned To | ||||||||||
Priority | normal | Severity | major | Reproducibility | always | |||||
Status | new | Resolution | open | |||||||
Platform | LINUX | Operating System | UBUNTU x64 | Operating System Version | 14/15/16 | |||||
Product Version | master (dev code) | |||||||||
Target Version | Fixed in Version | |||||||||
Summary | 0008177: Extreme memory usage as compared to 0.8.2.1 and 0.8.3 Post-Fixes | |||||||||
Description | I have tested this by loading up a region in 8x using bullet physics. When loading up the exact same region in the .9 master code the same region uses over double the amount of RAM regardless of using bullet or ubODE physics. regions are 4x4 var. on 8x region using approx 2700mb of ram on 9x region using approx 5600mb of ram This is not a leak , this is right after start up and logins enable. | |||||||||
Steps To Reproduce | I have tested this by loading up a region in 8x using bullet physics. When loading up the exact same region in the .9 master code the same region uses over double the amount of RAM regardless of using bullet or ubODE physics. regions are 4x4 var. on 8x region using approx 2700mb of ram on 9x region using approx 5600mb of ram This is not a leak , this is right after start up and logins enable. | |||||||||
Additional Information | I have not noticed this previously , it could have been there for a while . | |||||||||
Tags | No tags attached. | |||||||||
Git Revision or version number | ||||||||||
Run Mode | Grid (1 Region per Sim) , Grid (Multiple Regions per Sim) | |||||||||
Physics Engine | BulletSim, ubODE | |||||||||
Script Engine | ||||||||||
Environment | Mono / Linux64 | |||||||||
Mono Version | Other | |||||||||
Viewer | ANY | |||||||||
Attached Files | ||||||||||
![]() |
|
(0031923) kcozens (administrator) 2017-05-23 13:24 |
Which version of mono are you using? |
(0031924) BillBlight (developer) 2017-05-23 14:07 |
I noticed with 4.2 , then with 4.8 and now my servers evidently auto updated to 5 .. |
(0031925) BillBlight (developer) 2017-05-23 15:09 |
I am going to test on my backup sever on mono 3.2.8 , see if that does the same thing. .. |
(0031927) BillBlight (developer) 2017-05-23 16:26 edited on: 2017-05-23 16:33 |
Now tested on mono 3.2.8, and above this is just a random sampling, but same oar loaded .. (this is right after startup on .9x and a few hours later on 8.x) On 9.x Process memory: Physical 4469 MB On 8.x Process memory : 2430 MB I have tried on the .9x with both bullet and ubODE makes no noticeable difference. And just for informational purposes, on either system does it appear that it is using much cpu, which is pretty normal for opensim, but on the .9x it is about 50% higher, which is still not that much. Oh and just to verify I am testing with one region per instance. As stated before this is a 4x4 var, with about 17 thousand objects .. On the .9x it will spike sometimes and when it does you can hardly move at all. |
(0031932) BillBlight (developer) 2017-05-23 18:16 |
ok both sims tested on 5 different versions of mono , difference between versions is negligible. Sims have been up for an hour, started at the same time, with clean reboot of the server .. Show stats from each. .8x MEMORY STATISTICS Heap allocated to OpenSim : 672 MB Last heap allocation rate : 0.027 MB/s Average heap allocation rate: 0.13 MB/s Process memory : 2377 MBMEMORY STATISTICS .9x MEMORY STATISTICS Heap allocated to OpenSim : 3152 MB Last heap allocation rate : 0.145 MB/s Average heap allocation rate: 0.038 MB/s Process memory: Physical 5469 MB Paged 0 MB Virtual 8798 MB Peak process memory: Physical 5470 MB Paged 0 MB Virtual 8798 MB I think this is a little excessive .. |
(0031933) UbitUmarov (administrator) 2017-05-23 18:27 |
just a bit.. and total opposite to what I seen in win7 now |
(0031942) BillBlight (developer) 2017-05-23 18:51 |
Yes and as I said those sims are exact copies of each other , with NO avatars or NPC's on them .. A lot of scripts and objects , but no agents .. |
(0031965) UbitUmarov (administrator) 2017-05-25 22:46 |
testing with mono 5.0 with Bill starting region with the opensim-ode.sh as -- ulimit -s 262144 <-- for ode and ubdoe MONO_GC_PARAMS="minor=split,promotion-age=14" mono --desktop OpenSim.exe --- does have big impact on mono memory use bill is testing other option, he may report |
(0031966) BillBlight (developer) 2017-05-25 22:53 |
If you guys have not figured it out yet I am Bill Blight .. I am testing with some overly complicated settings that seem to have a better result on a region with 33k prims and 2700 scripts, but will test longer term and report back .. The settings I am using are .. ulimit -s 1048576 export MONO_GC_PARAMS="nursery-size=16m,evacuation-threshold=90,soft-heap-limit=1280m,save-target-ratio=0.1,default-allowance-ratio=1.0,promotion-age=14,minor=split,major=marksweep,no-lazy-sweep,alloc-ratio=50" export MONO_GC_DEBUG="xdomain-checks,clear-at-gc" export MONO_ENV_OPTIONS="--desktop" export MONO_THREADS_PER_CPU=250 These do appear to slow down region startup for some reason , but after the region is up it seems to run just fine, with some better memory management. |
(0031970) UbitUmarov (administrator) 2017-05-26 15:27 |
Compile as Release does help a bit on all aspects Just if something goes wrong, there is less information to debug it |
(0031972) BillBlight (developer) 2017-05-27 19:39 |
Well let's see how can I put this, umm , still crap .. 33k objects 2700 scripts Region on .8x running for 24 hours MEMORY STATISTICS Heap allocated to OpenSim : 751 MB Last heap allocation rate : 2.048 MB/s Average heap allocation rate: 0.914 MB/s Process memory : 3514 MB Runs fine , no lag what so ever .. .9x up for just over an hour .. MEMORY STATISTICS Heap allocated to OpenSim : 1399 MB Last heap allocation rate : 0 MB/s Average heap allocation rate: 0 MB/s Process memory: Physical 4452 MB Paged 0 MB Virtual 7780 MB Peak process memory: Physical 4452 MB Paged 0 MB Virtual 7781 MB Can hardly move on the region .. |
(0031973) UbitUmarov (administrator) 2017-05-27 20:05 |
crude win test,just after logins enable, no avatar log in region 0.82, bullet 2x size MEMORY STATISTICS Heap allocated to OpenSim : 54 MB Last heap allocation rate : 0.016 MB/s Average heap allocation rate: 0.015 MB/s Process memory : 376 MB 0.91 region 3x + 1x on same instance. 1x has same prims as above, 3x has a few more MEMORY STATISTICS Heap allocated to OpenSim : 58 MB Last heap allocation rate : 0.104 MB/s Average heap allocation rate: 0.104 MB/s Process memory: Physical 171 MB Paged 152 MB Virtual 740 MB Peak process memory: Physical 267 MB Paged 249 MB Virtual 763 MB |
(0031974) BillBlight (developer) 2017-05-27 20:07 |
yeah yeah yeah .... Win win win .. :P Who's gonna pay for my windows licenses? But I guess the issue is .. .8x does not have this memory issue as .9x does on linux ... What is the root cause of .9x using more memory on linux? |
(0031975) UbitUmarov (administrator) 2017-05-27 20:11 |
driving same car on identical terrain, both regions on bullet 0.91 150ms avg physics time 0.82 320ms avg physics time ( but ok 0.82 stats was bad ) with the avatar log and a few laps with car 0.82 MEMORY STATISTICS Heap allocated to OpenSim : 91 MB Last heap allocation rate : 0.034 MB/s Average heap allocation rate: 0.05 MB/s Process memory : 405 MB 0.91 MEMORY STATISTICS Heap allocated to OpenSim : 64 MB Last heap allocation rate : 0.111 MB/s Average heap allocation rate: 0.104 MB/s Process memory: Physical 187 MB Paged 163 MB Virtual 746 MB Peak process memory: Physical 267 MB Paged 249 MB Virtual 771 MB |
(0031976) UbitUmarov (administrator) 2017-05-27 20:23 |
avatar login with clear cache in similar position, full inv fetch objects and textures 0.82 MEMORY STATISTICS Heap allocated to OpenSim : 81 MB Last heap allocation rate : 0.054 MB/s Average heap allocation rate: 4.028 MB/s Process memory : 428 MB 0.91 MEMORY STATISTICS Heap allocated to OpenSim : 68 MB Last heap allocation rate : 3.524 MB/s Average heap allocation rate: 3.318 MB/s Process memory: Physical 210 MB Paged 219 MB Virtual 756 MB Peak process memory: Physical 306 MB Paged 288 MB Virtual 779 MB after avatar log out 0.82 MEMORY STATISTICS Heap allocated to OpenSim : 96 MB Last heap allocation rate : 0.009 MB/s Average heap allocation rate: 0.045 MB/s Process memory : 429 MB 0.91 MEMORY STATISTICS Heap allocated to OpenSim : 64 MB Last heap allocation rate : 0.125 MB/s Average heap allocation rate: 0.107 MB/s Process memory: Physical 195 MB Paged 203 MB Virtual 746 MB Peak process memory: Physical 306 MB Paged 288 MB Virtual 779 MB so not sure what to say :( |
(0031977) UbitUmarov (administrator) 2017-05-27 20:28 |
after avatar return ( kicks gc sometimes) 0.82 MEMORY STATISTICS Heap allocated to OpenSim : 101 MB Last heap allocation rate : 0.048 MB/s Average heap allocation rate: 0.78 MB/s Process memory : 433 MB 0.91 MEMORY STATISTICS Heap allocated to OpenSim : 61 MB Last heap allocation rate : 0.098 MB/s Average heap allocation rate: 0.108 MB/s Process memory: Physical 183 MB Paged 164 MB Virtual 745 MB Peak process memory: Physical 306 MB Paged 288 MB Virtual 779 MB |
(0031978) BillBlight (developer) 2017-05-27 20:28 |
So essentially we are just screwed if we want to continue running on linux , we are stuck with .8x ... |
(0031979) UbitUmarov (administrator) 2017-05-27 20:30 |
versions where urrent simulator: OpenSim 0.9.1.0 Dev (Win/.NET) ( master ) Previous simulator: OpenSim 0.8.2.1 Post_Fixes (Win/.NET)] |
(0031980) UbitUmarov (administrator) 2017-05-27 21:20 |
your VM, mono5, 1x region, identical oar. ubOde started with ulimit -s 262144 <-- for ode and ubdoe MONO_GC_PARAMS="minor=split,promotion-age=14" mono --desktop OpenSim.exe MEMORY STATISTICS Heap allocated to OpenSim : 40 MB Last heap allocation rate : 0 MB/s Average heap allocation rate: 0.002 MB/s Process memory: Physical 189 MB Paged 0 MB Virtual 4710 MB Peak process memory: Physical 273 MB Paged 0 MB Virtual 4788 MB same win 0.91 as before but also on ubOde now MEMORY STATISTICS Heap allocated to OpenSim : 52 MB Last heap allocation rate : 0.094 MB/s Average heap allocation rate: 0.102 MB/s Process memory: Physical 169 MB Paged 143 MB Virtual 769 MB Peak process memory: Physical 262 MB Paged 239 MB Virtual 786 MB |
(0031981) UbitUmarov (administrator) 2017-05-27 22:22 |
sorry about car timings where us not ms |
(0031982) UbitUmarov (administrator) 2017-05-27 23:30 |
before going to sleep did some crossings and little prims rez on 0.91 regions above, both win and mono on ubode, stats as I type Linux VM mono 5 EMORY STATISTICS Heap allocated to OpenSim : 46 MB Last heap allocation rate : 0 MB/s Average heap allocation rate: 0.004 MB/s Process memory: Physical 250 MB Paged 0 MB Virtual 4757 MB Peak process memory: Physical 300 MB Paged 0 MB Virtual 4808 MB win 91 2 regions as above EMORY STATISTICS eap allocated to OpenSim : 50 MB ast heap allocation rate : 0.113 MB/s verage heap allocation rate: 0.112 MB/s rocess memory: Physical 185 MB Paged 160 MB Virtual 779 MB eak process memory: Physical 346 MB Paged 320 MB Virtual 920 MB win 0.82 just being idle since last stats repo above MEMORY STATISTICS Heap allocated to OpenSim : 272 MB Last heap allocation rate : 0.015 MB/s Average heap allocation rate: 0.045 MB/s Process memory : 521 MB so 0.82 is great? well maybe.. this is just idle test |
![]() |
|||
Date Modified | Username | Field | Change |
2017-05-23 13:11 | BillBlight | New Issue | |
2017-05-23 13:21 | BillBlight | Category | [MISC] Compiling / Building => [REGION] OpenSim Core |
2017-05-23 13:24 | kcozens | Note Added: 0031923 | |
2017-05-23 14:07 | BillBlight | Note Added: 0031924 | |
2017-05-23 15:09 | BillBlight | Note Added: 0031925 | |
2017-05-23 16:26 | BillBlight | Note Added: 0031927 | |
2017-05-23 16:29 | BillBlight | Note Edited: 0031927 | View Revisions |
2017-05-23 16:31 | BillBlight | Note Edited: 0031927 | View Revisions |
2017-05-23 16:33 | BillBlight | Note Edited: 0031927 | View Revisions |
2017-05-23 18:16 | BillBlight | Note Added: 0031932 | |
2017-05-23 18:27 | UbitUmarov | Note Added: 0031933 | |
2017-05-23 18:51 | BillBlight | Note Added: 0031942 | |
2017-05-25 22:46 | UbitUmarov | Note Added: 0031965 | |
2017-05-25 22:53 | BillBlight | Note Added: 0031966 | |
2017-05-26 15:27 | UbitUmarov | Note Added: 0031970 | |
2017-05-27 19:39 | BillBlight | Note Added: 0031972 | |
2017-05-27 20:05 | UbitUmarov | Note Added: 0031973 | |
2017-05-27 20:07 | BillBlight | Note Added: 0031974 | |
2017-05-27 20:11 | UbitUmarov | Note Added: 0031975 | |
2017-05-27 20:23 | UbitUmarov | Note Added: 0031976 | |
2017-05-27 20:28 | UbitUmarov | Note Added: 0031977 | |
2017-05-27 20:28 | BillBlight | Note Added: 0031978 | |
2017-05-27 20:30 | UbitUmarov | Note Added: 0031979 | |
2017-05-27 21:20 | UbitUmarov | Note Added: 0031980 | |
2017-05-27 22:22 | UbitUmarov | Note Added: 0031981 | |
2017-05-27 23:30 | UbitUmarov | Note Added: 0031982 |
Copyright © 2000 - 2012 MantisBT Group |