Chat log from the meeting on 2015-07-28

From OpenSimulator

Jump to: navigation, search
[11:08] Aine.Caoimhe @refugegrid.com:8002: a little tricky having a dev meeting with no devs in attendance
[11:09] Richardus.Raymaker @hypergrid.org:8002: hee vivian. finding my pins back ? can be exiting if you crawl around the bugs :P
[11:10] Aine.Caoimhe @refugegrid.com:8002: @Richardus....do any of the regions you're hosting have a Shoutcast board on them?
[11:10] Aine.Caoimhe @refugegrid.com:8002: (re http://opensimulator.org/mantis/view.php?id=7647)
[11:11] Richardus.Raymaker @hypergrid.org:8002: Hi aine. no never used a shoutcast board. i just knew this word since this year
[11:11] Richardus.Raymaker @hypergrid.org:8002: offcorse i know what shoutcast is
[11:12] Richardus.Raymaker @hypergrid.org:8002: or at least i neve rplaced one
[11:12] Andrew Hellershanks: nebadon is on his way
[11:12] Aine.Caoimhe @refugegrid.com:8002: I was just curious if you'd seen that memory issue
[11:13] Snakedance.Moonwing @snaketopia.darktech.org:8002: hello :)
[11:13] Nebs Metal Bar Stool v1.5 (w/sit & launch): Hello Nebadon Izumi, enjoy your sit..
[11:13] Nebadon Izumi kommt in Chat-Reichweite (8.35 m).
[11:13] Nebadon Izumi: hello
[11:13] Richardus.Raymaker @hypergrid.org:8002: but my system, seems stable
[11:13] Nebadon Izumi: sorry I am late everyone
[11:13] Aine.Caoimhe @refugegrid.com:8002: I've had one sitting in my test region for the last couple days and it's doing nothing at all to eat memory and when I looked through the (billion!) scripts in it I couldn't see anything that looked horid
[11:13] Richardus.Raymaker @hypergrid.org:8002: hii snake
[11:13] Snakedance.Moonwing @snaketopia.darktech.org:8002: i;ve come to snoop :)
[11:13] Vivian Klees flüstert: better late than never :)
[11:13] Snakedance.Moonwing @snaketopia.darktech.org:8002 kommt in Chat-Reichweite (9.39 m).
[11:13] Aine.Caoimhe @refugegrid.com:8002: but both Ken and Seth have encountered sudden memory bloating from it
[11:13] Andrew Hellershanks: oh, nebadon is wearing on of his interesting avatars today.
[11:14] Nebadon Izumi: Ken said he figured out why
[11:14] Nebadon Izumi: it was a scrpt
[11:14] Aine.Caoimhe @refugegrid.com:8002: oh really:?
[11:14] Nebadon Izumi: which is what I figured it was
[11:14] Nebadon Izumi: ya
[11:14] Nebadon Izumi: he posted IAR of the object
[11:14] Aine.Caoimhe @refugegrid.com:8002: do you know which one?
[11:14] Nebadon Izumi: its a Shoutcast Board
[11:14] Nebadon Izumi: its on the Mantis
[11:14] Aine.Caoimhe @refugegrid.com:8002: yes
[11:14] Andrew Hellershanks: Aine, I am a dev. Just not a very active one in the main parts of OS.
[11:14] Nebadon Izumi: he removed object
[11:14] Nebadon Izumi: and memory bloat stopped
[11:14] Aine.Caoimhe @refugegrid.com:8002: I looked through the scripts there and couldn't see anything suspicioius
[11:14] Nebadon Izumi: so we have a start point
[11:14] Nebadon Izumi: which is good
[11:14] Aine.Caoimhe @refugegrid.com:8002: oh, sorry Andrew
[11:14] Nebadon Izumi: ya it could be anything honestly
[11:15] Nebadon Izumi: Xengine needs a lot of love
[11:15] Nebadon Izumi: no doubt its the source of tons of issues
[11:15] Aine.Caoimhe @refugegrid.com:8002: it's all keyframe type texture calls....no dynamic draws
[11:15] Andrew Hellershanks: Aine, np. :) I'm also a core member.
[11:15] Aine.Caoimhe @refugegrid.com:8002: so all the function calls are pretty run of the mill
[11:15] Richardus.Raymaker @hypergrid.org:8002: is the shoutcast baord using dynamic textures ?
[11:15] Aine.Caoimhe @refugegrid.com:8002: it would be interesting to track down what's causing it though
[11:15] Aine.Caoimhe @refugegrid.com:8002: no
[11:16] Aine.Caoimhe @refugegrid.com:8002: it's all LSL
[11:16] Aine.Caoimhe @refugegrid.com:8002: has about 30+ scripts in it
[11:16] Aine.Caoimhe @refugegrid.com:8002: and a lot of very short duration llSleeps
[11:16] Aine.Caoimhe @refugegrid.com:8002: but no dynamic textures
[11:16] Andrew Hellershanks: ewww
[11:16] Andrew Hellershanks: short duration llSleeps?
[11:16] Aine.Caoimhe @refugegrid.com:8002: it uses a texture with the alphabet and numbers on it
[11:16] Aine.Caoimhe @refugegrid.com:8002: and then uses texture tile settings to display the one it wants
[11:17] Aine.Caoimhe @refugegrid.com:8002: sort of like akeyframing it
[11:17] Aine.Caoimhe @refugegrid.com:8002: yeah....like llSleep (0.2)
[11:17] Snakedance.Moonwing @snaketopia.darktech.org:8002: i hope its alright if i sit here quietly and listen :)
[11:17] Andrew Hellershanks: Snakedance, everyone is welcome
[11:17] Nebadon Izumi: absolutely Snakedance
[11:17] Aine.Caoimhe @refugegrid.com:8002: but as many as 20+ scripts calling that llSleep, frequently
[11:17] Wolf.Freeman @free-world.spdns.de:8002 kommt in Chat-Reichweite (19.21 m).
[11:17] Snakedance.Moonwing @snaketopia.darktech.org:8002: ok :)
[11:17] Andrew Hellershanks: Did I saw ewww already? :)
[11:17] Aine.Caoimhe @refugegrid.com:8002: but I would have expected that to cause thread starvation, not a sudden memory bloat
[11:18] Andrew Hellershanks: Aine, what do you get for the script runtime for that device?
[11:18] Wolf.Freeman @free-world.spdns.de:8002 kommt in Chat-Reichweite (3.19 m).
[11:18] Nebadon Izumi: sleep = bad
[11:18] Aine.Caoimhe @refugegrid.com:8002: negligible
[11:18] Richardus.Raymaker @hypergrid.org:8002: if you can look in the script aine , mabye you can see how it works
[11:18] Nebadon Izumi: I would avoid sleep at all costs
[11:18] Richardus.Raymaker @hypergrid.org:8002: hi neb
[11:18] Andrew Hellershanks: sleep is fine. Short sleep can be bad.
[11:18] Aine.Caoimhe @refugegrid.com:8002: 0.008ms when idle
[11:18] Wolf.Freeman @free-world.spdns.de:8002: hello to all
[11:18] Nebadon Izumi: if you have lots of sleeps expect trouble
[11:18] Aine.Caoimhe @refugegrid.com:8002: I avoid sleeps completely
[11:18] Nebadon Izumi: it will eventually lead to threading issues
[11:18] Nebadon Izumi: light use is ok
[11:19] Andrew Hellershanks: I don't use sleeps less than 1 second if I need them.
[11:19] Nebadon Izumi: just dont get too reliant on it
[11:19] Aine.Caoimhe @refugegrid.com:8002: the only time I ever call it is right after deleting a notecard and about to create a new notecard with the same name
[11:19] Andrew Hellershanks: I had to use one in between two calls that triggered an llHTTPRequest or the second call got lost.
[11:19] Aine.Caoimhe @refugegrid.com:8002: that way the old one is out of inventory before the new one gets created
[11:19] Nebadon Izumi: yea timing can be very problematic
[11:19] Nebadon Izumi: because of how we handle threads
[11:20] Richardus.Raymaker @hypergrid.org:8002: ll sleep are doing complex and change state to create a sleep. then i choice for short llsleep (0.2 or lower)
[11:20] Aine.Caoimhe @refugegrid.com:8002: and I also use it right after an avi first sits to give it time to have the sit status kick in
[11:20] Aine.Caoimhe @refugegrid.com:8002: otherwise I really, really avoid it
[11:21] Aine.Caoimhe @refugegrid.com:8002: anyway....it would be very good to know *what* in that shoutcast board is causing that occasional memory bloat
[11:21] Andrew Hellershanks: some of the other stuff for the device makes me think of xytext
[11:21] Nebadon Izumi: yea I havent had a chance to look at it yet
[11:21] Nebadon Izumi: but im guessing its chock full of timers and sleeps
[11:21] Aine.Caoimhe @refugegrid.com:8002: yes
[11:21] Aine.Caoimhe @refugegrid.com:8002: it is
[11:21] Nebadon Izumi: its the only real way to make those things work
[11:21] Aine.Caoimhe @refugegrid.com:8002: but why would those cause memory bloating?
[11:21] Nebadon Izumi: threads hung
[11:21] Nebadon Izumi: never letting go
[11:22] Andrew Hellershanks: why does it need all those scripts and timers/sleeps?
[11:22] Krysss.Galatea @login.digiworldz.com:8002 kommt in Chat-Reichweite (19.32 m).
[11:22] Aine.Caoimhe @refugegrid.com:8002: oh
[11:22] Aine.Caoimhe @refugegrid.com:8002: do you think it somehow loses a thread and it never gets released
[11:22] Nebadon Izumi: sleeps just kill threading
[11:22] Nebadon Izumi: yes
[11:22] Krysss.Galatea @login.digiworldz.com:8002 kommt in Chat-Reichweite (18.36 m).
[11:22] Dahlia Trimble kommt in Chat-Reichweite (19.27 m).
[11:22] Nebadon Izumi: xengine is not so great at threading
[11:22] Andrew Hellershanks: Is the shoutcast board open source?
[11:22] Nebadon Izumi: the problem is a sleep takes a full thread
[11:22] Aine.Caoimhe @refugegrid.com:8002: and that would be worse on a linux box because it will never get cleaned up
[11:22] Nebadon Izumi: sleeps basically shut xneinge off
[11:22] Aine.Caoimhe @refugegrid.com:8002: yes, it's opensource
[11:23] Aine.Caoimhe @refugegrid.com:8002: I don't ahve one in my suitcase
[11:23] Dahlia Trimble kommt in Chat-Reichweite (19.89 m).
[11:23] Nebadon Izumi: yea 1 sec
[11:23] Aine.Caoimhe @refugegrid.com:8002: but there's an iar of it in the Mantis
[11:23] Nebadon Izumi: he sent me a copy
[11:23] Nebadon Izumi: crap
[11:23] Andrew Hellershanks: Pass me a copy sometime and I'll have a look at it
[11:23] Nebadon Izumi: its no mod no copy
[11:23] Nebadon Izumi: figures 1 sec
[11:23] Dahlia Trimble: hi
[11:23] Andrew Hellershanks: uh oh. SOmething just went poof?
[11:23] Aine.Caoimhe @refugegrid.com:8002: I've had it sitting in my test region for 3 days and it hasn't done anything to the memory
[11:24] Andrew Hellershanks: That was odd. Part of the building vanished for a moment/
[11:24] Nebadon Izumi: there you go ANdrew
[11:24] Aine.Caoimhe @refugegrid.com:8002: but that's running under .NET so maybe Windows is killing those threads
[11:24] Andrew Hellershanks: ty, Nebadon
[11:24] Nebadon Izumi: xengine is no better in Windows than Linux
[11:24] Nebadon Izumi: in terms of threading
[11:24] Nebadon Izumi: sleeps would work better if we did microthreading i suspect
[11:25] Nebadon Izumi: then no single sleep could kill an entire thread
[11:25] Aine.Caoimhe @refugegrid.com:8002: well....mostly trying to think of differences between Ken & Seth vs my setup
[11:25] Aine.Caoimhe @refugegrid.com:8002: and they're both on mono-hosted regions vs my test region is .NET
[11:25] Wolf.Freeman @free-world.spdns.de:8002: what is better ?
[11:25] Wolf.Freeman @free-world.spdns.de:8002: co-op odr aboard
[11:26] Andrew Hellershanks: I know people who have shoutcast streams so I can do testing with them.
[11:26] Richardus.Raymaker @hypergrid.org:8002: did co-op not have problems with llsleep ?
[11:26] Nebadon Izumi: well in mono most people are running single AppDomain
[11:26] Shoutcast Radio & DJ Board.1.9/works flüstert: Reading config notecard...
[11:26] Nebadon Izumi: where Windows is usaully running multiple app domains
[11:26] Aine.Caoimhe @refugegrid.com:8002: yes
[11:26] Nebadon Izumi: running seperate app domains
[11:26] Wolf.Freeman @free-world.spdns.de:8002: true
[11:26] Nebadon Izumi: wats a ton more memory
[11:26] Richardus.Raymaker @hypergrid.org:8002: it works fine on windowx :)
[11:27] Richardus.Raymaker @hypergrid.org:8002: x = s :O
[11:27] Nebadon Izumi: that could be why windows appears to use more memory
[11:27] Nebadon Izumi: then there is 32bit vs 64bit
[11:27] Nebadon Izumi: 64bit always appears to use more memory
[11:27] Nebadon Izumi: could be many reasons why honestly
[11:27] Aine.Caoimhe @refugegrid.com:8002: mhmmm
[11:28] Dahlia Trimble: references are twice the size in 64 bit
[11:28] Nebadon Izumi: right
[11:28] Aine.Caoimhe @refugegrid.com:8002: differences in memory use are many; but once you have a sort of baseline for your own setup your eyebrows tend to go up when there's a drastic change
[11:28] Aine.Caoimhe @refugegrid.com:8002: the other day I suddenly dropped from a little over 1GB down to 250MB and haven't the faintest idea why
[11:29] Aine.Caoimhe @refugegrid.com:8002: then after the next startup it was back up to 875MB which is "normal" for me
[11:29] Dahlia Trimble: memory is very difficult to measure in a managed environment
[11:29] Andrew Hellershanks: You also wonder what's gone wrong when the memory usage drops drastially.
[11:29] Aine.Caoimhe @refugegrid.com:8002: so how it was managing to only use 250 for a day or two is mystifying
[11:29] Andrew Hellershanks: s/drastially/drastically/
[11:30] Aine.Caoimhe @refugegrid.com:8002: and I checked to make sure the system was reporting that same amount so it wasn't just a case of Opensim mis-reporting
[11:30] Nebadon Izumi: only thing I could think is maybe xengine or something actually hung up
[11:31] Nebadon Izumi: and even though things appeared to be working, may not have been
[11:31] Dahlia Trimble: a lot depends on how often and how successful garbage collection runs
[11:31] Aine.Caoimhe @refugegrid.com:8002: doubt it....everything was working fine
[11:31] Nebadon Izumi: well it could be also how .net and mono handle resources
[11:31] Nebadon Izumi: as resources tighten on the machine
[11:31] Aine.Caoimhe @refugegrid.com:8002 nods
[11:32] Nebadon Izumi: they may get more aggressive in garbage collection
[11:32] Aine.Caoimhe @refugegrid.com:8002: I've just never seen that happen before
[11:32] OtakuMegane Desu: Garbage collection in seems to be a weird and sometimes unpredictable thing in most applications.
[11:32] Nebadon Izumi: .net is way more aggressive than mono is
[11:32] Dahlia Trimble: mono GC has always had problems
[11:32] Aine.Caoimhe @refugegrid.com:8002: over time mine normally climbs at if fails to clear our meshes that I'm working on
[11:32] Aine.Caoimhe @refugegrid.com:8002: so eventually I have to restart
[11:32] Nebadon Izumi: orders of magnitude more aggressive
[11:32] Aine.Caoimhe @refugegrid.com:8002: but i've never seen it suddenly drop so drastically
[11:32] Nebadon Izumi: yea me either
[11:32] Nebadon Izumi: not that drastic
[11:33] Dahlia Trimble: I have
[11:33] Nebadon Izumi: usually for me its the other way
[11:33] Nebadon Izumi: one day a region is using 3 times more memory
[11:33] Aine.Caoimhe @refugegrid.com:8002: to only 1/4 of the baseline start-up value?
[11:33] OtakuMegane Desu: The only two things I can think is either a little-run garbage collection triggered or something in the simulator that's normally bloating memory was gone or inactive.
[11:34] Aine.Caoimhe @refugegrid.com:8002: yeah....it's just the regions on that instance are pretty light.....4 of them with maybe 250 scripts total and a bunch of assorted mesh objects so it's usually very stable
[11:35] Nebadon Izumi: and this is on Windows right Aine?
[11:35] Aine.Caoimhe @refugegrid.com:8002: and I hadn't done anything unusual at all in that 24-hour period
[11:35] Aine.Caoimhe @refugegrid.com:8002: yes, Win7-64
[11:35] Nebadon Izumi: are you running AppDomainLoading = true?
[11:35] Nebadon Izumi: 1gb for such light regions seems very high
[11:35] OtakuMegane Desu: I've had my sandbox run wild and even crash because someone rezzed less than that. Never figurte out why, just something knocked stuff out of whack.
[11:35] Nebadon Izumi: When Right Plaza starts up initallly
[11:35] Nebadon Izumi: its only using like 900mb or something
[11:36] Nebadon Izumi: this region is insane
[11:36] Dahlia Trimble: I can't say I've ever seen .NET GC fail, but I've seem Mono fail a *lot*
[11:36] Nebadon Izumi: AppDomainLoading=True does really shoot memory up high
[11:36] Nebadon Izumi: you might try false
[11:36] Nebadon Izumi: you can run false in windows
[11:36] Nebadon Izumi: it will use a lot less memory
[11:36] Aine.Caoimhe @refugegrid.com:8002: it's running set to true
[11:37] Aine.Caoimhe @refugegrid.com:8002: but usually memory isn't an issue for me
[11:37] Nebadon Izumi: worth experimenting
[11:37] Nebadon Izumi: yea
[11:37] Nebadon Izumi: it might help with threading too though
[11:37] Aine.Caoimhe @refugegrid.com:8002: and it's not a case of bloat....it's a case of that sudden inexplicable drop
[11:37] Nebadon Izumi: yea
[11:37] Aine.Caoimhe @refugegrid.com:8002: it was running 4 regions with all its stuff in it in just under 250MB of process memory
[11:37] Aine.Caoimhe @refugegrid.com:8002: which was....surprising
[11:38] Nebadon Izumi: 4 regions in a single instance?
[11:38] Aine.Caoimhe @refugegrid.com:8002: yes
[11:38] Nebadon Izumi: hmm
[11:38] Richardus.Raymaker @hypergrid.org:8002: the biggest region 2var's 3x3 only use 1GB on my windows. and appdom must be true
[11:38] Richardus.Raymaker @hypergrid.org:8002: one instance btw
[11:38] Nebadon Izumi: im leaning towards something with AppDomains malfunctioning
[11:38] Aine.Caoimhe @refugegrid.com:8002: normally those 4 regions are using about 800MB on start-up
[11:38] Nebadon Izumi: seeing that drastic
[11:38] OtakuMegane Desu: That itself seems high
[11:38] Nebadon Izumi: its the only thing I can think that would swing so far with memory like that
[11:39] Nebadon Izumi: thats one of the big problems with running multiple region instances
[11:39] Aine.Caoimhe @refugegrid.com:8002: keep in mind that when I say a "bunch of mesh" I'm talking about a *lot* of high poly models :p
[11:39] Nebadon Izumi: you are really cutting up resource availability
[11:39] Nebadon Izumi: thats why i refuse to use it myself
[11:40] Nebadon Izumi: always 1:1 for my stuff
[11:40] Richardus.Raymaker @hypergrid.org:8002: Naughty aine.. not sounds really optimized
[11:40] Aine.Caoimhe @refugegrid.com:8002: it's not a public-accessible set of regions
[11:40] Nebadon Izumi: personally I wish we didnt allow it
[11:40] Andrew Hellershanks: refuse to use mesh??
[11:40] Aine.Caoimhe @refugegrid.com:8002: they're my testing/experimentation/NDA workplace
[11:40] Nebadon Izumi: no
[11:40] Nebadon Izumi: refuse to have support for multiple region simulators
[11:40] OtakuMegane Desu: That works unless you have many regions. dealing with 9 instances is enough as it is.
[11:40] Nebadon Izumi: but I wont go there
[11:40] Nebadon Izumi: I know most people like it
[11:40] Aine.Caoimhe @refugegrid.com:8002: so it doesn't need to be optimized
[11:40] Andrew Hellershanks: That makes ore sense
[11:41] Andrew Hellershanks: s/ore/more/
[11:41] Aine.Caoimhe @refugegrid.com:8002: and if anything gets out of whack I just restart
[11:41] Aine.Caoimhe @refugegrid.com:8002: anything that I have open to public access is 1:1
[11:41] Nebadon Izumi: ah
[11:41] Nebadon Izumi: ya
[11:41] OtakuMegane Desu: It should be much less necessary once vars can properly coexist among other regions
[11:41] Richardus.Raymaker @hypergrid.org:8002: nebadon it's ok. if you don't abuse it. and there's the problem there are some people that abuse it and put 30 sims on one simulator and pump it full with prims etc.
[11:41] Nebadon Izumi: so Aine
[11:41] Nebadon Izumi: the only ones you have trouble with
[11:41] OtakuMegane Desu: THat was usually my reason for large numbers of regions
[11:41] Nebadon Izumi: are Multiple Region Simulators?
[11:41] Nebadon Izumi: or no?
[11:42] Dahlia Trimble: I like multiple region instances :P
[11:42] Nebadon Izumi: it is convienient for sure Dahlia
[11:42] Aine.Caoimhe @refugegrid.com:8002: I almost never have any trouble at all with mine....and suddenly freeing up a ton of memory isn't a "problem" :p...just a cuirosity
[11:42] Nebadon Izumi: but I think ultimately is the source of many problems for people
[11:42] Andrew Hellershanks: I used to run up to 9 regions in one but that was ages ago.
[11:43] Nebadon Izumi: well it could be a problem
[11:43] Aine.Caoimhe @refugegrid.com:8002: the one where Ken is experiencing issues with shoutcast board I think is a *lot* of regions running on it
[11:43] Richardus.Raymaker @hypergrid.org:8002: it saves me resources to but in this case 23 vars on one simuator and it dont have any impact
[11:43] OtakuMegane Desu: Multiple region instances aren't bad. I most use them for "borders" or non-square/rectangular areas that are lightweight
[11:43] Dahlia Trimble: I don't see people having extraordinary problems with them
[11:43] Nebadon Izumi: to me Sudden huge memory shift downward
[11:43] Nebadon Izumi: to me that says something is not loading
[11:43] Aine.Caoimhe @refugegrid.com:8002: but I don't know whether he's running 1:1 or something else
[11:43] Nebadon Izumi: and maybe your just not noticing
[11:43] OtakuMegane Desu: I'd never try to run say, multiple plazas in one instance though
[11:43] Andrew Hellershanks: Aine, check the region stats for the number of active scripts and number of script lines per second
[11:43] Nebadon Izumi: to me thats just as alarming as memory use trippeling or something
[11:43] Richardus.Raymaker @hypergrid.org:8002: uhh. that i would not do too otaku
[11:44] OtakuMegane Desu: lol
[11:44] Nebadon Izumi: yea it would be good to capture the stats
[11:44] Nebadon Izumi: for comarison
[11:44] Nebadon Izumi: when memory is normal vs memory is high
[11:44] Andrew Hellershanks: hehe... how many copies of this region under one Simulator instance would it take to grind things to a halt?
[11:44] Nebadon Izumi: definitely focus on scripts
[11:44] Dahlia Trimble: so many conclustions drawn from a problem that's not understood
[11:44] Nebadon Izumi: my guess is thats where the problem is
[11:44] OtakuMegane Desu: Usually the only times I have issues with multi-region instances is if it's a huge number like 40+ or they're heavily loaded.
[11:44] Richardus.Raymaker @hypergrid.org:8002: 2-3
[11:44] OtakuMegane Desu: Otherwise mostly-empty regions sit there very nicely
[11:45] OSG8B POSTER: active
[11:45] Morgana XSmit kommt in Chat-Reichweite (17.20 m).
[11:45] Aine.Caoimhe @refugegrid.com:8002: it's 4 regions so it's reporting them separately
[11:45] Aine.Caoimhe @refugegrid.com:8002: doing the math....
[11:45] Andrew Hellershanks: sure. You could have a nice sailing area that are all under a single instance
[11:46] Morgana XSmit kommt in Chat-Reichweite (17.28 m).
[11:46] Nebadon Izumi: ya mostly its just to compare that you have the same script loads running etc..
[11:46] OtakuMegane Desu: Or just make a var, now.
[11:47] Aine.Caoimhe @refugegrid.com:8002: 190 scripts loaded.....0 scripts waiting to load.....roughly 1.3 million events processed
[11:47] Aine.Caoimhe @refugegrid.com:8002: 4 sensors....9 timers
[11:47] Aine.Caoimhe @refugegrid.com:8002: 8 listeners
[11:47] Nebadon Izumi: I think its pretty clear from what Ken_S has shown that all it takes is 1 bad script
[11:48] Nebadon Izumi: and your memory consumption can double or tripple or more
[11:48] Aine.Caoimhe @refugegrid.com:8002: yes
[11:48] Andrew Hellershanks: Aine, if you have webstats enabled you can get the ActScr and ScLPS number.
[11:48] Andrew Hellershanks: 190 scripts isn't that many
[11:48] Nebadon Izumi: ya in Ken_S case it was 1 script
[11:48] Aine.Caoimhe @refugegrid.com:8002: yeah....it's very light....as I say it's my test instance (and I don't have webstats enabled)
[11:48] Nebadon Izumi: so many times people have come to IRC
[11:49] Andrew Hellershanks: Bad scripts often show up on top scripts with high run times
[11:49] Nebadon Izumi: all the sudden their region stars acting odd
[11:49] Nebadon Izumi: lag, or more memorh
[11:49] Nebadon Izumi: so many times its 1 item in the region
[11:49] Nebadon Izumi: a lost vehicle, a physical prim stuck on border
[11:49] Dahlia Trimble: I bet I could crash a sim with one script ^.^
[11:49] Aine.Caoimhe @refugegrid.com:8002: shoutcast board was tiny when I last looked at it
[11:49] Nebadon Izumi: hehe im sure you could Dahlia
[11:49] Aine.Caoimhe @refugegrid.com:8002: 0.008ms iitc
[11:49] Nebadon Izumi: i bet I could as well :)
[11:49] Andrew Hellershanks: what about when it isn't idle, or does it only go active for a very short period of time?
[11:50] Andrew Hellershanks: If it is the shoutcast board, disable its scripts then see if the region perks up again
[11:50] Aine.Caoimhe @refugegrid.com:8002: I haven't monitoried it closely enough when in use....I had hoped to get to that yesterday but got tied up in something else
[11:50] Nebadon Izumi: Andew as soon as he removed the board things got better was my understanding
[11:51] Aine.Caoimhe @refugegrid.com:8002: the one that's quite a lot more concerning is a case where memory isn't being released after people leave a region
[11:51] Andrew Hellershanks: ok. Sounds like something in the device causing problems. Shouldn't be just its prims.
[11:51] Richardus.Raymaker @hypergrid.org:8002: Nebdon, is the board using lists etc ?
[11:51] Nebadon Izumi: I havent had a chance to review ityt Richardus
[11:51] Andrew Hellershanks: That shoutcast box is no mod so I'll have to see what parts of it have the no mod perms.
[11:52] Nebadon Izumi: I sent you a copy Richardus
[11:52] Nebadon Izumi: anyone else wants a copy just let me know
[11:52] Dahlia Trimble: I'll take one
[11:52] Andrew Hellershanks: Is the shoutcast box something that originated over in SL?
[11:52] Aine.Caoimhe @refugegrid.com:8002: in my Hedonism sim after start-up it's using hardly anything (200MB), then during a party as people arrive it will slowly climb to somewhere in the 700-800MB range (sometimes higher)....but then after everyone has left and the region is empty it never settles back down to under 600MB (sometimes even higher)
[11:52] Dahlia Trimble: ty
[11:52] Nebadon Izumi: fedex delivery dispatched :)
[11:53] Aine.Caoimhe @refugegrid.com:8002: so that excess 400+MB is never released
[11:53] Richardus.Raymaker @hypergrid.org:8002: meh, no mod. so cannot look here
[11:53] Aine.Caoimhe @refugegrid.com:8002: and if I do another party there before restarting the instance it will continue to grow like that
[11:53] Nebadon Izumi: ever?
[11:53] Andrew Hellershanks: Aine, it might be using sensors to determine how many people are listening to the stream?
[11:53] Aine.Caoimhe @refugegrid.com:8002: never
[11:53] Aine.Caoimhe @refugegrid.com:8002: i t will *never* release until instance start
[11:53] Nebadon Izumi: ya thats kind of unusual in .net
[11:53] Richardus.Raymaker @hypergrid.org:8002: sensor cannot know how many listen
[11:53] Nebadon Izumi: that likely means its outside of AppDomains
[11:53] Nebadon Izumi: unless your running false
[11:54] Aine.Caoimhe @refugegrid.com:8002: that's in Hedonism which Seth hosts on the RG servers so ApDomains= false on that one
[11:54] Nebadon Izumi: oh
[11:54] Nebadon Izumi: thats why
[11:54] Nebadon Izumi: what happens with False
[11:54] Andrew Hellershanks: Richardus, assumes that the people in the area are listening perhaps? Or using it for some kind of statistic?
[11:54] Nebadon Izumi: you get a single AppDomain for all scripts
[11:54] Nebadon Izumi: you can never unload them
[11:54] Nebadon Izumi: thats expected behavior for mono
[11:54] Nebadon Izumi: unfortunately
[11:55] Aine.Caoimhe @refugegrid.com:8002: so even though I disallow scripts it's going to just eat memory for every worn script that enters the region?
[11:55] Nebadon Izumi: one of the proposes design changes by Justin was to seperate Avatar app domains
[11:55] Nebadon Izumi: so tha each avatar has thier own domain
[11:55] Nebadon Izumi: that gets closed when they leave the region
[11:55] Aine.Caoimhe @refugegrid.com:8002 nods
[11:55] Aine.Caoimhe @refugegrid.com:8002: would make sense
[11:55] Andrew Hellershanks: If the device is from SL I've sometimes made small changes to scripts from there to really improve their performance.
[11:55] Nebadon Izumi: what they do In SL
[11:55] Nebadon Izumi: with app domains
[11:56] Nebadon Izumi: is that when the app domain hits a certain threshold
[11:56] Nebadon Izumi: they start a new app domain and transfer over only active states
[11:56] Nebadon Izumi: then close old one
[11:56] Nebadon Izumi: a rough GC basically
[11:56] Nebadon Izumi: they can even move app domain to a new region
[11:56] Aine.Caoimhe @refugegrid.com:8002 nods
[11:56] Nebadon Izumi: and move all avatars simulanouesly
[11:56] Dahlia Trimble: how did you find that out?
[11:57] Nebadon Izumi: there was a guy that used to come here a lot
[11:57] Nebadon Izumi: that worked on it
[11:57] Nebadon Izumi: I forget his name now
[11:57] Nebadon Izumi: he used to pass me details on exploits he would find
[11:57] Richardus.Raymaker @hypergrid.org:8002: latif or soemthing like that ?
[11:57] Nebadon Izumi: and I would pass them along to Melanie or Diva
[11:57] Andrew Hellershanks: If it used something like linked lists you could just drop the entry that points to an app domain block and it would free things up faster. No need for app domain on or off.
[11:57] Nebadon Izumi: to fix
[11:59] Andrew Hellershanks: This hour flew by
[12:00] Dahlia Trimble: bye all :)
[12:00] Vivian Klees: bye Dahlia
[12:00] Andrew Hellershanks: I'm going to hit a sandbox and unpack this shoutcast stuff.
[12:01] Andrew Hellershanks: See you next week.
[12:01] Aine.Caoimhe @refugegrid.com:8002: have a good week, all
[12:01] Nebadon Izumi: ya dang I cant remember the guys name oh well
[12:01] Simulator Version v0.5 ruft: OpenSim 0.8.2.0 Dev        8aa75f2: 2015-07-26 14:09:17 +0300 (Unix/Mono)
[12:01] Aine.Caoimhe @refugegrid.com:8002 is going to head out too
[12:01] Nebadon Izumi: ok thanks everyone
[12:01] Nebadon Izumi: see you next week! :)
[12:01] Andrew Hellershanks: Scroll down your list of friends/
[12:02] Nebadon Izumi: I did I am not sure why he isnt on my friends list
[12:02] Nebadon Izumi: he was a very secretive kind of person heh
[12:02] Nebadon Izumi: he was not a Linden
[12:02] Andrew Hellershanks: that's probaby why
[12:02] Nebadon Izumi: he was a contractor hired to find exploits
[12:02] Nebadon Izumi: in SL
[12:02] Andrew Hellershanks: secretive so didn't exchange friendship
[12:02] Nebadon Izumi: and he helped find them here in his free time as well
[12:02] Andrew Hellershanks: oh, cool.
[12:02] Nebadon Izumi: he was very knowledgable of the back end of SL
[12:03] Andrew Hellershanks: ok, I'm off to a sandbox then logging out. ttyl
[12:03] Nebadon Izumi: see you guys!
[12:03] Snakedance.Moonwing @snaketopia.darktech.org:8002: :)
[12:03] Nebadon Izumi: Sheera an you post logs?
[12:03] Nebadon Izumi: I missed 12 minutes in begining again
[12:03] Sheera.Khan @hypergrid.org:8002: yes Neb - just arrived back
[12:03] Nebadon Izumi: thanks!
Personal tools
About This Wiki