Simulator User Group/Transcripts/2011.09.27

From Second Life Wiki
Jump to: navigation, search


Prev 2011.09.23 Next 2011.09.30

List of Speakers

Andrew Linden Kallista Destiny
Kelly Linden Latif Khalifa
Liisa Runo Motor Loon
Pauline Darkfury Qie Niangao
Rex Cronon Sahkolihaa Contepomi
Simon Linden TankMaster Teichmann
Vincent Nacon


[12:00] Motor Loon: dadaaa

[12:00] Sahkolihaa Contepomi: Hey Andrew.

[12:01] Liisa Runo: hiyas

[12:01] Andrew Linden: hello

[12:01] Kallista Destiny: Hi Andrew

[12:01] Sahkolihaa Contepomi: And, hey Simon.

[12:01] Kallista Destiny: Hi Simon

[12:01] Simon Linden: Hello everyone

[12:02] Motor Loon: Nice hat simon

[12:03] Andrew Linden: Ok news...

[12:03] Rex Cronon: hello everybody

[12:04] Motor Loon:

[12:04] Andrew Linden: There was an update to the main server channel today, but it was just an isolated fix for one bug... lemme find it.

[12:04] Andrew Linden: SVC-7283

[12:04] JIRA-helper:

[#SVC-7283] llRemoteLoadScriptPin and llGiveInventory fail when installing non-full perm items into worn attachments

[12:04] Motor Loon: problem all solved then?

[12:05] Latif Khalifa: So you decided not to ship the physics lsl bug with it?

[12:05] Andrew Linden: We already have a report that this fix introduced a new bug: SCR-7321

[12:05] Andrew Linden: so no, problem not all solved

[12:05] Motor Loon: ah ;-(

[12:06] Vincent Nacon: The issue you are trying to view does not exist.

[12:06] Kallista Destiny: Security?

[12:06] Andrew Linden: unfortunately, this new bug existed on all three RC channels for a week (30% of the grid?) and no one noticed until today.

[12:06] Motor Loon: "Rules of the game"

[12:06] Andrew Linden: er... SVC-7321

[12:06] JIRA-helper:

[#SVC-7321] llRemoteLoadScriptPin() does not allow injection of an O:VMCT script into an O:PERM_ALL worn attachment

[12:07] Motor Loon: Guess people was too busy whining to actually test the bugfix ,-)

[12:07] Vincent Nacon: they got business to focus on, that's why

[12:07] Sahkolihaa Contepomi: Hey Kelly.

[12:07] Kallista Destiny: Hi Kelly

[12:08] Rex Cronon: hello

[12:08] Andrew Linden: My "slightly faster scripts" project should hit one of the RC channels this week

[12:08] Motor Loon: ey Kelly

[12:08] Kallista Destiny: Anyth8ng that speds up scripts is a win

[12:08] Andrew Linden: probably BlueSteel

[12:09] Vincent Nacon: but only slightly

[12:09] Kallista Destiny: 1% is btter that a poke in the eye with a sharp stick

[12:09] Pauline Darkfury: hi folks :)

[12:10] Andrew Linden: oh yeah, I sent off that email summary of the "returnability of encraoching objects" to Patch Linden, so hopefully he'll help push it out soon.

[12:10] Vincent Nacon: how unrelated but yeah, I guess

[12:10] Pauline Darkfury: yay, thanks, Andrew, on behalf of everyone that cares about mainland :)

[12:10] Andrew Linden: That's all for my news for today.

[12:10] Vincent Nacon: mainland? what mainland?

[12:11] Liisa Runo: mainland is the real SL

[12:11] Vincent Nacon: I was kidding

[12:11] Sahkolihaa Contepomi: Andrew, do you know how much memory the script VM gets per sim? I was under the impression that it was 500MiB, but sims I'm an EM of, I see using up to 1GiB. So now I don't know.

[12:11] Pauline Darkfury: The "usage" number you can see isn't actually usage

[12:12] Kelly Linden: What number are you looking at for 'script vm' memory size?

[12:12] Pauline Darkfury: It's a worst case scenario that is very unlikely to ever be reached or even approached in practice

[12:12] Sahkolihaa Contepomi: Under about land > script info > region

[12:12] Kelly Linden: There is no current limit.

[12:12] Sahkolihaa Contepomi: Ah. I've heard it was 500MiB. Good to know.

[12:13] Kelly Linden: 500 is just a tad on the high side.

[12:13] Pauline Darkfury: Tbh, script count is still by far the best number to look at for load, combined with the script events rate

[12:13] Kelly Linden: but not outragious probably.

[12:13] Pauline Darkfury: And script time, of course

[12:13] Sahkolihaa Contepomi: My viewer was saying one sim I help look after, Deshima, uses around 800MiB.

[12:13] Motor Loon: ..but memory use is the real killer aint it?

[12:13] Andrew Linden: Motor, you have a funny hat (is that a petroleum distilery on the back of that hat?)

[12:14] Pauline Darkfury: No, by my reckoning excessive script CPU is a bigger killer in practice than excessive script memory

[12:14] Kelly Linden: If you hit the wall, memory is worse.

[12:14] Pauline Darkfury: 150 badly behaved scripts can bring a sim to a near standstill

[12:14] Kelly Linden: Script CPU should only effect other scripts.

[12:14] Pauline Darkfury: Yeah, if you actually do trip past the physical memory, it's going to be horrible

[12:14] Latif Khalifa: unless they're rezzing stuff

[12:14] Kelly Linden: That has more to do with what those scripts are doing that is 'badly behaved'

[12:15] Pauline Darkfury: Ok, yeah, the knock on effects from what the script is doing

[12:15] Kelly Linden: Right, rezzing is hard on the sim. So lots of rezzing - by script or bot or agent - has a negative impact on performance.

[12:15] Motor Loon: ah well, guess anything is bad if you misuse it enough

[12:16] Kallista Destiny: Onewould hope that regions limit their VM us so they NEVER start paging.

[12:16] Kallista Destiny: by refusing to rez new objcts etc

[12:17] Kallista Destiny: Question, is rezzing an objects which is a copy of an already existing in the region better?

[12:17] Pauline Darkfury: Mono scripts copied after compilation have 1 copy of bytecode per region, so it's better in that regard

[12:17] Vincent Nacon: no, not really because each one got their own Key

[12:18] Vincent Nacon: maybe with textures

[12:18] Andrew Linden: slightly better in some cases. if the source asset exists in the simulator cache then it doesn't have to wait for it to download

[12:18] Andrew Linden: however, it still has to parse it

[12:18] Kallista Destiny: nods

[12:18] Vincent Nacon: mesh...? no idea

[12:18] Pauline Darkfury: Yeah, you've got 1 set of textures, sounds, anims, Mono scripts, config NCs, etc

[12:18] Motor Loon: is mesh cached on the server?

[12:18] Andrew Linden: making a duplicate of an existing object (CTRL + d or SHIFT-drag) doesn't read asset

[12:19] Kallista Destiny: it would be interesting to know how to make an 'easy to parse' object

[12:19] Andrew Linden: yes, I believe mesh data is cached at the simulator host

[12:19] Motor Loon: k

[12:19] Kallista Destiny: Hi Tank

[12:19] TankMaster Teichmann: hey

[12:19] Vincent Nacon: heya Noob

[12:19] Andrew Linden: the simple answer to an "easy to parse" object would be a single-prim object without any scripts

[12:20] Kallista Destiny: lol

[12:20] Sahkolihaa Contepomi: Just been asked - do regions in general, have a RAM limit (aka for everything)?

[12:20] Pauline Darkfury: One easy to parse object :p

[12:20] Andrew Linden: No. The 'simulator' server (which runs the region) does not have a memory limit.

[12:20] Sahkolihaa Contepomi: Thanks.

[12:20] Kelly Linden: Koli: 4gb (ie it is a 32bit process, so has 4gb available)

[12:21] Sahkolihaa Contepomi: Had a feeling that would be the only limit.

[12:21] Kallista Destiny: and the class 5 machines have 16GB?

[12:21] Kelly Linden: It will start paging LONG before it hits that though, and performance will tank.

[12:21] Simon Linden: We watch some memory (physics usage) and react if they start to bloat

[12:21] Sahkolihaa Contepomi: clears up my confusion. Thanks. ^^

[12:21] Pauline Darkfury: Mmm, that's a thing, has moving to 64 bit been considered recently? (I know it was a long time ago)

[12:21] Andrew Linden: physics memory usage can spike very high under some situations

[12:22] Vincent Nacon: I hope not, Pauline

[12:22] TankMaster Teichmann: i see my bros are here :)

[12:22] Kelly Linden: Pauline: not receently. It has implications on mono script size.

[12:22] Vincent Nacon: it would cost them more in hardware upgrades

[12:22] Kallista Destiny: That begs the question does the Simulator have a way of detecting that it's about to stsrt paging?

[12:22] Pauline Darkfury: Not if the hardware is already x86_64.

[12:22] Kelly Linden: I don't think there is a hardware upgrade or cost associated with moving to 64bit for the sim.

[12:22] Vincent Nacon: I meant HDDs

[12:23] Vincent Nacon: and RAMs

[12:23] Pauline Darkfury: If there's known negatives, such as some serious memory bloat issues, that could be a showstopper for it

[12:23] Kallista Destiny: and if so can it restrict itself so it DOESN'T start paging?

[12:23] Simon Linden: Kallista - no, there's no special checks I know of about the paging threshold

[12:23] Simon Linden: we leave that all up to the OS to sort out

[12:23] Kallista Destiny: Paging is the performance killer

[12:24] Kallista Destiny: Is a Simulator one process?

[12:24] Kallista Destiny: with threads?

[12:24] Sahkolihaa Contepomi: Thankfully Linux handles memory nicely.

[12:24] Motor Loon: Not being able to get memory would kill too

[12:24] Simon Linden: There are multiple threads, Kallista

[12:24] Andrew Linden: yes, the simulator is a single process most of the time, with multiple threads

[12:24] Simon Linden: Some of the lag work over the years has been moving jobs to side threads

[12:24] Andrew Linden: but it forks autosaves once an hour

[12:24] Sahkolihaa Contepomi: Hm. Is the debian install you guys are using 32-bit or 64-bit?

[12:24] Simon Linden: But it's mostly one main thread

[12:25] TankMaster Teichmann: yes

[12:25] Simon Linden: I have to leave early today, so I'll see you all next time

[12:25] Vincent Nacon: seeya

[12:25] Rex Cronon: tc

[12:25] Simon Linden: bye :)

[12:25] Sahkolihaa Contepomi: Aw, see you simon.

[12:25] Kallista Destiny: nods So at least therotetically you could set the per simulator VM limit to be no more than its sahre ot the machine

[12:25] Pauline Darkfury: Take care, Simon, have a good one :)

[12:25] Kallista Destiny: By Simon

[12:26] Kallista Destiny: share

[12:26] Kallista Destiny: to avoid paging.

[12:27] Kallista Destiny: I've gone through enormous contortions to avoid pagin in RT applications.

[12:28] Andrew Linden: RT applications? interfacing with hardware or test equipment?

[12:29] Pauline Darkfury: Tbh, with the monthly cost of regions, I personally feel quite strongly that the solution to paging is to have a medium and long term plan in place to make sure that most non-abuse cases have enough memory

[12:29] Kallista Destiny: Hardware, EEG machines detecting Seizures and dumping the data to an Analog tape, and to a Pollygraph

[12:30] Kallista Destiny: with a lot of pre-event data... it was fun

[12:30] Andrew Linden: Pauline, what do you mean by "non-abuse cases"?

[12:30] Pauline Darkfury: The 7s already have more memory that the 5s, I believe. The 5s are now getting relatively long in the tooth, if they are still the same spec as when first introduced. They have to be reaching the stage where it would be more economic to progressively replace them than maintain them

[12:30] Meeter: Timecheck : User Group is half over

[12:30] Kallista Destiny: 5s are what 4 years old?

[12:31] Vincent Nacon: 3 I think

[12:31] Pauline Darkfury: Well, basically provision enough memory on average to cover 99.999% of current cases where someone isn't deliberately trying to kill an entire host

[12:31] Latif Khalifa: Andew, with $295/mo for a sim, it wouldn't be unreasonable to expect to have 1GB ram available

[12:32] Motor Loon: You aint paying for server hosting

[12:32] Latif Khalifa: i know, i would be paying $30 for class 5

[12:32] Kallista Destiny: Oh then what is the $295/mo about?

[12:32] Pauline Darkfury: Well, tbh, in many ways we are paying for hosting, just a more complex form of hosting, with huge central DBs, etc

[12:33] Kallista Destiny: $30/month doesn't even cover to power and cooling bill.

[12:33] Motor Loon: I disagree... we're also paying for R&D and support etc. etc. etc.

[12:34] Latif Khalifa: I know, that's why we pay 5x - 10x as much as hosting would cost. but i say it's not unreasonable to expect to have 1GB ram for that kind of money

[12:34] Andrew Linden: Personally I view it like a hosting service. Support is just necessary. So is R&D if we hope to stay in business.

[12:34] Rex Cronon: if i am not wrog u can rent your own server for 100 per month

[12:35] Latif Khalifa: Kallista, you should check out what 50 eur buys you these days

[12:35] Kallista Destiny: All in all I think the ~300 USD a month is quite reasonable.

[12:35] Latif Khalifa: i7-quad with 8gb ram

[12:36] Motor Loon: well, you are free to go rent a dedicated server - but you might not see much effect on your Second Life experience from it ;-)

[12:36] Pauline Darkfury: We've talked about improvements to the Region Conductor. If better balancing of the load across the server farms is done there, progressively replacing older low-memory hardware with newer high-memory hardware could solve the general issues with script memory (which I personally don't believe are really as common as some think), without causing disruption to existing use cases

[12:36] Kallista Destiny: no that is what the $1000 start up fee buys

[12:36] Vincent Nacon: this is becoming the wrong meeting to discusse about billing for the server

[12:37] Latif Khalifa: I'm just saying it wouldn't be unreasonable to expect to have 1GB script ram for ~300 USD (on one querter of a server).

[12:37] Liisa Runo: i agree with Latif. And my calculations say that over 80% of money LL get from tier is pure profit. And my calculations give pretty damn good salary for every linden, including the one who leans the office bathrooms

[12:37] Liisa Runo: cleans*

[12:37] Kallista Destiny: 1/4 of a class 5 + 1/12 of a rack and wireing etc...

[12:37] Rex Cronon: imo, ll might do better if it were to just license others to run the sim servers:)

[12:38] Latif Khalifa: Kallista, click on that link above for what you get for 50 and then talk to me about wiring :P

[12:38] Vincent Nacon: anyway... Andrew, when do you think I can expect LSL function codes for density, friction, gravity etc etc

[12:38] Vincent Nacon: ?

[12:38] Kelly Linden: I think those are on LeTigre now.

[12:38] Andrew Linden: Vincent, you mean for setting those parameters?

[12:38] Motor Loon: Yep they are

[12:38] Vincent Nacon: ah ok

[12:38] Motor Loon: + scripting physics shapes

[12:39] Andrew Linden: What are the LSL method names Motor?

[12:39] Vincent Nacon: not llSPP I hope

[12:39] Rex Cronon: i think we need bot setter and getter function for those values

[12:39] Vincent Nacon: can't find it

[12:40] Motor Loon:

[12:40] Vincent Nacon: ty

[12:40] Qie Niangao: that reminds me.. any interest (besides mine) in a script function to *set* PARCEL_DETAILS_SEE_AVATARS ?

[12:40] Vincent Nacon: ahh llSetPhysicsMaterial

[12:40] Kallista Destiny: llisp( ((())))

[12:40] Motor Loon: much better than llSPP

[12:40] Vincent Nacon: still kinda long name

[12:41] Andrew Linden: Oh yeah... that llGetMassMKS() is for Vincent, I think.

[12:41] Vincent Nacon: oy... I didn't even ask fo rit

[12:41] Vincent Nacon: for*

[12:41] Pauline Darkfury: Mmmm, talking of new LSL methods, can you confirm that OBJECT_PRIM_EQUIVALENCE is supported and expected to give correct results now? It's there, appears to work, but was absent from the wiki (I added it).

[12:41] Kelly Linden: No.

[12:41] Kelly Linden: It is not supported at this time.

[12:41] Andrew Linden: maybe I'm misremembering... I'll have to browse old transcripts.

[12:41] Vincent Nacon: but thanks, I guess

[12:42] Kelly Linden: Keep your eyes peeled for "DRTSIM-98" Pauline.

[12:42] Pauline Darkfury: Ok, thanks :)

[12:42] Vincent Nacon: I just don't think it's good idea to add another function that is basically the same of a old one with proper info

[12:42] Pauline Darkfury: I'll amend the wiki to reflect it not working yet

[12:42] Vincent Nacon: I mean it's hardly a "fix"

[12:43] Kallista Destiny: I suspect that OBJECT_PRIM_EQUIVALENCE will b e a moving target for some time

[12:43] Vincent Nacon: I understand llGetMass isn't correct as it should have been but to the point that Density shouldn't have to use the actual "correct" value when it doesn't match up with llGetMass

[12:44] Vincent Nacon: so it's easier to fix density than having to add another llGetMass function for it

[12:44] Motor Loon: oh oh ... about - could one of you check the testregions are ok, I tried earlier today, but couldnt teleport to either of them on ADITI

[12:44] Motor Loon: looked ok, just wouldnt let me in

[12:45] Andrew Linden: using wacky units for density was deemed more objectionable by some

[12:45] Vincent Nacon: I know... it's better not to add more trouble than adapting "Lindogram"

[12:45] Andrew Linden: sure, I'll look at the keyframe demo regions...

[12:45] Motor Loon: Thanks

[12:46] Kallista Destiny: I'd much rather have 'real' units for desnity and weight. When will these be pplied to Avatars. My Scale code is really an ugly hack.

[12:46] Andrew Linden: yeah, they're down. I wonder why...

[12:47] Vincent Nacon: yeah avatar's mass is another issue

[12:47] Motor Loon: it was quite a few hours ago Andrew

[12:47] Motor Loon: so they have been down for a while

[12:47] Vincent Nacon: we're lighter than gas, even with this new Mass function... we're still lighter than gas

[12:47] Kallista Destiny: Vincent, nice understatement.

[12:48] Vincent Nacon: eh?

[12:48] Liisa Runo: (btw, has anyone noticed that lately moving_end() seem to get triggered when prim cross region border? Not always but often)

[12:48] Vincent Nacon: what's to confuse with?

[12:49] Kallista Destiny: Scale scripts are full of adhoc constants to make weight computations give some approximation of reality

[12:49] Motor Loon: reality.... I've heard about that

[12:50] Andrew Linden: well, I can't figure out why the keyframe regions won't come up after a cursory glance. I'll have to dig into it later.

[12:50] Andrew Linden: I'll point out they are down on our #deploy channel...

[12:50] Vincent Nacon: maybe Falcon is still working on it?

[12:51] Andrew Linden: No, Falcon asked that they be brought up. The code was deployed but the regions won't start for some reason.

[12:51] Motor Loon: ..someone spilled coffee in the serverroom again...

[12:51] Vincent Nacon: did it have my suggestion in already? that'd be quick

[12:52] Vincent Nacon: the animation flag types

[12:52] Andrew Linden: dunno Vincent

[12:53] Kallista Destiny: Or the restroom upstairs had an overflow that dribbled onto the disk-farm

[12:53] Vincent Nacon: clean your room, Motor

[12:53] Motor Loon: I'll just clean the monitors ,-)

[12:54] Rex Cronon: because u r looking at a your lcd in a dark room. the fly is attrated to light

[12:54] Rex Cronon: attracted*

[12:54] Vincent Nacon: and soda spits

[12:54] Kallista Destiny: that too

[12:54] Vincent Nacon: correction.... beer spits

[12:54] Kallista Destiny: Or he's on Austrailai or anywhere in the Deep South of the Su

[12:55] Motor Loon: Why should be crapping on the lamps next to the monitors instead then...

[12:55] Kallista Destiny: US

[12:55] Kallista Destiny: Lamps too hot, flys not totally stupid.

[12:55] Vincent Nacon: anyway, 5 minutes left, any other topic?

[12:55] Meeter: Timecheck : User Group is almost over

[12:56] Rex Cronon: there r those stripes with glue on them that u hang from the ceiling. get some

[12:56] Rex Cronon: i have

[12:56] Motor Loon: Actually they are... 90% of them die after shitting on my LCD's ... getting pretty good at catching them in flight

[12:56] Rex Cronon: can llGetOnwerKey(uuid key) work for anything, not just prims?

[12:56] Latif Khalifa: Motor, I think we have heard quite enough who is crapping where in your place, thank you

[12:56] Vincent Nacon: well no

[12:56] Motor Loon: yw Latif

[12:57] Vincent Nacon: because it's assigned to prim it's in

[12:57] Rex Cronon: so, llgetownerkey?

[12:57] Vincent Nacon: err... oh wait wrong one

[12:57] Andrew Linden: no, I think the key argument to llGetOwnerKey() is for an object_id

[12:57] Rex Cronon: well. the sim has already a list of all the keys in it

[12:57] Pauline Darkfury: llGetOwnerKey won't work for AVs in certain types of relationship ;)

[12:57] Vincent Nacon: it would be interesting to find out who do I belong to if I used my key

[12:57] Rex Cronon: how hard is to look at all and not just prims?

[12:58] Liisa Runo: get a cat, super effective destroyer of everything that moves, like flies, also the best possible sleep aid, purrrrrr

[12:58] Motor Loon: lol Vincent

[12:58] Andrew Linden: all what? what kinds of things would you be looking for the owners?

[12:58] Kelly Linden: What do you want to use it for Rex?

[12:58] Kallista Destiny: Pauline, does that mean If I'm not married it won't work?

[12:58] Rex Cronon: i can get a list on animations that animate me, and i want to know who owns them

[12:59] Andrew Linden: hrm... where do the animations come from?

[12:59] Motor Loon: hm... that would be useful

[12:59] Andrew Linden: a dance ball you clicked? your attachments?

[12:59] Kallista Destiny: Umm I think that the Phoenix animator explorer can gt they for you

[12:59] Rex Cronon: deformers:(

[12:59] Vincent Nacon: uhh Rex, getting the key that would be the creator, not the owner

[12:59] Qie Niangao: except... how did you get a handle on the animation's key?

[12:59] Latif Khalifa: what would be much more useful is animation changed event, so AO's don't have to be in a thight loop checking what animation state avatar is

[12:59] Pauline Darkfury: The problem there is that multiple objects in the same region could own a copy of the same anim UUID, couldn't they?

[12:59] Rex Cronon: nope, each anim has an owner in that sim

[13:00] Andrew Linden: Pauline is right. There can be multiple "references" to an asset uuid

[13:00] Rex Cronon: qie, try it:)

[13:00] Meeter: Thank you for coming to the Server User Group

[13:00] Vincent Nacon: Qie, you can get it from AnimationList

[13:00] Andrew Linden: Oh. I have to go at 13:00 today. I have another meeting that starts now.

[13:00] Pauline Darkfury: Thanks, Andrew :)

[13:00] Vincent Nacon: seeya

[13:00] Sahkolihaa Contepomi: See you, Andrew.

[13:00] Andrew Linden: Thanks for coming everyone.

[13:00] Pauline Darkfury: Have a good one :)

[13:00] Rex Cronon: tc andrew

[13:00] Liisa Runo: thanks everybody

[13:00] Motor Loon: okies

[13:01] Vincent Nacon: you too

[13:01] Kallista Destiny: Thanks andrew

[13:01] Sahkolihaa Contepomi: And see you too, Kelly.

[13:01] Motor Loon: Have funs

[13:01] Latif Khalifa: thanks Andrew, Simon, Kelly :)

[13:01] Kallista Destiny: Thanks Kelly

[13:01] Pauline Darkfury: And thanks, Kelly, as well :)

[13:01] Qie Niangao: thanks Andrew, Simon, Kelly

[13:01] Qie Niangao: (I thought animation list was supposed to spit out null key for any anims I don't have full perms on)


Prev 2011.09.23 Next 2011.09.30