Chat log from the meeting on 2013-07-23

From OpenSimulator

Jump to: navigation, search
[10:59] Nebadon Izumi is online.
[10:59] Richardus Raymaker: wow empty here
[10:59] Richardus Raymaker: hi neb
[11:00] Justin Clark-Casey: hi folks
[11:00] Grid.Master @ Hi Justin
[11:00] Richardus Raymaker: hmm. ask it here to hi justin
[11:01] Nebadon Izumi: hello Justin
[11:01] Richardus Raymaker: kokua or wright. i cant get outy of fly mode when iim landing my avatar
[11:01] Nebadon Izumi: rut roh
[11:01] Nebadon Izumi: its the sim im sure
[11:01] Grid.Master @ Happened to me here... also ..when i came in.
[11:01] Justin Clark-Casey: I would guess something physics
[11:01] Grid.Master @ have to his stop flying...sighs... >>> Firestorm
[11:01] Richardus Raymaker: i think i can only stop when i press the button 'stop flying'
[11:02] Richardus Raymaker: yes master, but thats normal todo
[11:02] Andrew Hellershanks is online.
[11:02] Richardus Raymaker: yup. oly stop button works
[11:02] Nebadon Izumi: ya ok i let diva and radams1 know just now on IRC
[11:02] Nebadon Izumi: we will look into it
[11:03] Nebadon Izumi: is same thing happening on ODE do you know Richardus?
[11:03] Richardus Raymaker: i know lots of things ging on, juts spotted that one
[11:03] Nebadon Izumi: this region is running BulletSim
[11:03] Nebadon Izumi: wondering if its physics related or something related to what divas been doing
[11:03] Grid.Master @ ah.... ok
[11:05] Richardus Raymaker: not seen it on my older bullet sandbox. so... small guess.
[11:05] Nebadon Izumi: ok i'll test that later
[11:05] logger sewell: Hi all
[11:05] Nebadon Izumi: hello logger
[11:05] Nebadon Izumi: seems like light attendance today
[11:05] Grid.Master @ waves
[11:05] Nebadon Izumi: hope we do better at the load test
[11:05] Teravus Ousley is online.
[11:06] Nebadon Izumi: we might have trouble breaking 50 if more people dont show up
[11:06] logger sewell: it's summer time
[11:06] Nebadon Izumi: ya
[11:06] Teravus Ousley is offline.
[11:06] Frank Northmead: Hello everyone ?
[11:06] Andrew Hellershanks: That's no excuse ;-)
[11:06] Nebadon Izumi: hello
[11:06] Andrew Hellershanks: Afternoon, all.
[11:06] Vivian Klees: hello
[11:07] Andrew Hellershanks: or morning, all
[11:07] Grid.Master @ waves again
[11:08] Richardus Raymaker: hi andrew
[11:08] Richardus Raymaker: btw neb. since last problem im no running Kokua
[11:08] Teravus Ousley is online.
[11:08] Andrew Hellershanks: I made a few minor changes to ossearch this past week.
[11:08] Nebadon Izumi: nice, this was the event stuff?
[11:08] Andrew Hellershanks: I'm thinking that I need to make a minor change to the database structure to make it easy/easier to edit of existing events.
[11:08] Andrew Hellershanks: yup.
[11:09] Andrew Hellershanks: I created a page (not published) to create event listings.
[11:09] Nebadon Izumi: cool
[11:09] Andrew Hellershanks: It allowed me to create and delete events. Problem with editing them is the need to remember the parcel where the event was being held.
[11:10] Justin Clark-Casey coughs
[11:10] Andrew Hellershanks: I have the edit page but right now it would forget what parcel was selected.
[11:10] Nebadon Izumi: the whole viewer event thing is problematic for opensimulator
[11:10] Nebadon Izumi: good for small grids i guess
[11:10] Nebadon Izumi: but the time zone thing is very problematic
[11:10] Andrew Hellershanks: how is there a timezone issue?
[11:11] Nebadon Izumi: well when you have people all spread over the place regions in different time zones
[11:11] Richardus Raymaker: hi robert
[11:11] Andrew Hellershanks: event listings are stored in UTC date
[11:11] logger sewell: the search should pull from the grid server time
[11:11] Robert Adams: justin, I have an addition that allows reading all the StatManager data as JSON from the web.... the question is, how best can that be added to master
[11:11] Nebadon Izumi: the viewer doesnt display in UTC does it?
[11:11] Richardus Raymaker: depends wich viewere and how its set
[11:11] Robert Adams: there might need to be some security but I don't know what's appropriate
[11:12] Andrew Hellershanks: Events are stored in UTC format but the creation you just specify a time/date in grid time (whatever that happens to be). It defaults to Pacific time.
[11:12] Justin Clark-Casey: robert: yes, I would say definitely meed thought about security, that's one of my beefs with the last system
[11:12] Simulator Version v0.5 shouts: OpenSim 0.7.6 Dev          90528c2: 2013-07-23 01:13:13 +0100 (Unix/Mono)
[11:12] Dahlia Trimble is online.
[11:12] Andrew Hellershanks: I'd love to see event markers appear on the map. Melanie seems to think it isn't easy to do.
[11:13] Justin Clark-Casey: weren't you saying something about reusing the current auth system?
[11:13] Richardus Raymaker: Andrew, is the groups stuff a bit secure now if its running on differet server/network then the simulator ? i mean some protection so only ypour servers can use it ?
[11:13] Justin Clark-Casey: effectively having to request as a user??
[11:13] Andrew Hellershanks: Seeing land for sale was also an issue until Melanie provided some three or four lines for ini files that fixed that issue.
[11:13] Andrew Hellershanks: I'm hoping event listings might have a similar simple solution.
[11:13] Robert Adams: that's one way to go.... make it so you at least need to be a user on the grid
[11:14] Nebadon Izumi: great Andrew sounds like good progress
[11:14] Andrew Hellershanks: yup.
[11:14] Justin Clark-Casey: robert: that still makes me a bit queasy tbh but maybe that's just becvause I',m not used to the idea
[11:14] Justin Clark-Casey: robert: so this call would deliver stats info as one block of json?
[11:14] Robert Adams: with that solution, the displayer would log in, get a capability and then access the stats through that cap
[11:15] Justin Clark-Casey: I wouldn't be a fan of having to actually login
[11:15] Andrew Hellershanks: WHen you open the map, I see queries to ossearch looking for events. I know where in the code you send data that makes the viewer place markers on the map. The problem is to link the data coming back from the event listing request to the code that sends the marker data.
[11:15] Justin Clark-Casey: for most monitoring you don't want to be foreced to have a user actually logged in
[11:15] Robert Adams: the current implementation lets you specify category, container and stat to filter what you get back in the JSON
[11:15] Justin Clark-Casey: robert: how is it specified? via paramters or a RESTful approach?
[11:16] Richardus Raymaker: hi dahlia
[11:16] Dahlia Trimble: hi
[11:16] Frank Northmead: hello dahlia
[11:16] Robert Adams: http://serverHTTPport/StatManager?cat=category&cont=container&stat=statistic
[11:16] Grid.Master @ waves at Dahlia
[11:16] Andrew Hellershanks: Nebabdon, yes, I was pleased with how its gone so far. I want to make a change to the event creation page to use jQuery based selection tools for date and time and have one for the start and one for the end times instead of start time and duration. That will be a fairly simple change.
[11:17] Justin Clark-Casey: robert: how about GET StatManager/category or StatManager/Category./container or StatManager/Category/Container/Statistics ?
[11:18] Robert Adams: ick.... the stuff after the "?" is called the "query".... you can have several or none specified in any ordeer
[11:19] Justin Clark-Casey: robert: right, but the path approach is the RESTful one. I like the simplicity
[11:19] Richardus Raymaker: press stop flying
[11:19] Teravus Ousley: Landing on prim is broken.
[11:19] Richardus Raymaker: yup
[11:19] George Equus: Hi everyone
[11:19] Frank Northmead: landing is rboken, you need to stop flying
[11:20] Robert Adams: you don't want to break any bone when landing :)
[11:20] Teravus Ousley: This is a result of the collision plane work a few days ago. I warned diva to test this.
[11:20] Justin Clark-Casey: robert: Anyway, I presume if you request by container then you receive evreything structured in json as one would expect?
[11:20] Nebadon Izumi is offline.
[11:20] Justin Clark-Casey: one of my grips about the last system is that evreything was shoved into one uncategorizedbag...
[11:20] George Equus: tremendous lag... had to try three times to get here two of which i kept walking backwards... perpetually :(
[11:21] Nebadon Izumi is online.
[11:21] Dahlia Trimble: no lag for me
[11:21] Nebadon Izumi: lol woops
[11:21] Nebadon Izumi: closed wrong viewer
[11:21] Robert Adams: it is always returning {category: { container: {stat: { statInfo }}}}... just less stuff if you specify a filter
[11:21] Frank Northmead: Does anyone know if Melanie is going to have time to get the region crossing code into head/core before the conference?
[11:22] Justin Clark-Casey: there does seem to be a tremendous amount of in/out traffic
[11:22] Justin Clark-Casey: frank: why?
[11:22] Justin Clark-Casey: in/out udp traffic
[11:22] Frank Northmead: Just curious ?
[11:22] Justin Clark-Casey: well, it's being held up on the next release, which in turn is being held up by the very high volume of current changes
[11:22] Nebadon Izumi: no i would expect not Frank
[11:22] Robert Adams: I volunteered to maintain a branch with any crossing changes she might make
[11:23] Robert Adams: (keep it rebased to master)
[11:23] Justin Clark-Casey: Melanie does not seem to like branches
[11:23] Nebadon Izumi: however some of the stuff we are doing now in theory could improve border crossing
[11:23] Nebadon Izumi: we have not really been testing that though
[11:23] Robert Adams: that would allow you (justin) to wait for a settled point to get a release while region crossing improvements were made
[11:23] Frank Northmead: okay, it is probably best to not stir the pot anymore, the current performance enhancements are really awesome and more than enough pot stirring for a while ?
[11:23] Richardus Raymaker: cool.
[11:24] Justin Clark-Casey: I would love a period of stability
[11:24] Justin Clark-Casey: but have to make big changes to make the conf possible
[11:24] Nebadon Izumi: Melanies_T time has been limited lately and I cant really say for sure she will have a lot of extra time between now and the conference
[11:24] Frank Northmead: understand
[11:24] Andrew Hellershanks: You are hoping the next release will be out before the conference?
[11:25] Nebadon Izumi: hopefully
[11:25] Andrew Hellershanks nods
[11:25] Justin Clark-Casey: I really hope so
[11:25] Frank Northmead: Is Wright Plaza at head? It feels a little laggy?
[11:25] Nebadon Izumi: it is at head
[11:25] Frank Northmead: :(
[11:25] Nebadon Izumi: were using very low CPU
[11:25] Nebadon Izumi: this region has a crap ton of content
[11:25] Justin Clark-Casey: what do you mean by feels a little laggy
[11:25] Richardus Raymaker: it feels ok for me
[11:25] Nebadon Izumi: if your feeling lag its probably your viewer
[11:25] Renato Schnyder: anyone can send me a ladmark for CRAFT
[11:25] Justin Clark-Casey: some of the agentupdate restrictions might generate that feel
[11:26] Frank Northmead: feels okay now, I had my draw distance at 1024 - oops ?
[11:26] Richardus Raymaker: i think loading lag.
[11:26] Justin Clark-Casey: lot of asset requests going on though, which is generating this lag
[11:26] Nebadon Izumi: ya thats viewer lag
[11:26] Richardus Raymaker: and thats not only opensim problem
[11:26] Nebadon Izumi: this region alone has like 4000 textures
[11:26] Nebadon Izumi: thats not including the region next door with all of the photography
[11:26] Richardus Raymaker: how can you see the amount of textures nebadon ?
[11:26] Justin Clark-Casey: do you know if those asset requests flow through this region?
[11:26] Justin Clark-Casey: I hope not...
[11:27] Nebadon Izumi: dahlia does counts with her viewer testing :)
[11:27] Nebadon Izumi: hmm I am not sure Justin
[11:27] Nebadon Izumi: seems unlikely they do
[11:27] Justin Clark-Casey: yeah, we are seieng quite a few GetTexture requests atm
[11:27] Nebadon Izumi: ya im sure its the textures in this region
[11:27] Nebadon Izumi: as people log in
[11:27] Nebadon Izumi: this region has in insane amount of textures
[11:28] Nebadon Izumi: even by SL standards
[11:28] Nebadon Izumi: haha
[11:28] Justin Clark-Casey: might need to look at throttling asset requests
[11:28] Nebadon Izumi: ya i think we talked about that previously
[11:28] Nebadon Izumi: probably a good thing
[11:28] Justin Clark-Casey: yes
[11:28] Nebadon Izumi: question
[11:28] Nebadon Izumi: are assets throttled for IARs and OARs?
[11:28] Nebadon Izumi: we should probably consider that as well if its not
[11:28] Justin Clark-Casey: saving?
[11:28] Nebadon Izumi: yes
[11:29] Justin Clark-Casey: the archives do only request one at a time
[11:29] Nebadon Izumi: iars in particular
[11:29] Justin Clark-Casey: but are not otherwise throttled
[11:29] Nebadon Izumi: k
[11:29] Justin Clark-Casey: that's a lot better than viewers which aren't restricted to one request at a time :)
[11:29] Nebadon Izumi: oars are not as bad cause genearlly the content is cached
[11:29] Nebadon Izumi: but my inventory
[11:29] Nebadon Izumi: my gigantic inventory seems to be pretty harsh
[11:29] Justin Clark-Casey: it really busts cpu?
[11:29] Richardus Raymaker: assets throttling, can avoid router flood and jamming the connection. if iam right.
[11:30] Nebadon Izumi: i'll have to try again its been a while since i tried honestly
[11:30] Teravus Ousley wonders if Nebadon knows the size of the full IAR of OSGrid->Nebadon Izumi
[11:30] Richardus Raymaker: onl would set the number of requestion in config
[11:30] Nebadon Izumi: about 2gb Teravus
[11:30] Teravus Ousley: hehe
[11:30] Justin Clark-Casey: throttling would make it take even longer though :)
[11:30] Richardus Raymaker: only 2GB ?
[11:31] Nebadon Izumi: its not the amount of time i am concerned with
[11:31] Vivian Klees: not including all the alts
[11:31] Nebadon Izumi: im more concerned with load on simulator and robust
[11:31] Teravus Ousley: Might want to start putting nebadon's inventory on a storage cloud
[11:31] Justin Clark-Casey: nebadon: one thing that might help is to make the cache limited by size instead of what's in the region
[11:31] Nebadon Izumi: that 1 person creating iar can create
[11:31] Justin Clark-Casey: nebadon: I think with the settings you have, any assets downloaded to create an iar get wiped when the cache is refreshed
[11:31] Justin Clark-Casey: whereas it may make a lot more sense to keep much of that data cached - disk is cheap
[11:32] Nebadon Izumi: large caches do bad though
[11:32] Justin Clark-Casey: how so?
[11:32] Nebadon Izumi: ive noticed when your cache exceeds around 100,000 items
[11:32] Nebadon Izumi: it can get wierd and slow
[11:32] Marcus Llewellyn: At least as things stand right now, since only a SimOp can initiiate an IAR save/restore, you can choose a time when the simulator isn't otherwise busy to perform a backup.
[11:32] Justin Clark-Casey: that may be a factor of the implementation than any intrisic reason
[11:32] Teravus Ousley: What's the limiting factor? Network IO? Physical IO? CPU?
[11:32] Andrew Hellershanks: nebadon, that might have something to do with the file system of the machine running the software.
[11:32] Nebadon Izumi: not sure ive never looked that closely at it
[11:32] Nebadon Izumi: im mostly going on what i see in world happening
[11:32] Justin Clark-Casey: I imagine handling 100k files on thre system
[11:33] Justin Clark-Casey: in a naive fashion instead of maintaing an index or similar (guessing)
[11:33] Nebadon Izumi: ya hard to say, my findings are very unscientific
[11:33] Nebadon Izumi: and mostly about personal experience
[11:33] Justin Clark-Casey: lol
[11:33] Justin Clark-Casey: I know I keep banging on about unscientific :)
[11:34] Justin Clark-Casey: hoping for a nirvana that isn't realistic
[11:34] Nebadon Izumi: haha
[11:34] Teravus Ousley: would be useful to run a resource monitor to try and identify the limiting factor.. so we can better come up with a reasonable solution for the generalized issue of massive inventory.
[11:34] Nebadon Izumi: ya
[11:34] Nebadon Izumi: i mean my inventory reads 45,000 items
[11:34] Nebadon Izumi: but just think about all the content inside of those objects
[11:35] Nebadon Izumi: scripts and boxes of textures and stuff
[11:35] Nebadon Izumi: its 100,000's of items total
[11:35] Justin Clark-Casey: hell of a lot of chained objects
[11:35] Nebadon Izumi: ya
[11:35] Justin Clark-Casey: at the end of the day, you're looking at a *lot* of data
[11:35] Nebadon Izumi: ya untar/zipped my inventory is over 4gb
[11:35] Richardus Raymaker: thats why 2GB sounds so small nebadon. mine where already bigger
[11:36] Justin Clark-Casey: nebadon: I think better caching has to be part of that improvement
[11:36] Richardus Raymaker: aaah. ok thats better
[11:36] Nebadon Izumi: ive cleaned my inventory on many occasions
[11:36] Justin Clark-Casey: but first time is going to be tough
[11:36] Nebadon Izumi: ya
[11:36] Nebadon Izumi: there has been minimal time spent on flotsam
[11:36] Justin Clark-Casey: in theory, you could run a completely separate proces for doing that
[11:36] Justin Clark-Casey: something which contacts the asset/inventoryt service etc.
[11:36] Nebadon Izumi: one thing i noticed though with Flotsam
[11:36] Nebadon Izumi: we did a test here last week and we had Memory and Disk cache enabled
[11:36] Nebadon Izumi: simultaneously
[11:37] Nebadon Izumi: and the region seemed to start up faster
[11:37] Nebadon Izumi: interestingly enough
[11:37] Nebadon Izumi: ot ran pretty good, we ended up disabling it because i was unsure about excessive memory use
[11:37] Richardus Raymaker: i would say, opensim puts data first in memory cache. and later it moves it to disk
[11:37] Justin Clark-Casey: ineresting, memory cache shouldn't make much difference to that
[11:37] Nebadon Izumi: but i do plan on revisiting that
[11:37] Justin Clark-Casey: I haven't looked a lot at the memory side of flotasm but afaik it's quite a naive imple,mentation
[11:38] Justin Clark-Casey: I think it does obey whatever limit you set it, though?
[11:38] Nebadon Izumi: i was always afraid to run flotsam memory cache because of leaks and instability
[11:38] Richardus Raymaker: never looked at memory cache. we already now have never enough memory
[11:38] Nebadon Izumi: but it ran here at Wright Plaza for 3-4 days without any problems
[11:38] Teravus Ousley: Individual user inventories are not highly cacheable.. however, caching web proxies may improve performance. Also dedicating more resources to serving the individual inventory data requests may improve performance assuming that not all of the inventory resources are being served from a single physical machine.. and that there are potentially multiple copies of the data. :)
[11:39] Justin Clark-Casey: I believe osgrid runs multiple inventory service copies and load balances via nginx
[11:39] Justin Clark-Casey: right neb?
[11:39] Nebadon Izumi: right
[11:39] Nebadon Izumi: 5 inventory servers
[11:39] Nebadon Izumi: on round robin via nginx
[11:40] Teravus Ousley: One of the issues, I thought was the immediacy of the response to the request before the viewer decides to give up. Is that still an issue?
[11:41] Justin Clark-Casey: to an inventory request?
[11:41] Teravus Ousley: Yes
[11:41] Justin Clark-Casey: I don't remember that personally. I know there is na issue where somteimes the viewer does not keep downloading inventory but stops with unfulfilled requests
[11:42] Nebadon Izumi: inventory has actually been working pretty good for me lately
[11:42] Justin Clark-Casey: whether that has changed I don't know
[11:42] Justin Clark-Casey: there were afew tweaks recencly
[11:42] Nebadon Izumi: my concern was with IAR saving mostly
[11:42] Teravus Ousley: Could we potentially put all inventory item requests in a queue and respond to the viewer as they get processed so as not to have an inventory resource free-for all on the individual region?
[11:42] Nebadon Izumi: it seems to have a high impact on the region and robust
[11:42] Nebadon Izumi: higher than i would like to see
[11:42] Justin Clark-Casey: teravus: inv request are now being throttled via a modified http polling
[11:43] Nebadon Izumi: it really takes a long running big inventory like my own to notice it
[11:43] Teravus Ousley: Ok. I've seen that code.
[11:43] Justin Clark-Casey: I think it's the fact that asset requests remains a free for all which is causing some issues
[11:43] Teravus Ousley: It's UBER threaded.
[11:43] Justin Clark-Casey: like lag seen here whgen people login, some of which is almost certainly due to that
[11:43] Richardus Raymaker: whats long, im used to a few hours from the p[ast. just turn it on and go slepe :)
[11:43] Teravus Ousley: .. so.. Lots of asset requests = resource free-for-all
[11:43] Nebadon Izumi: i mean personally i wouldnt care if my inventory took 24 hours to make an IAR
[11:44] Nebadon Izumi: if the impact to the grid and region was near zero
[11:44] Justin Clark-Casey: teravus: yeah
[11:44] Justin Clark-Casey: nebadon: not all may agree
[11:44] Justin Clark-Casey: perhaps there could be a throttle option
[11:45] Nebadon Izumi: but hey if there is a way to keep it fast but also have zero impact
[11:45] Nebadon Izumi: im also all for that
[11:45] Justin Clark-Casey: lol
[11:45] Nebadon Izumi: i dont think anyone wants to see overall services degrade because of it
[11:45] Justin Clark-Casey: I think to make it nice (at least the first time) there is a cpu/time tradeoff
[11:45] Nebadon Izumi: im not even sure its high cpu
[11:45] Nebadon Izumi: that is the problem
[11:45] Nebadon Izumi: i think like you said its maybe too many requests being allowed to happen
[11:46] Vivian Klees: what about a priority of system folders being the first to load
[11:46] Nebadon Izumi: the highest impact seems to be on asset services
[11:46] Justin Clark-Casey: requests are only made one at a time, not concurrently
[11:46] Justin Clark-Casey: so it shouldn't be that
[11:46] Justin Clark-Casey: however, that did change some time back, so maybe you tested before that change?
[11:46] Nebadon Izumi: hard to say
[11:46] Nebadon Izumi: i can try again soon
[11:47] Justin Clark-Casey: yeah, would be good to knwo
[11:47] Justin Clark-Casey: that area has changed,. though maybe a year or so ago now
[11:47] Justin Clark-Casey: or maybe not quite that long, but in that region
[11:47] Nebadon Izumi: ah ya its been less than a year since i tried
[11:48] Nebadon Izumi: its good we have a good handle on all requests though
[11:49] Nebadon Izumi: I think all of this stuff we are doing now is long over due :)
[11:49] Justin Clark-Casey: we are getting better
[11:49] Nebadon Izumi: ya
[11:49] Nebadon Izumi: the improvements in past few weeks have been enourmous
[11:49] Nebadon Izumi: this region right now is using less than 50% cpu
[11:49] Nebadon Izumi: thats amazing
[11:50] Nebadon Izumi: 22.6% cpu is reading aprox
[11:50] Nebadon Izumi: with 21 avatars
[11:50] Grid.Master @ sweet
[11:50] Justin Clark-Casey: yeah, not too bad at steady state. I am surprised at the amount of inbound udp
[11:50] Richardus Raymaker: thats good
[11:50] Nebadon Izumi: im not
[11:50] Justin Clark-Casey: but that's probably quite normal, just havent' seen it before
[11:50] Nebadon Izumi: the viewer is a greedy bastard
[11:50] Nebadon Izumi: lol
[11:51] Justin Clark-Casey: 700 receives per second?
[11:51] Nebadon Izumi: ya LL seems to have taken the we dont care what the viewer does approach
[11:51] Frank Northmead: Justin's typing animation appears to stutter to me, does anyone else see that?
[11:51] Andrew Hellershanks: Nebadon, you'll be taming a viewer next? ;)
[11:51] Nebadon Izumi: and rely on their simulators to control everything
[11:51] Justin Clark-Casey: although I am now wondering it I've screwd up this stat
[11:51] Dahlia Trimble: viewers send a *lot* of AgentUpdate packets
[11:51] Nebadon Izumi: I think the next thing we need to focus on after teleports are smoothed
[11:52] Nebadon Izumi: is log ins
[11:52] Nebadon Izumi: and the big cpu spikes
[11:52] Justin Clark-Casey: high instant numbers aren't pulling up the average
[11:52] Nebadon Izumi: though some of those spikes might get smoothed out with teleports too
[11:52] Robert Adams: I've always wondered why the standard viewer generated so much traffic.... LL must pay for bandwidth so reducing same would save big bucks
[11:52] Teravus Ousley: 10x21 = 210 AgentUpdatePackets/sec
[11:52] Nebadon Izumi: since the #s are similar
[11:52] Justin Clark-Casey: nebadon: I have a feeling it is to do with asset requets
[11:52] Nebadon Izumi: ya
[11:52] Dahlia Trimble: AgentUpdate packets are pretty small
[11:52] Nebadon Izumi: im sure getting that under control will be a big boost
[11:52] Teravus Ousley: They are small :D
[11:53] Richardus Raymaker: sure LL dont buye flatrate ?
[11:53] Frank Northmead: we also still have the UserUMMTGUN7
[11:53] Richardus Raymaker: good one frank. how is that justin ?
[11:53] Teravus Ousley cleared his cache from the last time.... so will be able to asses this time on the KN load test
[11:54] Justin Clark-Casey: tbh, I think that may be resolved
[11:54] Justin Clark-Casey: but it will only go away once all viewers have cleared cache and all grids have updated to very recent changes
[11:54] Frank Northmead: hmm, okay, I am at head, but have not cleared cache recently
[11:54] Justin Clark-Casey: I haven't been seeing any UMMTGUN7, even in load tests
[11:55] Frank Northmead: I am seeing one here now.
[11:55] Teravus Ousley: I have not heard anything about UMMOUGUN993 in a while so I figured it had probably been fixed even though I hadn't bothered to clear my cache in a while :)
[11:55] Frank Northmead: GUN7 said "AgentUpdate packets are pretty small" earlier for me.
[11:56] Justin Clark-Casey: I thin kthere comes a point where everyone starts ignoring it so you stop hearing about it :)
[11:56] Teravus Ousley: Frank, yes. It will happen if your viewer has that UUID->Avatar name
[11:56] Teravus Ousley: .. cached
[11:56] Frank Northmead: okay, after the load test I will clear my cache.
[11:56] Dahlia Trimble: I am NOT a GUN7 :P
[11:56] Justin Clark-Casey: ahh duuuuh
[11:56] Teravus Ousley: doesn't matter when it cached it but if it did and you haven't cleared your cache.. it will remain a GUN
[11:56] Justin Clark-Casey: all instance samples are displaying 3 times higher tha nthey should
[11:56] Teravus Ousley: Yeah.. Dahlia is a GUN to Sr Frank
[11:56] Dahlia Trimble: :(
[11:56] Marcus Llewellyn: I've seen GUN7's recently aftera cleared cache on one of my regions. It was an NPC I always have running. The console said something like it could find a bound name.
[11:57] Marcus Llewellyn: couldn't, rather
[11:57] Justin Clark-Casey: so we're seeing 250 inbound rather than 750
[11:57] Teravus Ousley: That meshes with the AgentUpdate info
[11:57] Dahlia Trimble: everyone go to low graphics mode so AgentUpdate packets increase :)
[11:57] Justin Clark-Casey: marcus: ok - could you file a but report? It should have a bound name really
[11:58] Marcus Llewellyn: Sure.
[11:58] Justin Clark-Casey: you can also do "show name <uuid>" on the console to check the simulator records directly
[11:59] Teravus Ousley: Yes, landing fail. I already chatted with diva about it :)
[11:59] Marcus Llewellyn adds a sticky note to the side of his tower. :)
[11:59] Nebadon Izumi: anyone else have anything they want to talk about before we start heading over to CC grid?
[11:59] Robert Adams: time to kit the keynote regions?
[11:59] Robert Adams: hti
[11:59] Robert Adams: hit
[12:00] Robert Adams: sheesh
[12:00] Nebadon Izumi: yea were at to of the hour
[12:00] Nebadon Izumi: if no one else has anything we can start heading over
[12:00] Kayaker Magic: I'd like to keep the \93moving a seated avatar\94 issue warm, with this observation:
[12:00] Marcus Llewellyn: We're spreading out for this one, yes? Not all in KN1?
[12:00] Teravus Ousley: For the longest time, the collision plane was a static plane defined that moved when the avatar landed so the viewer will interpret the landing
[12:00] Kayaker Magic: Lots of people in OpenSim use a posing script called MLP (Magic Love Pose? But used in regular chairs)
It is a total hack, you right-click, touch, select a pose from a menu, it rezzes a pose-ball (only works in public rez areas) then you sit on the ball.
All just so you can adjust your position after you sit.
[12:00] Frank Northmead: address please
[12:00] Kayaker Magic: I made some chairs that you sit on in the 'usual' way, by left clicking. People used to MLP can't get it to work. MPL is has trained them to right-click and they have forgoten that you can just left-click on a CLICK_ACTION_SIT object.
[12:00] Marcus Llewellyn: 1
[12:00] Dahlia Trimble: load test? uri please?
[12:00] Teravus Ousley: At some point either myself or Jhurliman or both did a raycast against the ground to determine where the plane should actually be
[12:00] Nebadon Izumi: ok Kayaker that certianly is an important note
[12:00] Justin Clark-Casey: kayaker: yeah sorry, perf work is blowing eveything else out of the water atm, at least for me
[12:01] Justin Clark-Casey: I know this issue has been around for a looong time
[12:01] Teravus Ousley: anyway.. :) Happy Hypergridding TO KN
Personal tools
About This Wiki