Chat log from the meeting on 2014-06-24

From OpenSimulator

Jump to: navigation, search
[10:59]  Andrew Hellershanks is Online
[10:59]  Connecting to in-world Voice Chat...
[11:00]  Connected
[11:00]  Richardus Raymaker: hi andrew
[11:00]  Connecting to in-world Voice Chat...
[11:00]  Connected
[11:00]  Tiffany Magic: Hi Justin
[11:00]  Richardus Raymaker: hi justin
[11:00]  Justin Clark-Casey: hi tiffany, folks
[11:00]  Shez Oyen: Justin :)
[11:01]  Mata Hari: hi Justin
[11:01]  Tiffany Magic falls out of my chair and slowly climbs back up and sits down.... Justin is in regular clothes and has hair???? I haven't been here in awhile. *grin*
[11:01]  Richardus Raymaker: typ, its now hard te see when juistin is here
[11:02]  Justin Clark-Casey: I had to get dressed for something... ah yes the vwbpe talk :)
[11:02]  Andrew Hellershanks: Hey, Richardus
[11:02]  Andrew Hellershanks: yea, quick take a picture. Justin is wearing a regular set of clothing. :)
[11:02]  Richardus Raymaker: why does HG still not show maps from the grid you visit ?
[11:03]  Justin Clark-Casey: I expect it looks at the wrong place
[11:03]  Mata Hari: I ran into something earlier today that I think is a bug but I'm not 100% sure because I've never done it in SL......if you have group permissions set to allow group members to edit objects in a region it seems to allow it just fine, but if you add a script to an object no other group member can edit it even if you set the script perm with the "share group" flag
[11:03]  Justin Clark-Casey: might be difficult to fix without more explicit hg support in viewers
[11:03]  Mata Hari: they can still edit the rest of the object though
[11:04]  Justin Clark-Casey: I don't know either, I haven't tried that in sl
[11:04]  Sarah Kline: hi
[11:04]  vegaslon plutonian: yes for group members to edit a script they must also have edit rights to your prims
[11:04]  Justin Clark-Casey: hi sarah
[11:04]  Mata Hari: my guess is it's a bug
[11:04]  Richardus Raymaker: hi sarah
[11:04]  Andrew Hellershanks: Interesting. I haven't often used the group settings.
[11:04]  Mata Hari: (otherwise what would the "share group" flag function be?)
[11:05]  Richardus Raymaker: othe rproblem, fix-phantom on mega seems not to work. but who want to use a mega. not me
[11:05]  acryline erin is Online
[11:05]  Mata Hari: I was trying to make it painless for a group of people to work simultaneously on a region
[11:05]  vegaslon plutonian: you have to set "alot" of stuff for someone to get to a script and even in secondlife only works half the time
[11:05]  Richardus Raymaker: and the last one, can be viewer problem. in singularity on var region the coordinates you see are wrong. (still 256m based)
[11:06]  Nebadon Izumi is Online
[11:06]  Richardus Raymaker: at least the viewer still thinks it need to display normal sims and coordinates. but TP wokrs fine. except the z direction
[11:06]  vegaslon plutonian: ultimately easier to pull the script from the object so it goes next user, then make your edit and delete the old script and stick in the new
[11:06]  Mata Hari: I set the script and the object to essentially full perm.....was trying everything I could think of to allow someone else to edit the script
[11:06]  Nebs Metal Bar Stool v1.5 (w/sit & launch): Hello Nebadon Izumi, enjoy your sit..
[11:07]  Richardus Raymaker: oh thats not dust but nebadon
[11:07]  Nebadon Izumi: hello everyone, sorry im late, was in the blender zone
[11:07]  Mata Hari: the only way they could was to enter god mode, force ownership change and then edit it
[11:07]  Nebadon Izumi: lol
[11:07]  Richardus Raymaker: hi nebadon
[11:07]  Nebadon Izumi: or should i say Blender Tunnel
[11:07]  Nebadon Izumi: hehe
[11:07]  Mata Hari: and funny thing is even after forcing owner if you leave god mode you can't edit the script any longer
[11:07]  Nebadon Izumi: ya my cache was cleared
[11:07]  Nebadon Izumi: give me a minute to rez up
[11:08]  Richardus Raymaker: yeah kokua have the same map coordinate bug
[11:08]  Richardus Raymaker: hi robert
[11:09]  Richardus Raymaker: do you know if the sim send map coodinates to the viewer map ?
[11:09]  Robert Adams: hi all
[11:09]  Richardus Raymaker: the coordinates show wrong on var\
[11:09]  Mata Hari: hi Robert
[11:09]  Andrew Hellershanks: Richardus, only z is wrong?
[11:09]  Richardus Raymaker: open map, click on var. and you see
[11:09]  vegaslon plutonian: ya diva just fixed that in master
[11:09]  Richardus Raymaker: z did go wtong when i tried to TP up to skybox manual.
[11:10]  Richardus Raymaker: but thgats because x and y are wrong to in map coordinates
[11:10]  Nebadon Izumi: also with divas fix
[11:10]  Andrew Hellershanks: odd. Z would be the one part of the coord I would expect should be unaffected by varregion.
[11:10]  Nebadon Izumi: if you log into an old region and then teleport to new region
[11:10]  Nebadon Izumi: it will still be broken on new region
[11:10]  Nebadon Izumi: unfortunately, i learned that the hard way
[11:10]  Justin Clark-Casey: tp broken?
[11:11]  Nebadon Izumi: no just map coordinates will be wrong
[11:11]  Nebadon Izumi: you have to log into a new region
[11:11]  Justin Clark-Casey: as displayed on the map?
[11:11]  Justin Clark-Casey: oh, so you can't actually target it to tp?
[11:11]  Nebadon Izumi: ya the problem is if you search for a region from a old not updated sijmulator
[11:11]  Nebadon Izumi: it caches those results
[11:11]  Nebadon Izumi: so even when you arrive at the new simulator it will still seem broken
[11:11]  Nebadon Izumi: because results from the older simulator
[11:11]  acryline erin is Offline
[11:12]  Richardus Raymaker: also it seems teleport is still many times broken until the var is loaded terrain
[11:12]  Nebadon Izumi: not much we can do about that other than encourage people to upgrade soon
[11:12]  vegaslon plutonian: that is a task
[11:12]  Justin Clark-Casey: ho hum
[11:13]  Nebadon Izumi: however
[11:13]  Nebadon Izumi: people still using 0.2 would be denied anyway
[11:13]  Nebadon Izumi: so its not as bad as it really sounds
[11:13]  Richardus Raymaker: lol
[11:13]  Richardus Raymaker: where at 0.3 now ?
[11:13]  Andrew Hellershanks: If a .version file was created automatically from a standard build people would at least know when they are TP'ing between regions running different code.
[11:13]  Nebadon Izumi: for the simulation version yes
[11:14]  Nebadon Izumi: its in Grid.ini I think
[11:14]  Nebadon Izumi: so anyone updating regularly doesnt have to change anything
[11:14]  Nebadon Izumi: it happens automatically
[11:14]  Justin Clark-Casey: it should only be an issue if someone has overriden it manually, for some reason
[11:15]  Nebadon Izumi: ya thats possible but good administration practices should avoid things like taht
[11:15]  Nebadon Izumi: this is the setting
[11:15]  Nebadon Izumi: ConnectorProtocolVersion = "SIMULATION/0.3"
[11:15]  Nebadon Izumi: in Grid.ini
[11:15]  Nebadon Izumi: if for some reason your not updating your Grid.ini
[11:15]  Nebadon Izumi: you really should be
[11:15]  Justin Clark-Casey: I guess it's possible people are reusing an old Grid.ini
[11:16]  Nebadon Izumi: possible but probably low occurance of that
[11:16]  Justin Clark-Casey: which may show the value of having defaults in actual code rather than always explicit in config
[11:16]  Nebadon Izumi: those that are probably know to update regularly i would hope
[11:16]  Justin Clark-Casey: I mean, explicitly set
[11:16]  Nebadon Izumi: eh
[11:16]  Nebadon Izumi: if someone wants to change it they will just change it there
[11:16]  Nebadon Izumi: and then its even more work
[11:17]  Richardus Raymaker: need to check that line
[11:17]  Justin Clark-Casey: no
[11:17]  Justin Clark-Casey: I mean, have it commented out rather than set
[11:17]  Justin Clark-Casey: anyway, it's an esoteric point, not really worth the time to try and work through
[11:17]  Richardus Raymaker: in the past i played once with the setting. so need to check if its active still
[11:18]  Richardus Raymaker: not solveing invisible maptiles after HG i think
[11:18]  Nebadon Izumi: ah i see what you mean
[11:19]  Robert Adams: probably shouldn't bring up the topic of redoing the configuration files into something better ;-)
[11:19]  Justin Clark-Casey: that way, a change in opensim itself would not be overriden by somebody's copied Grid.ini
[11:19]  Justin Clark-Casey: robert: heh, like what?
[11:19]  Richardus Raymaker: all database settings in one file.
[11:19]  Justin Clark-Casey: although I guess that would almost negate the poitn of grid.ini
[11:20]  Robert Adams: and a better organization of config-includes so one can override settings there...
[11:20]  Justin Clark-Casey: honestly, so much stuff there needs to be improved
[11:20]  Justin Clark-Casey: shoving everything in bin/ is just brain-dead, but changing it means working through all the protests about changing any of the laout
[11:20]  Andrew Hellershanks: Richardus, I would like to see the code verify the db settings are valid when they are specified instead of waiting until a DB connection is attempted and get an odd failure/exception as a result that doesn't mention a problem with the connection string.
[11:20]  Robert Adams: my configuration is to not modify any released ini files and override the parameters with files in bin/config directory
[11:21]  Robert Adams: but that doesn't work with all the stuff in config-include
[11:21]  Richardus Raymaker: a check at startup would be good idea. and then start opensim
[11:21]  Richardus Raymaker: the only change i really made is seperate vivox file
[11:21]  Andrew Hellershanks: Richardus, yea. In the early days I had issues because I forgot to set a string or messed it up.
[11:22]  Richardus Raymaker: before notepad++ i guess :)
[11:22]  Robert Adams: oh, and I noticed that a few months ago, the Nini parameter inclusion stuff was disabled again
[11:22]  George Equus: Evening everyone
[11:22]  Richardus Raymaker: i got some EOC error at the point warp3d where created on var ..
[11:22]  Andrew Hellershanks: Afternoon, George.
[11:22]  Richardus Raymaker: hello george
[11:22]  Andrew Hellershanks: Does Warp3D handle vars?
[11:23]  Casias Falta: hey eq
[11:23]  Justin Clark-Casey: warp3d is a memory leaking mess
[11:23]  Richardus Raymaker: seems to look fine. a bit grainy and some terrain parts are weird colored but it works andrew\
[11:24]  Richardus Raymaker: it still is ?
[11:24]  Andrew Hellershanks: Justin, yea, heard that before. It also still has that off by one error in the right side of the tiles.
[11:24]  Richardus Raymaker: maby under linux
[11:24]  Andrew Hellershanks: Richardus, ok.
[11:24]  logger sewell: hey all
[11:24]  Justin Clark-Casey: hi logger
[11:24]  Andrew Hellershanks: Hey, logger.
[11:24]  Richardus Raymaker: hi logger
[11:24]  George Equus: Hi
[11:25]  Mata Hari: I've had some serious issues with HG tping between here and AVW recently....essentially I have to Ruth myself and remove all attachments before making the TP and then wear everything again once I'm there.....else I'm invisible....Bob (who runs AVW) thinks it's because OSG has his old url cached somewhere but I can't see how that would be causing the issue
[11:25]  Andrew Hellershanks: Is there some equivalent to valgrind for mono/Windows that would help track down reference counting issues?
[11:26]  Justin Clark-Casey: I know where it is - it's because whoever wrote warp3d didn't release the graphics objects used
[11:26]  Justin Clark-Casey: or dispose of them rather
[11:26]  Justin Clark-Casey: I fixed exactly the same issue with the normal map generation but haven't had time to do it in warp3d, where it's somewhat more complex to do
[11:26]  Andrew Hellershanks: Justin, file a mantis and point out where the problem is and someone may jump in and fix it.
[11:26]  Andrew Hellershanks: :)
[11:27]  Justin Clark-Casey: maybe
[11:27]  Robert Adams: adding a whole mini-renderer to OpenSim core to make map tiles is not a good idea in general... Warp3D needs to be taken out of the simulator and something else should be done
[11:27]  Justin Clark-Casey: robert: like what?
[11:27]  Andrew Hellershanks: If warp3d could be run ins some sort of standalone configurtion it would make it easier to test.
[11:27]  Richardus Raymaker: justin thats makeing a mantis. he possible need to fix itself at the end :)
[11:28]  Dahlia Trimble is Online
[11:28]  Justin Clark-Casey: it's easy enough to test. Do continual manual regreneration of warp 3d tiles and watch memory usage permanently increase
[11:29]  Cuteulala Artis is Online
[11:29]  Andrew Hellershanks: Robert: I've also thought about whether map tile generation could be run as a separate process outside of the main grid code. The problem is that it needs all the information about what is in a region.
[11:30]  Nebadon Izumi: whats wrong with warp3D?
[11:30]  Nebadon Izumi: you would still need something to render tiles
[11:30]  Nebadon Izumi: SL does tiles very differently
[11:30]  Dahlia Trimble: hi
[11:31]  Nebadon Izumi: they have a crawler, basically a modified viewer that crawls the grid
[11:31]  Justin Clark-Casey: hi dahlia
[11:31]  Richardus Raymaker: hi dahlia
[11:31]  Nebadon Izumi: and basically renders a tile the same way the viewer would
[11:31]  Andrew Hellershanks: nebaon, right. That is the other way to do it.]
[11:31]  Robert Adams: the separate process can just connection to the region... if the map generator can have a 'mapping' account... it could just log in and fetch the region data
[11:31]  Nebadon Izumi: that feels heavy to me
[11:31]  Robert Adams: libomv could provide the connection
[11:31]  Andrew Hellershanks: Nebadon, sounds like a job for someone who knows the viewer code.
[11:32]  Nebadon Izumi: i would be against using the viewer myself
[11:32]  Nebadon Izumi: its way to heavy
[11:32]  Dahlia Trimble: I thought SL used a hacker up viewer to make their map tiles
[11:32]  Nebadon Izumi: would require a dedicated machine
[11:32]  Nebadon Izumi: they do dahlia
[11:32]  Richardus Raymaker: not option for opensim i think
[11:32]  Nebadon Izumi: no i dont think so either honestly
[11:32]  Dahlia Trimble: why not?
[11:32]  Nebadon Izumi: wouldnt it require a GPU?
[11:32]  Dahlia Trimble: could be an option for a larger grid
[11:32]  Richardus Raymaker: for small grid or standalone it would be running on same system as server
[11:33]  Nebadon Izumi: cant expect grids to have a machine like that to make map tiles
[11:33]  Richardus Raymaker: and thats to heavy
[11:33]  Nebadon Izumi: ya you cant run a viewer from command line either
[11:33]  Dahlia Trimble: I know how to write a renderer but no way do I have time to do it
[11:33]  Nebadon Izumi: I dont understand why having the region do it is bad?
[11:34]  Nebadon Izumi: we talked about splitting it out before
[11:34]  Andrew Hellershanks: It would be an interesting project to strip a viewer down to where it could connect to a region and make a map tile.
[11:34]  Nebadon Izumi: but the problem is the tiler needs access to all the regions data
[11:34]  Nebadon Izumi: so even if its split out it needs to download the entire region before it can render
[11:34]  Justin Clark-Casey: if somebody really wanted to, they could take libomv, have it login, get the object, etc. data and then feed that back into warp3d
[11:34]  Nebadon Izumi: so basically you end up running 2 simulators
[11:34]  Nebadon Izumi: just to make tiles
[11:34]  Justin Clark-Casey: would be a lot of work for minimal gain though
[11:35]  Dahlia Trimble: well the region needs to grab all the assets to render anyway
[11:35]  Nebadon Izumi: ya and it needs to build the scene graph right?
[11:35]  Nebadon Izumi: basically its like running 2 simulators at that point
[11:35]  Dahlia Trimble: it has the scene graph already
[11:35]  Andrew Hellershanks: yup. For regions with lots of prims it requires getting a lot of data from the DBss
[11:35]  Nebadon Izumi: not if its in a seperate instance
[11:35]  Nebadon Izumi: how would that work?
[11:35]  Dahlia Trimble: all the prims are in memory already
[11:36]  Dahlia Trimble: its the textures
[11:36]  Nebadon Izumi: i mean if you have the map tile creation running in its own process
[11:36]  Nebadon Izumi: outside of OpenSim.exe
[11:36]  Nebadon Izumi: would there be some kind of connector?
[11:36]  Dahlia Trimble: nebadon, how long does it take to render map tile for this region?
[11:37]  Nebadon Izumi: it would still have to pull all the data from the simulator
[11:37]  Andrew Hellershanks: hm... the other option (still not great) would be if there was a way for a separate process to request the scene data so the maptile could be generated.
[11:37]  Nebadon Izumi: rendering is fast
[11:37]  Nebadon Izumi: its the asset fetch thats slow
[11:37]  Nebadon Izumi: if cache is cleared
[11:37]  Nebadon Izumi: 20+ minutes
[11:37]  Nebadon Izumi: if cache is full
[11:37]  Nebadon Izumi: seconds
[11:37]  Dahlia Trimble: but it still needs to decode all those textures, and there are around 4000 of them here
[11:37]  Nebadon Izumi: ya it goes pretty quick here
[11:38]  Nebadon Izumi: with full cache and scripts already compiled
[11:38]  Nebadon Izumi: this region starts up in like 2 minutes or something
[11:38]  Nebadon Izumi: thats with Warp3D
[11:38]  Dahlia Trimble: Id be surprised if it goes quick. It takes minutes for my best machine to decode all the textures in wright plaza using openjpeg
[11:39]  Nebadon Izumi: its a 8 core cpu
[11:39]  Nebadon Izumi: with 24gb ram
[11:39]  Dahlia Trimble: openjpeg is not threaded
[11:39]  Dahlia Trimble: in libomv
[11:39]  Nebadon Izumi: ya it starts up pretty quick though
[11:39]  Nebadon Izumi: it doesnt take more than 2-3 minutes total
[11:40]  Nebadon Izumi: I dont think warp3D does textures correctly though
[11:40]  Nebadon Izumi: i wouldnt be surprised if it skips most of them
[11:41]  Nebadon Izumi: 90% of the textures in this region wouldnt render
[11:41]  Nebadon Izumi: on the maptiles
[11:41]  Nebadon Izumi: they are inside of buildings
[11:41]  Andrew Hellershanks: It can skip any texture that is only on a prim in the vertical plane.
[11:41]  Dahlia Trimble: I doubt warp3d is smart enough to do occlusion culling
[11:42]  Nebadon Izumi: i am not even sure it renders textures properly at all
[11:42]  Andrew Hellershanks: Dahlia, fix the ref counting issue first, then the white line down the right side, then worry about occlusion culling. :)
[11:42]  Nebadon Izumi: the map tile here is really screwed up
[11:42]  Nebadon Izumi: odd
[11:42]  Justin Clark-Casey: I don't think Dahlia is planning on fixing anything :)
[11:42]  Justin Clark-Casey: in that area
[11:43]  Dahlia Trimble: Andrew, sorry I'm not touching that code :P
[11:43]  Mata Hari: lol
[11:43]  Andrew Hellershanks: Dahlia, I don't blame you.
[11:43]  Justin Clark-Casey: whoever wrote it clearly failed to realize that gfx objects need to be disposed
[11:43]  Mata Hari: how about the advanced mats stuff? is that a nightmare or relatively straighforward?
[11:43]  Robert Adams: Warp3d doesn't do occusion culling but it doesn't render things smaller then like 10m (not sure of exact number)
[11:43]  Mata Hari: (related to the constants you added last night)
[11:43]  Dahlia Trimble: If it were me I'd write a new one, but I dont have time so I wont :P
[11:44]  Andrew Hellershanks: Just put it on the ToDo list ;)
[11:44]  Nebadon Izumi: ya I am not against a new renderer
[11:44]  Nebadon Izumi: im just not convinced having a seperate process is a good solution
[11:44]  Justin Clark-Casey: I'm not against a new script engine
[11:44]  Dahlia Trimble pulls up Andrew's TODO list....
[11:44]  Andrew Hellershanks: Dahlia, mine is at least a full page in length.
[11:44]  Richardus Raymaker: A3 andrew ?
[11:44]  Nebadon Izumi: having it optionally be seperate is good
[11:45]  Mata Hari: lots of room on the other side then :)
[11:45]  Dahlia Trimble adds page 2
[11:45]  Nebadon Izumi: but it should also function without that as well
[11:45]  Andrew Hellershanks: Richardus, 8.5x11 for now.
[11:45]  Nebadon Izumi: kind of like Physics
[11:45]  Justin Clark-Casey: I doubt anybody is going to be doing that anytime soon
[11:45]  Richardus Raymaker: ohh tiny list :))
[11:45]  Nebadon Izumi: run map tiler on own thread option
[11:45]  Simulator Version v0.5 shouts: OpenSim 0.8.1 Dev          ca2379e: 2014-06-21 15:38:38 -0700 (Unix/Mono)
[11:45]  Andrew Hellershanks: Justin, I know of a number of issues with the parser that need fixing (or should be fixed).
[11:45]  reserect darkness: okies anyone any idea what port i have to open up for OSgrid
[11:45]  Nebadon Izumi: so it doesnt kill the heartbeat if its on a timer
[11:45]  reserect darkness: to get home lol
[11:46]  Andrew Hellershanks: Richardus, the other issue is that a lot of the items on the list are ones that require a lot of time to get done.
[11:47]  Richardus Raymaker: thats the real prblem andrew
[11:47]  Justin Clark-Casey: anyway, any other opensim topics today?
[11:47]  Andrew Hellershanks: Justin, I do want to look at the parser at some point. First thing is to add more regression tests.
[11:47]  Nebadon Izumi: well we need builders for OSCC
[11:47]  Justin Clark-Casey: andrew: I am always happy to see more regression tests
[11:47]  Nebadon Izumi: volunteers
[11:48]  Nebadon Izumi: if you are interested or know someone who might be
[11:48]  Justin Clark-Casey: oh yeah, plan is to start load tests again next week after this meeting
[11:48]  Nebadon Izumi: please have them contact Key Gruin or Myself
[11:48]  Nebadon Izumi: or just send them to
[11:48]  Andrew Hellershanks: Justin, I found out the parser allows an attempt to assign a string to an integer variable.
[11:48]  Nebadon Izumi: to register as a volunteer
[11:48]  Nebadon Izumi: we need all kinds of volunteers not just builders
[11:48]  Andrew Hellershanks: Justin, seems like an error in the BNF
[11:48]  Dahlia Trimble: is there a way the tests can be made smart enough to only run if the code they test is affected?
[11:49]  Andrew Hellershanks: Are there going to be more load tests for this years OSCC?
[11:49]  Justin Clark-Casey: I still haven't got round to writing out the instructinos to rebuild the parser, including the annoying hack
[11:49]  Nebadon Izumi: yes
[11:49]  Nebadon Izumi: first load test is the 1st
[11:49]  Nebadon Izumi: immediately after this meeting
[11:49]  Justin Clark-Casey: dahlia: no
[11:49]  Nebadon Izumi: we will announce that soon
[11:49]  Justin Clark-Casey: you can run individual test dlls ot individual tests manually, but of course, that's messy
[11:49]  Nebadon Izumi: but keep 3pm EST open if you can for July 1st
[11:50]  Andrew Hellershanks: Justin, You gave me enough info that I have gotten a handle on building the parser. I believe the hack is a sign that there is something wrong in the BNF that should be fixed so the hack isn't required.
[11:50]  Nebadon Izumi: and depending on how the first few load tests go will determine how often they occur after that
[11:50]  Andrew Hellershanks: dang, my cat is calling for me again.
[11:50]  Nebadon Izumi: I have a feeling we wont be doing as many load tests this year as last
[11:50]  Justin Clark-Casey: andrew: yeah, I might be an issue there, it's a strange hack to have to do
[11:50]  Nebadon Izumi: but i hope we can get a few really good ones in this year
[11:50]  Justin Clark-Casey: it might
[11:50]  Lucy Rosadisharon: hello
[11:50]  Lucy Rosadisharon (en -> es): Hola
[11:50]  Nebadon Izumi: and push some improvements down the pipe
[11:50]  Justin Clark-Casey: nebadon: there will have to be quite a few if one wants to get near 400
[11:50]  Nebadon Izumi: maybe
[11:51]  Nebadon Izumi: Im optimistic
[11:51]  Andrew Hellershanks: I'm not sure if I can be around on the 1st. It is Canada Day for me and I also have a visitor arriving from England that day.
[11:51]  Nebadon Izumi: :)
[11:51]  Justin Clark-Casey: hello lucy
[11:51]  Justin Clark-Casey: nebadon: wait till it fails, then you'll be totally pessimistic ;)
[11:51]  Lucy Rosadisharon: hi justin
[11:51]  Lucy Rosadisharon (it -> en): hi justin
[11:51]  Nebadon Izumi: ok no worries the first one will mostly be formalities making sure things still work
[11:51]  Nebadon Izumi: people can get in
[11:51]  Nebadon Izumi: for the first few weeks we will be definitely doing load tests
[11:51]  Andrew Hellershanks nods
[11:51]  Casias Falta: say again what the contact is to volunteer please
[11:51]  Nebadon Izumi: pretty regularly
[11:51]  Nebadon Izumi:
[11:51]  Nebadon Izumi: there is a be a volunteer link there
[11:51]  Casias Falta: yks
[11:52]  Casias Falta: tks
[11:52]  Nebadon Izumi: have a read and if your interested and have the time and ability register and we'll get you started :)
[11:52]  Billy.Bradshaw wonders if a deprecated list could accompany releases, so those features no longer supported could die gracefully
[11:52]  Nebadon Izumi: I cant think of anythiing we actually deprecated
[11:52]  Nebadon Izumi: has to be something but it eludes me
[11:52]  Andrew Hellershanks: Justin, apart from the script parser the other important thing to (re-)visit is handling of asset blobs.
[11:53]  Justin Clark-Casey: andrew: in what way?
[11:53]  Richardus Raymaker: oh justin..
[11:53]  Richardus Raymaker: how's the assets robust cleaner you worked on ?
[11:53]  Justin Clark-Casey: richardus: robust cleaner?
[11:53]  Richardus Raymaker: guess still not in 0.8
[11:53]  Andrew Hellershanks: Justin, It used to be a db could blow up if it got too big, now I hear it is a problem if it gets too large as it can slow down access.
[11:53]  Richardus Raymaker: you made some assets cleaner
[11:53]  Nebadon Izumi: i think he means deduping asset service
[11:53]  Justin Clark-Casey: I still need to activate the deduplicating asset service, which works absolutely fine
[11:54]  Richardus Raymaker: yes neb
[11:54]  Justin Clark-Casey: I just needed to finally work out the migration story
[11:54]  Andrew Hellershanks: Justin, there is my blobs on disk experiment, your notes in the wiki about using two tables for de duping, then there is SRAS.
[11:54]  Richardus Raymaker: but its not in 0.8robust ?
[11:54]  Justin Clark-Casey: I actually did have a neat way of doing it piecemal over time - just want to work out how to finally do it
[11:54]  Andrew Hellershanks: Justin, it would be great if that can be done before the next release.
[11:54]  Justin Clark-Casey: andrew: there is a whole XAssetService in OpenSimulator which does the deduping
[11:55]  Andrew Hellershanks: Is that in master?
[11:55]  Justin Clark-Casey: I want to, just any big change can meet with objections from certain
[11:55]  Justin Clark-Casey: quarters, so I always need to be prepared to meet them, which is a pita
[11:55]  Andrew Hellershanks: yeah. It will be a benefit to any large grid.
[11:56]  Justin Clark-Casey: activating instructions are at
[11:56]  Justin Clark-Casey: it prints a big warning say that it's experimental, because I may still want to change it before finalisation
[11:56]  Justin Clark-Casey: but it works just fine - if you save the same asset twice, even qwith different uuids, it only keeps one copy and two links to that copy
[11:57]  Andrew Hellershanks: I'm looking to switch to using SRAS but have to export all the data from a large asset table. If asset blobs never change I'm thinking I can migrate the blobs to disk over time up to a certain date then shutdown grid to do blobs since the given data and switch to full use of SRAS.
[11:57]  Nebadon Izumi: I would suggest waiting honestly
[11:57]  Nebadon Izumi: if its going to happen soon
[11:57]  Justin Clark-Casey: I think an interesting approach for SRAS would be to point it at an existing opensim asset service as a backing store
[11:57]  Andrew Hellershanks: Justin, ok, ty. I'll take a look at it out of interest.
[11:57]  Nebadon Izumi: you should do some test runs with whats in core now
[11:57]  Justin Clark-Casey: then any request for an asset could be pulled from robust and stored at that time
[11:57]  Justin Clark-Casey: which is what the xassetsservice does
[11:58]  Justin Clark-Casey: so with xassetservice, there should be no need to do a big bang migratio nby taking down the grid - it will just be done slowly over time
[11:58]  Justin Clark-Casey: but at some poitn, one would want to do a final conversion frmo the old store, which is the detail I need to work out
[11:58]  Nebadon Izumi: thats how we did sras too
[11:58]  Nebadon Izumi: it took weeks to convert sras
[11:58]  Tiffany Magic: Bye all... back to Virtual Highway for me.
[11:59]  Andrew Hellershanks: Nebadon, you did a gradual migration?
[11:59]  Nebadon Izumi: yes
[11:59]  Nebadon Izumi: took very long
[11:59]  Justin Clark-Casey: bye tiffany
[11:59]  Nebadon Izumi: we kept the grid up though
[11:59]  Nebadon Izumi: we did it without a single day of downtime
[11:59]  Justin Clark-Casey: how did you achieve that?
[11:59]  Nebadon Izumi: we had scripts
[11:59]  Nebadon Izumi: dave coyle is very good
[11:59]  Nebadon Izumi: heh
[12:00]  Justin Clark-Casey: I mean, did you keep the robust service online until everything was converted
[12:00]  Justin Clark-Casey: then shutdown the grid and convert anything remaining. or something similar?
[12:00]  Justin Clark-Casey: where shutdown would be for a much shorter time
[12:00]  Nebadon Izumi: well when we went from robust to sras back then the db was a lot smaller
[12:00]  Nebadon Izumi: but yes basically we dumped assets slowly
[12:00]  Nebadon Izumi: over a few weeks
[12:00]  Andrew Hellershanks: its 3 pm and my cat has come to see me since I didn't come up to see him. I also have to leave in half an hour to run some errands so I need to logout shortly.
[12:00]  Nebadon Izumi: we did shut the grid down briefly
[12:00]  Nebadon Izumi: couple hours
[12:00]  Justin Clark-Casey nods
[12:00]  Nebadon Izumi: so we could sync the very last bit
[12:01]  Nebadon Izumi: without disruption
[12:01]  Richardus Raymaker: bte andrew
[12:01]  Andrew Hellershanks: nebadon, yeah. I was thinking I could do that using my blobs to disk migration script.
[12:01]  Nebadon Izumi: yea
[12:01]  Justin Clark-Casey: I suspect with a backing approach, people might never bother with a full migration, though that premanently leaves some asstes in the old service and others in the new
[12:01]  Nebadon Izumi: we had something similar that adam wrote in C#
[12:01]  Nebadon Izumi: things were very different then though
[12:01]  Justin Clark-Casey: which is not ideal, I think some kind of final command is needed, which might be a console command
[12:01]  Nebadon Izumi: this was many years ago
[12:01]  Andrew Hellershanks: It was going to take 36 hours to export the data to move to a new machine then I still had to copy and re-import it.
[12:02]  Nebadon Izumi: yea
[12:02]  Nebadon Izumi: 300gb is going to take time
[12:02]  Nebadon Izumi: no matter what you do
[12:02]  Andrew Hellershanks: I was looking at 3 days down time.
[12:02]  Andrew Hellershanks: yeah. I have 400G.
[12:02]  Richardus Raymaker: 400G how many regions/users ?
[12:03]  Andrew Hellershanks: I don't remember the user count. 2 or 3 thousand(?) and around 250 some odd regions.
[12:03]  Richardus Raymaker: ohh. ok. then my 128GB ssd is enough for a long time
[12:04]  Andrew Hellershanks: Also depends on how long the grid has been running
[12:04]  Justin Clark-Casey: I do think a clearner approach is possible as well though, without having to do a full grid mark and sweep
[12:04]  Justin Clark-Casey: if one keeps occasional last access times for assets
[12:04]  Andrew Hellershanks: that would be helpful.
[12:04]  Justin Clark-Casey: then it depends if one is comfortable wiping an asset whose last recorded access time was, say, 3 years ago
[12:05]  Andrew Hellershanks nods
[12:06]  Andrew Hellershanks: I probably have textures in my inventory I haven't accessed in that long.
[12:06]  Justin Clark-Casey: you need to be careful that simulators don't end up permanently caching assets and never re-requesting them from the service
[12:06]  Justin Clark-Casey: andrew: right, so deleting on that basis might not really be viable
[12:06]  Justin Clark-Casey: unless one accepts that unused stuff will get deleted as well
[12:06]  Andrew Hellershanks: Someone also suggested the possibility of a reference count but I'm not sure if that would be any more reliable.
[12:07]  Justin Clark-Casey: it's practically impossible to do a reference count over a large distributed grid
[12:07]  Richardus Raymaker: it  would be intressting if unused stuff could be moved out of the database and packed into small files seperate on disk.
[12:07]  Andrew Hellershanks: If I'm not going to pay much attention to the cat he wants my chair. He is trying to push behind me. :)
[12:07]  Richardus Raymaker: i mean stuff thats not used for few years
[12:07]  Justin Clark-Casey: richardus: then why not just keep it in the db if you have it on disk anyway
[12:08]  Justin Clark-Casey: I am not aware of a penalty for having things in db rather than on disk
[12:08]  Richardus Raymaker: big db can slow things down, right ?
[12:08]  Andrew Hellershanks: If things are de-duped that still cuts down on disk usage
[12:08]  Justin Clark-Casey: I have not seen any convincing benchmarks demonstrating that
[12:08]  Andrew Hellershanks: Richardus, that's what I've heard.
[12:08]  Richardus Raymaker: only idea, dont hit the panic button
[12:09]  Justin Clark-Casey: anyway, I need to go as well
[12:09]  Richardus Raymaker: the biggest dubes you have  are plywood cubes etc
[12:09]  Justin Clark-Casey: thanks for the covnersation, folks
[12:09]  Richardus Raymaker: bye justin
[12:09]  Andrew Hellershanks: Justin, Items like NC's or scripts that you create and delete are pointless to keep in the asset table.
[12:09]  Justin Clark-Casey: yeah, elminating all those plywood cubes wouod be great
[12:09]  Sarah Kline: byes
[12:09]  Nebadon Izumi: kk Justin, and thanks everyone
[12:09]  logger sewell: Justin TC
[12:09]  Justin Clark-Casey: andrew: yeah
[12:09]  Andrew Hellershanks: but that is another issue on how to know if they can be removed.
[12:09]  Nebadon Izumi: good meeting, remember July 1st
[12:09]  Nebadon Izumi: big OSCC14 Load test
[12:09]  Richardus Raymaker: thats other good one andrew
[12:09]  Nebadon Izumi: mark it on your calendars
[12:09]  Nebadon Izumi: right after this meeting
[12:09]  Nebadon Izumi: we will teleport to the testing area
[12:09]  Mata Hari: Canada Day party load tst?
[12:10]  Nebadon Izumi: yes
[12:10]  Dahlia Trimble: bye all :)
[12:10]  Nebadon Izumi: bring your Maple Syrup
[12:10]  Mata Hari: will there be beer?
[12:10]  Richardus Raymaker: t byew
[12:10]  Andrew Hellershanks: yeah. I make a script or NC for an object then delete it or modify it. No one else has it so why not be able to reuse the asset or delete it from the db?
[12:10]  vegaslon plutonian: someone needs to bring a truck to load up the donuts
[12:10]  Andrew Hellershanks: ok, I need to get ready to go out for a bit.
[12:10]  Andrew Hellershanks: Interesting meeting today.
[12:10]  Richardus Raymaker: a script you edit 10's of times
[12:10]  Justin Clark-Casey: andrew; yeah, it's a case of detecting those situations reliably
[12:10]  Justin Clark-Casey: which might be possible, but could be hard
[12:10]  Robert Adams: where are we TPing to?
[12:10]  Andrew Hellershanks: The start is if you created it and own it.
[12:10]  Richardus Raymaker: type, save. oh other bug save.. etc
[12:11]  Casias Falta: Interesting discussion cu everyone
[12:11]  Sarah Kline: bi
[12:11]  logger sewell: tc all
[12:11]  Andrew Hellershanks: Richardus, exactly.
[12:11]  Justin Clark-Casey: tbh though, script size is dwarfed by texture and serialized object sizes
[12:11]  Cuteulala Artis: Ohhhh! boy we need a meeting reminder group :(
[12:11]  Andrew Hellershanks: Justin, true. It still bothers me a little.
[12:11]  Justin Clark-Casey: bye folks
[12:11]  Richardus Raymaker: hi cute..
[12:11]  Andrew Hellershanks: cya, justin
[12:11]  Cuteulala Artis: hi
Personal tools
About This Wiki