(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
[11:02 AM] Andrew Hellershanks: Hello, everyonel
[11:02 AM] Ubit Umarov: machine load is not high
[11:03 AM] Lyr.Lobo @cc.opensimulator.org:8002: Hello *grins*
[11:03 AM] Vincent.Sylvester @hg.zetaworlds.com: Yeah it's something network related making it slow, but eh eventually it loads so good enough
[11:03 AM] Vincent.Sylvester @hg.zetaworlds.com: Who knows what kind of malware already got into the poor thing... or should I say melware...
[11:06 AM] Ubit Umarov: well good news.. no new bugs added during last week!
[11:06 AM] Andrew Hellershanks: If you do a ping it might confirm if there are network issues.
[11:06 AM] Andrew Hellershanks: /me grins at Ubit.
[11:06 AM] Vincent.Sylvester @hg.zetaworlds.com: I did write something earlier today that I might contribute in given it apparently isn't something that can be handled any other way
[11:07 AM] Vincent.Sylvester @hg.zetaworlds.com: A validation function for http requests that gracefully returns whether a connection can even be made from the request parameters
[11:08 AM] Vincent.Sylvester @hg.zetaworlds.com: Currently http request will just fail with an error, which exposes the request address among other things publicly so not ideal for some
[11:08 AM] Vincent.Sylvester @hg.zetaworlds.com: Can't handle it with try catch either cause it's not an exception as such
[11:08 AM] Ubit Umarov: guess such thing is slower that actually doing a connection.... but who knows
[11:09 AM] Vincent.Sylvester @hg.zetaworlds.com: Well I certainly wouldn't recommend doing that every request, but perhaps as a way to first validate whether the parameters work or if the address can be resolved, such stuff mostly
[11:09 AM] Vincent.Sylvester @hg.zetaworlds.com: I had an issue with that for some reason DNS reporting a NAT IP for a public machine and since that's blacklisted by the filter it just blew up without a way to handle the error
[11:10 AM] Vincent.Sylvester @hg.zetaworlds.com: Perhaps osHTTPRequest throwing actual exceptions for try catch to handle could be a way to do it also
[11:10 AM] Vincent.Sylvester @hg.zetaworlds.com: The open nature of the errors potentially exposing data or just being annoying as they broadcast to everyone has been a topic over the years
[11:11 AM] Vincent.Sylvester @hg.zetaworlds.com: Quite a few times have I heard the suggestion to restrict the messages to the script owners
[11:11 AM] Andrew Hellershanks: Fixing the error message sounds like a simpler fix.
[11:11 AM] Vincent.Sylvester @hg.zetaworlds.com: I think that has merit, but perhaps estate owner and manager might benefit from knowing about script errors as well
[11:12 AM] Ubit Umarov: guess error to all is a nice tactic
[11:12 AM] Ubit Umarov: to have scripts fixed
[11:12 AM] Andrew Hellershanks: Either option would be better than the message going to everyone
[11:12 AM] Vincent.Sylvester @hg.zetaworlds.com: Well yeah, proper DNS is always a good idea, but there currently is no way to "test" a http request to make sure it doesn't blow up. If it does explode then there is no feedback to the script on what has failed specifically, so you can't really handle it either
[11:12 AM] Ubit Umarov: in that mainland thing
[11:13 AM] Andrew Hellershanks: No easy way to throw an error that ties it back to a line of code in a script?
[11:14 AM] Vincent.Sylvester @hg.zetaworlds.com: http request only returns null key in certain cases, otherwise just errors and returns nothing basically
[11:14 AM] Vincent.Sylvester @hg.zetaworlds.com: If it threw a fatal script exception then try catch would be able to handle that, but it's a custom error function
[11:18 AM] Vincent.Sylvester @hg.zetaworlds.com: Like I said multiple ways of doing it, my first attempt right now is just a validation function returning integers corresponding to failure modes otherwise just returning true. Use case is validating urls, dns or parameters and headers. Possibly could do a oshttprequest throwing actual exceptions for handling via try catch, but that's kind of undocumented behavior and not LSL-like so to speak... not that I care, but yeah
[11:19 AM] Ubit Umarov: you cant add much to httprequest
[11:20 AM] Ubit Umarov: at least nothing liek goes out like dns
[11:20 AM] Andrew Hellershanks: brb
[11:20 AM] Ubit Umarov: it is async for a reason...
[11:21 AM] Andrew Hellershanks: /me is back
[11:22 AM] Vincent.Sylvester @hg.zetaworlds.com: I have to test it still, but if you look at the lsl api code for that all I have right now is just copy of everything that can throw an error and just gave that a return code, not actually doing the request itself, just the checks basically
[11:22 AM] Ubit Umarov: ( ofc ther ossl things that decided to be sync where ... lets say a mistake.. (
[11:22 AM] Vincent.Sylvester @hg.zetaworlds.com: A sort of --dry-run deal
[11:22 AM] Ubit Umarov: like the notecard nonsense
[11:24 AM] Ubit Umarov: well at least there is some caching on those...
[11:24 AM] Vincent.Sylvester @hg.zetaworlds.com: Speaking of notecards. I saw the other day that "persistence ejected" and "persistence visitor" notecards are growing at about 10000 per week on ZetaWorlds. Haven't checked yet what script that is, but not a great rate to see that much new stuff
[11:25 AM] Vincent.Sylvester @hg.zetaworlds.com: Most of these I suspect are just entries pointing to the same data so not much waste besides the db entries
[11:26 AM] Vincent.Sylvester @hg.zetaworlds.com: Our "friends" from IMA apparently made a parcel dwell log thing as well, going up by about 1500 notecards per week, so much for their expertise I guess
[11:26 AM] Vincent.Sylvester @hg.zetaworlds.com: The fear of losing data due to restarts seems to compel some people to save notecards as often as every 5 minutes
[11:27 AM] Vincent.Sylvester @hg.zetaworlds.com: No checks of new data or anything like that either
[11:29 AM] Ubit Umarov: pff you just need newer TB ssds
[11:29 AM] Ubit Umarov: they may grow on trees... try plant some :p
[11:29 AM] Vincent.Sylvester @hg.zetaworlds.com: lol
[11:29 AM] Vincent.Sylvester @hg.zetaworlds.com: Was about to say
[11:30 AM] Vincent.Sylvester @hg.zetaworlds.com: Don't plant seagate or you'll poison the well
[11:33 AM] Andrew Hellershanks: :)
[11:34 AM] Andrew Hellershanks: I used to prefer Seagate. How times have changed.
[11:35 AM] Vincent.Sylvester @hg.zetaworlds.com: Supposedly their volume discounts is why they can get away with less than ideal mtbf rates
[11:39 AM] Andrew Hellershanks: Any other items for today? Any upcoming events of note?
[11:40 AM] Vincent.Sylvester @hg.zetaworlds.com: Oh right. I did some tests this week and found some interesting behavior of garbage collection in regards to maptile generation. Changing to Warp3D for the initial tile generation at startup would make memory consumption explode as somehow that memory got pretty stuck, gc wouldn't free it up after the maptile was made. Every new simulator would also just grow in memory usage. Doing normal tiles and only later switching to Warp3D would exhibit and increase in memory as well, but gc would eventually start clearing it all out
[11:40 AM] Vincent.Sylvester @hg.zetaworlds.com: Doing that daily now and seeing the bumps of that generation going up and down in memory now
[11:41 AM] Vincent.Sylvester @hg.zetaworlds.com: Have to do some specific tests if the initial memory used for tile generation is freed at all
[11:41 AM] Andrew Hellershanks: I think there is an outstanding bug report about memory usage by Warp3D.
[11:42 AM] Vincent.Sylvester @hg.zetaworlds.com: Well it's bound to be higher and a bit glitchy as it tries to render stuff only to eventually fail
[11:43 AM] Vincent.Sylvester @hg.zetaworlds.com: I suspect part of the memory usage is storing the maptile on scene rather than assets, so I expected more memory usage with that code, just not quite as much and with such odd behavior
[11:43 AM] Ubit Umarov: there is no bug... it is jsut that
[11:43 AM] Vincent.Sylvester @hg.zetaworlds.com: Like it would literally add 25mb per simulator started
[11:43 AM] Ubit Umarov: GC has its own mind abotu memory release
[11:44 AM] Ubit Umarov: well at least it is a bit better than on mono days
[11:44 AM] Vincent.Sylvester @hg.zetaworlds.com: Certainly
[11:44 AM] Ubit Umarov: and warp3d does use tons of memory yeah
[11:44 AM] Ubit Umarov: it full draws a region...
[11:45 AM] Ubit Umarov: so needs to get everything
[11:45 AM] Vincent.Sylvester @hg.zetaworlds.com: Really the odd part for me is this weird linear memory increase, it's as if it needs 25mb extra for each dotnet app
[11:46 AM] Vincent.Sylvester @hg.zetaworlds.com: The first simulator using 400mb, next 425, 450 etc. all running empty regions with nothing in it
[11:46 AM] Ubit Umarov: pff just buy more ram...
[11:46 AM] Ubit Umarov: it may grow on trees also :p
[11:47 AM] Vincent.Sylvester @hg.zetaworlds.com: I wonder if starting simulators one after the other without giving them time to settle down might play a part
[11:48 AM] Vincent.Sylvester @hg.zetaworlds.com: I don't think it's code, it's something in the runtime adding memory per instance for some reason
[11:48 AM] Vincent.Sylvester @hg.zetaworlds.com: Once they all started and had time they show the same memory consumption as expected
[11:49 AM] Vincent.Sylvester @hg.zetaworlds.com: I know that dotnet caches a ton of sys info and whatnot at start of an application so perhaps there is a bug there, have yet to check their issue tracking for anything related
[11:50 AM] Vincent.Sylvester @hg.zetaworlds.com: Given it's not idea to rapidly start simulators one after the other anyways unless you have plenty of resources to spare maybe just a quirk though
[11:50 AM] Vincent.Sylvester @hg.zetaworlds.com: Wouldn't be the first
[11:51 AM] Andrew Hellershanks: I ran one system where I started about 5 regions at a time. No more than that due to the CPU load.
[11:52 AM] Vincent.Sylvester @hg.zetaworlds.com: Unhooking the maptile generation from the startup process has reduced the load massively since I can just disable that entirely now and generate the maptiles later
[11:52 AM] Vincent.Sylvester @hg.zetaworlds.com: Next to the db calls it's the heaviest thing
[11:53 AM] Ubit Umarov: pff a 14900ks will eat regions per lunch... or breakfeast or dinner... may even eat your wallet!
[11:54 AM] Vincent.Sylvester @hg.zetaworlds.com: Or degrade and explode
[11:55 AM] Ubit Umarov: or/and who knows :)
[11:58 AM] Andrew Hellershanks: That t-rex doesn't know its place. It used to stay out of the bar.
[11:58 AM] Ubit Umarov: that was before he learned about beer
[11:59 AM] Andrew Hellershanks: :)
[11:59 AM] Joe Magarac: Gotta go, next meeting. Thanks!
[11:59 AM] Andrew Hellershanks: We are almost at the top of the hour. Any last minute items for today?
[11:59 AM] Ubit Umarov: have fun
[11:59 AM] Andrew Hellershanks: ok, Joe. See you next week.
[12:00 PM] Andrew Hellershanks: I'm noticing it must be summer. A couple of the other regular attendees have been absent for a few weeks.
[12:01 PM] Ubit Umarov: well it is winter in australia
[12:01 PM] Andrew Hellershanks: As there appears to be nothing further for today I will call this meeting to a close. Thank you all for coming. See you again next week.
[12:01 PM] Ubit Umarov: and other places down there >(
[12:01 PM] Andrew Hellershanks: True.
