Chat log from the meeting on 2018-04-24

From OpenSimulator

Jump to: navigation, search

Todays topics:
how to get a copy of the httptest branch
a short course in git and some tools
why httptests is still not master (.net version and Jenkins)
suspected memory leak in ROBUST
some discussion of mono-, kernel- and network-settings
remarks about the necessity of advancing to more recent versions of mono a.s.o.

[11:06] Sheera Khan: Does anyone happen to know if the setting of day_night_offset is implemented in 0.9.1?
[11:07] Andrew.Hellershanks Sheerah, is that in a region setting within the viewer or in a scripting call?
[11:07] Sheera Khan: It should be a setting in OpenSim.ini I was told
[11:08] Sheera Khan: I didn't test it myself, a resident wanted to know that and I havn't even heard of it ^^
[11:08] Andrew.Hellershanks If the setting is still in the 0.9 ini files I would say yes but it would be easy to test.
[11:09] Sheera Khan: my last region is offline as my server is gone :-(
[11:10] Andrew.Hellershanks Sheera, the setting is still in the OpenSimDefaults.ini file.
[11:10] Sheera Khan: yes, that I could verify...
[11:10] Sheera Khan: But I was told the setting didn't change anything
[11:11] Sheera Khan: I couldn't verify that though
[11:12] Andrew.Hellershanks There may be some confusion as to the purpose of that setting. The comment say it is Day to Night ratio. That means it isn't strictly speaking an offset. It would depend on how it is implemented.
[11:13] Andrew.Hellershanks The code says it shifts the suns orbit which affects the day/night ratio.
[11:14] Ubit Umarov: suns orbit .. thats so funny :)
[11:14] Andrew.Hellershanks :)
[11:14] Bill.Blight Yeah I always thought it just , as it said adjusted the amount of day vs night, not an offset to current time
[11:14] Sheera Khan: someone didn't get the memo I guess about the heliocentric model
[11:16] Sheera Khan: but I heard the Flat-Earthers are on the rise again, so....
[11:16] Bill.Blight uggg
[11:17] Kayaker Magic: Andrew, I asked you this just the other day, but then got distracted by RL, so I have to ask again: What is the git syntax on Linux to get a copy of the httptest branch? I go to the folder where I do my normal git pull? Or do I make a new folder for this? Then type git branch ...?
[11:17] Bill.Blight git checkout httptests
[11:18] Andrew.Hellershanks Kayaker, you get the git repo for master then checkout the branch.
[11:18] Bill.Blight I would advise then copying that to a new folder for compiling purposes
[11:19] Bill.Blight if you plan on making change to create a patch
[11:19] Kayaker Magic: So make a new copy of the opensim folder before doing this?
[11:19] Andrew.Hellershanks git branch local-branch-name remote-branch-name
[11:19] Bill.Blight nono
[11:19] Kayaker Magic: cd ..
[11:19] Andrew.Hellershanks That will let you create a local branch that trackes a remote branch.
[11:19] Bill.Blight cd into your opensim folder where you pulled it ..
[11:19] Bill.Blight git checkout httptests
[11:19] Misterblue Waves: git checks out the sources into a 'working directory' which you checkout and checkin changes into the git change storage....
[11:19] Bill.Blight then copy the contents to a new folder to compile
[11:19] Ubit Umarov: it is like bill is telling
[11:20] Misterblue Waves: so you update the working directory by checking-out the branch and/or version you want
[11:20] Kayaker Magic: ls
[11:20] Sheera Khan: but you have to clone the repo before you can dothe checkout, right?
[11:20] Bill.Blight compiling in your git pull/clone directory can pollute it with unneeded files
[11:20] Misterblue Waves: usually, you're on the 'master' branch but to change the working directory to another branch you say 'git checkout httpstest'
[11:20] Bill.Blight yes clone it
[11:21] Andrew.Hellershanks Sheera, correct.
[11:21] Andrew.Hellershanks Kayaker, if you have already cloned the OS repo you can make an extra copy of it so you can have more than one version of code available.
[11:21] Misterblue Waves: you can see differences between the working directory and the change store with 'git status' and 'git diff'
[11:21] Andrew.Hellershanks without constantly switching between versions.
[11:22] Kayaker Magic: So make a copy of my opensim folder, and git checkout will convert one of them to httptests, then build there.
[11:22] Misterblue Waves: if you're building in the working directories, what Andrew said is a good idea
[11:22] Andrew.Hellershanks I usually use git gui to do branch related operations as it is too easy to wind up not on a branch.
[11:23] Andrew.Hellershanks I have three branches of OpenSim checked out on my machine with each in seaprate directories.
[11:23] Bill.Blight there is a nice linux utility called tig which lets you browse commits via the command line, it is called tig
[11:23] Misterblue Waves: but you can look between branches and commits with git... 'git checkout httpstest'; git diff master -- OpenSim/Region/PhysicsModule.cs'
[11:23] Andrew.Hellershanks Bill, interesting. I have been using ViewVC with a web browser to explore repos.
[11:24] Bill.Blight tig is pure console app
[11:24] Bill.Blight after install go to your repo root and run tig
[11:24] Bill.Blight pretty nice
[11:25] Andrew.Hellershanks It is available for my distro. The description says it is ncurses based.
[11:25] Andrew.Hellershanks That should make it quite easy to use.
[11:25] Bill.Blight it is nice because it will show any local branches or commits you have made that have not been pushed
[11:25] Kayaker Magic: I'm getting errors about the commits I did to change llAtan, it refuses to change branches
[11:26] TG.Lucan Ubit, when running httptests are there any specific observations ?
[11:26] Bill.Blight that is becaused they are un pushed , and don't think you can push them
[11:26] Bill.Blight maybe do a git stash
[11:26] Andrew.Hellershanks You could try stashing your changes but at this point you are better off making a new directory/copy to hold the http tests branch.
[11:26] Bill.Blight ^^
[11:27] Ubit Umarov: not that i remember now
[11:27] TG.Lucan k
[11:27] Bill.Blight one of the reasons I make all code changes in a copy, test them then if they work I move them back in, but I use my own github .. You might want to fork it Kayaker if you want to commit changes to storage
[11:28] Kayaker Magic: Made a copy of the opensim folder, typed git stash in that one, then git branch httptests, finally git pull. So if I build there I will have the httptests version to play with?
[11:28] Bill.Blight ummm
[11:28] Bill.Blight no
[11:28] Kayaker Magic: I just want to test, not make changes.
[11:28] Bill.Blight you need to 'git checkout httptests'
[11:28] Bill.Blight not branch
[11:28] Andrew.Hellershanks Kayaker, use 'git status' to make sure what branch you are on.
[11:29] Bill.Blight git branch creates a local branch
[11:29] Kayaker Magic: git status says I am on branch httptests
[11:29] Andrew.Hellershanks Not if you give it the information that you are creating a local branch that will track a remote branch.
[11:29] Bill.Blight yes the local one you just created
[11:29] Ubit Umarov: no idea why git branch was mentioned here.. whatever
[11:29] Andrew.Hellershanks git branch local-branch-name remote-branch-name Make a new local branch that tracks a remote branch
[11:29] Misterblue Waves: 'stash' stores all the uncommitted changes to a stack... so you can 'git stash' (making a clean working tree, 'get pull' ( to get changes', then 'git pop' to reapply the changes you saved with t he 'git stash'
[11:29] Bill.Blight it does not have the pulled code from httptests it will just be a copy of the branch you were on when you ran it
[11:30] TG.Lucan Unless you are making changes to code the only stash request is for osslEnable.ini so rename that if you have any settings
[11:30] Bill.Blight ok so do this .. git checkout master, git branch -d httptests, git checkout httptests
[11:30] Andrew.Hellershanks TG.Lucan, Kayaker has uncommited changes in his source tree.
[11:31] Andrew.Hellershanks I give up.
[11:31] TG.Lucan oh
[11:32] Bill.Blight yeah
[11:32] Andrew.Hellershanks Bill, that's the long way around but I suppose it might get him there in the end.
[11:32] Bill.Blight he created a local branch
[11:32] Bill.Blight can't pull the httptests branch if he has a local branch
[11:33] Kayaker Magic: When I run the resulting opensim.exe, how will I know I built the right one? What will the region version be?
[11:33] TG.Lucan check version at the console
[11:33] Bill.Blight snail-dev I think
[11:34] Andrew.Hellershanks Kayaker, before you build run "git log". If you have the right version checked out you will see "commit f29358744d892f81edb7825cf97f6bbad90e1c4e" at the top of the page.
[11:34] TG.Lucan #show version
[11:35] Misterblue Waves: if you have a local branch (a branch that has never been pushed to a repository)... you can use 'rebase'... have you local branch all commited, 'git checkout httpstest', git pull origin httpstest'; 'git checkout localbranch', 'git rebase httpstest'... rebase undoes the commits in the local branch, merges the changes from httpstest and then reapplies the commits from the local branch
[11:36] Bill.Blight yeah rebasing is an option but in this case he'd be easier to just delete the local branch he created and pull the fresh one
[11:36] Andrew.Hellershanks Now Kayaker has about 4 ways to get where he wants to be.
[11:37] Bill.Blight Better option in this case, create a new directory all together such as /home/newos
[11:37] Bill.Blight cd to that
[11:37] Bill.Blight clone it
[11:37] Kayaker Magic: git log shows the hash Andrew said it should, build is done, has a lot fewer warnings than the regular 0.9.1 dev build, so I may have built it.
[11:37] Misterblue Waves: I have been known to export all my new commits (git format-patch), do a clean clone of a repository, and then reapply my commits (git apply)
[11:37] Bill.Blight cd to the opensim dir
[11:37] Bill.Blight and git checkout httptests
[11:38] Andrew.Hellershanks Kayaker, if git log shows that commit hash at the top of the list you have the right code checked out..
[11:39] TG.Lucan 2 warnings for the last httptests we did
[11:39] Ubit Umarov: of course looking to the top commit hash is a waste of time, since it will change on as soon there is a new commit
[11:40] Bill.Blight yep
[11:40] Bill.Blight better to just make sure you are on the remote branch with a checkout
[11:40] Ubit Umarov: if you use git bash shell it does show current branch
[11:40] Kayaker Magic: To try out Yengine, I have to edit that into OpenSim.ini? Or is the example Opensim.ini set up for Yengine in this build.
[11:40] Ubit Umarov: git bash on windows
[11:40] Andrew.Hellershanks Yes, but for this moment it tells Kayaker if he is on the right branch or not.
[11:40] Bill.Blight example is th ere
[11:40] Bill.Blight there
[11:41] Ubit Umarov: well but this is logged, other ppl will read it later
[11:41] Bill.Blight I can almost guarantee he is not, because he ran git branch httptests from the master branch which just made a local branch of master called h ttptests
[11:41] Ubit Umarov: so, be more precise on instructions if you please
[11:41] Kayaker Magic: Well, I'll try running this after the meeting. If it doesn't work I'll grab a whole new copy into yet another folder and start over....
[11:42] Ubit Umarov: it simple as bill tried to tell under the noise..
[11:42] Andrew.Hellershanks Ubit, I did provide more specific information but its been ignored so I gave up trying.
[11:42] Ubit Umarov: do a fresh git clone , then git checkout httptests
[11:42] Ubit Umarov: thats just that
[11:43] Kayaker Magic: Yeah, that sounds like the best way for me
[11:44] Bill.Blight maybe we need to hold a git class sometime
[11:45] Kayaker Magic: That I would like!
[11:45] Andrew.Hellershanks Rather feels like we are doing that now.
[11:45] Ubit Umarov: httptests is version 0.9.1 since it should had been master by now..
[11:45] Kayaker Magic: Something more organized....
[11:46] Ubit Umarov: but has a name on version
[11:46] Misterblue Waves: why don't you just merge it into master?
[11:46] Misterblue Waves: are there compatibility problems?
[11:46] Bill.Blight I'd suggest Youtube as well a lot of good git tutorials on there
[11:46] Ubit Umarov: our box is on mono 2.x
[11:46] Bill.Blight httptests is pushed to .net 4.6
[11:46] Bill.Blight the box's mono can't handle it
[11:46] Ubit Umarov: Region (root) # show version
Version: OpenSim Snail Dev (SIMULATION/0.3 - SIMULATION/0.6)
[11:47] Misterblue Waves: ah.... mono downward compatability
[11:47] Bill.Blight the box is like Ubuntu 12
[11:47] Andrew.Hellershanks Oh. I think the problem with updating the mono on the box is that it also has an old kernel.
[11:48] Misterblue Waves: since new development is happening on httpstest, might want to swap the branches: master -> ForMono2 and httpstest->master
[11:48] TG.Lucan Has anyone tested later mono with ubuntu 12?
[11:48] Ubit Umarov: and also that jenkins possible will not work with mono 5x
[11:48] Bill.Blight Jenkins works fine with mono 5.x do it daily
[11:49] Ubit Umarov: i said THAT jenkins :p
[11:49] Andrew Hellershanks: Why should Jenkins care which version of mono is being used?
[11:49] Misterblue Waves: how to upgrade? is it a complete rebuild of that server?
[11:50] Ubit Umarov: mel said she may try to compile mono 3 on it.. but had no time so far
[11:50] Ubit Umarov: i mean mono 5
[11:50] Bill.Blight don't think it is going to work due to old glibc
[11:50] Andrew Hellershanks: IIRC, you will need 3 before you can get to either 4 or 5.
[11:51] Misterblue Waves: sounds like a complete site rebuild which is a pain
[11:51] Bill.Blight yes, it was let get too far behind
[11:52] Bill.Blight and much to a lot of peoples dismay, httptests will not run on windows 7 anymore
[11:52] Ubit Umarov: yeap and so we are stuck
[11:52] Andrew Hellershanks: Ubit, I could attempt the build on the machine. I just have some other things I need to do first.
[11:52] Ubit Umarov: will not run on win7 why ??
[11:52] TG.Lucan we build mono in user space, so no dependancies on stock mono installed on the box
[11:52] Bill.Blight I thought you could not get 4.6 on win 7
[11:53] Ubit Umarov: you can even on vista
[11:53] Ubit Umarov: only XP is out
[11:53] Bill.Blight ohh I was not aware must be 4.7 you can't get
[11:53] Bill.Blight I was off by a version
[11:53] Ubit Umarov: vista will be out if we pull to .net 4.6
[11:53] Ubit Umarov: oops
[11:53] Ubit Umarov: to 4.6.2
[11:54] TG.Lucan we are running all opensim instances on 4.6.2 with no known issues
[11:54] Andrew Hellershanks: We are close to the top of the hour and spent most of it talking about git. Are there any other topics for todays meeting?
[11:55] Bill.Blight yep I was smoking crack evidently Win7 goes to .NET Framework 4.7.1
[11:55] Kayaker Magic: The guy I know who was trying to narrow down the source of the memory leak in Robust says:
[11:55] Kayaker Magic: I started small, just separating the asset service from everything else and it was not possible because HGassets takes a dependence on being in the same instance as assets. Then a public HG service has a dependence on the HG asset, but that then meant two instances required listening on the public port 8002 which is not permissible.]
[11:55] Ubit Umarov: but he main dif on 4.6.2 is that gc run times handles pinned memory a little better, they claim
[11:55] Ubit Umarov: ( on windows )
[11:56] TG.Lucan yes Ubit
[11:56] Ubit Umarov: you started small ?? what small ?
[11:57] Kayaker Magic: Small meaning "lets just split asset service into one instance"
[11:57] Kayaker Magic: "and everything else into another instance"
[11:57] Kayaker Magic: But did not work, so we are no closer to figuring out which service is leaking GB per day.
[11:57] Bill.Blight I'm not sure what memory leak people call bad, my robust has been running for 2 weeks and it is using 3.9 gig
[11:58] Bill.Blight and has sat there for many days
[11:58] Ubit Umarov: those things need to be grouped correctly
[11:58] Andrew Hellershanks: Bill, so bad that it can crash the server.
[11:58] Kayaker Magic: Andrew has this same problem on another grid, so there are two of us with terrible mem leaks in Robust.
[11:58] Bill.Blight I'm on mono
[11:59] Bill.Blight I wonder if these peeps are not recompiling their robust under the new mono
[11:59] Andrew Hellershanks: I forgot to write down the difference between the ini files. I think it was only in groups(?). There was very little difference.
[11:59] Andrew Hellershanks: Bill, the version of mono used doesn't matter.
[12:00] Bill.Blight actually it does, when basic code changes in mono , you need to recompile
[12:00] Ubit Umarov: andrew don't say nonsense...
[12:00] Bill.Blight they are changing the underlying code so complilation matters
[12:00] Ubit Umarov: it does depend on the runtime used
[12:00] Kayaker Magic: I think Andrew means he tried recompiling with different monos and it did not fix the leak.
[12:00] Ubit Umarov: gc is a thing they keep changing from version to version
[12:01] Ubit Umarov: ( and mono gc is not as ms one )
[12:01] Ubit Umarov: at least so far... possible ms will also merge that
[12:01] Bill.Blight Is that what you meant Andrew? If so sorry for jumping out like that
[12:02] TG.Lucan You can run dot net core on linux
[12:02] Bill.Blight sure you can .. if you are brave
[12:03] TG.Lucan then we are brave :)
[12:03] Ubit Umarov: well i do have those reports on mem leaks, but no clues and can't repo :(
[12:04] Bill.Blight I've tried to watch to see if I can repo it as well, can't
[12:05] Bill.Blight another thing to check with these memory leaks is the underlying system and kernel, it may not be a mono issue at all
[12:05] TG.Lucan We don't have that many visitors daily, maybe 2 or 3. So our public grid may not be representative
[12:05] Ubit Umarov: keep in mind, we do run with the desktop option on mono
[12:05] Bill.Blight yes
[12:05] Ubit Umarov: that does seem a lot better for opensim than server
[12:05] Bill.Blight if you do --server it will grow and grow and grow
[12:05] Ubit Umarov: again, at least so far
[12:06] Misterblue Waves: you mean running on desktop mono doesn't grow, but running on server mono does?
[12:06] TG.Lucan Ubit, is that both linux and win?
[12:06] Bill.Blight I have seen using the --server switch grow much larger in mem usage due to the way it handles the GC
[12:07] Bill.Blight my mono env values ..........................export MONO_GC_PARAMS="nursery-size=32m,promotion-age=14,minor=split,major=marksweep,no-lazy-sweep,alloc-ratio=50"
export MONO_GC_DEBUG=""
export MONO_ENV_OPTIONS="--desktop"
[12:07] TG.Lucan I run both --server --debug, don;'t see grow
[12:08] Ubit Umarov: well bin/ has a simpler version :)
[12:08] Bill.Blight yes but as I said could be platform specific , I don't see it grow on ubutu 18 even with --server
[12:08] Ubit Umarov: ( and )
[12:08] Bill.Blight well it grows larger than --desktop but it levels out
[12:08] TG.Lucan Linux we are RedHat / Fedora
[12:09] Bill.Blight httptests runs better in release mode btw
[12:09] Ubit Umarov: well i don't remember in past, but all c# now runs significatly faster in release than debug
[12:10] Ubit Umarov: and that also has impact on peak memory use
[12:10] TG.Lucan yes, but if there are any exceptions, you need the output
[12:10] Bill.Blight But I guess I should qualify my testing, I use a flavor of httptests
12:11] Ubit Umarov: yes tg it is a prob.. but debug mode is no longer nice on running regions
[12:11] Ubit Umarov: JIT optimizations are all off etc etc
[12:11] Ubit Umarov: and that does mean a lot now
[12:11] TG.Lucan We will recieve the hg safari the next couple of weeks, maybe that will show more
[12:12] Ubit Umarov: in past JIT was not that good.. no no big diff i guess
[12:13] Bill.Blight I've had the safari over a few times and they have not manged to kill me yet ... LOL , not for lack of trying ... HAHA
[12:13] Ubit Umarov: ms finally noticed that c# has no future if has bad performance
[12:13] TG.Lucan We run the public test grid on a laptop, 32 bits 2GB.. very sensitive to stuff
[12:14] Bill.Blight eek
[12:14] Bill.Blight LOL
[12:14] TG.Lucan And safari have been there before
[12:14] Ubit Umarov: well 32bit uses less memory and is faster
[12:14] TG.Lucan sure, we have fine tuned the Linux.. thats our thing
[12:15] Bill.Blight but does not handle the load or physics load over long periods as well as 64 bit, or regions with a ton of stuff ..
[12:15] Ubit Umarov: but the new JIT only works on 64bit, so 64b maybe faster on somethings now
[12:15] tx.0h how you have fine tuned the linux?
[12:15] Ubit Umarov: think on 4.7.2 the new jit will also do 32b
[12:16] Bill.Blight yeah my lixux networking config does not even resemble stock
[12:16] Ubit Umarov: because by default on win ANYCPU means 32bit
[12:16] TG.Lucan Minimise overheads, streamlined thread cleanup etc
[12:16] Bill.Blight my sysctl.conf is a mile long
[12:17] Ubit Umarov: ( httptests prebuild does change that to be the machine native size )
[12:17] TG.Lucan Linux 4.1.13.. a few years old now
[12:17] Ubit Umarov: i mean the win native size
[12:18] Ubit Umarov: but i did run a copy of wright plaza in 32bit with room to spare
[12:19] Ubit Umarov: ( but a 35k prims region should run in 64bits )
[12:19] Bill.Blight one thing to note, which may or may not affect it on linux, is windows default tcp ttl is 128 most linux is way lower than that, I have see increasing it to the windows level has stopped timeouts
[12:20] TG.Lucan dunno about that :)
[12:21] Ubit Umarov: possible on ubuntu messed up tcp stack :p
[12:21] TG.Lucan One thing I will say, is the public port private port is a bit silly at present
[12:22] Bill.Blight ummm why do you say that
[12:22] TG.Lucan Why not have private address as well?
[12:22] Ubit Umarov: bc no reason for that
[12:22] TG.Lucan It is a network thing
[12:22] TG.Lucan well there is ubit
[12:22] Bill.Blight why waste an address
[12:23] Ubit Umarov: and private/public is on a grid context only
[12:23] TG.Lucan well it means you are echoing all your local traffic via the loopback
[12:23] Bill.Blight how so?
[12:23] TG.Lucan and you are opening up the private port to allow the loopback.. nuts
[12:24] Ubit Umarov: toy mean hg ?
[12:24] TG.Lucan ok if you have regions external to robust like OSgrid
[12:24] Ubit Umarov: what loopback ?
[12:25] Ubit Umarov: ok i correct
[12:25] TG.Lucan public interface returns request to the local yes?
[12:25] Ubit Umarov: and private/public is on closed grids context only
[12:25] Ubit Umarov: NO
[12:25] TG.Lucan as I said a network thing
[12:25] Ubit Umarov: means ports that should not be visible to world on a closed grid
[12:26] Ubit Umarov: ie ports only for regions connections
[12:26] Bill.Blight Each service responds on their respective ports that they were asked the question on and to the IP they were asked the question on, they only go to loopback if they are contacted via the loopback interface
[12:26] Ubit Umarov: that private/public only makes sense in closed grids context
[12:27] TG.Lucan ok.. could be very long discussion. maybe another time
[12:27] Bill.Blight I traffic monitor my ports on my robust and my loopback has almost no traffic, so not seeing what you are saying
[12:27] Ubit Umarov: and you can have dif ips also
[12:27] Bill.Blight ports and interfaces I should say
[12:28] TG.Lucan one sec
[12:28] Ubit Umarov: those public/private things are only helpers
[12:28] Ubit Umarov: and those const things
[12:28] Bill.Blight if you have regions on that local machine and you bind them to and not the public IP you may see traffic on your loopback interface
[12:29] TG.Lucan what ip address do you use for private ports?
[12:29] Bill.Blight I use the same as for the public
[12:29] TG.Lucan there is the point
[12:29] Bill.Blight not the public ports that is an entirely different discussion
[12:29] TG.Lucan not required
[12:29] Bill.Blight umm
[12:30] Bill.Blight it is if you have 6 servers running regions all over the world
[12:30] TG.Lucan you set those to the local address, assuming it is on a local network
[12:30] Ubit Umarov: yes TG you can do hat
[12:30] Ubit Umarov: that on a closed grid
[12:30] Bill.Blight bind your regions to exteral IP not loopback
[12:30] TG.Lucan if the whole thing is public facing, that is different
[12:30] Ubit Umarov: but loopback is confusing
[12:30] Bill.Blight the whole thing is not
[12:31] Bill.Blight private ports are firewalled
[12:31] Ubit Umarov: its not going to the firewall
[12:31] Ubit Umarov: so yes.. you can set those on an internal to grid only IP also
[12:32] TG.Lucan another time..
[12:32] Bill.Blight I love a good geek speak talk
[12:33] TG.Lucan I want to go listen to music :)
[12:33] Ubit Umarov: as i said those names public/private and those const things are just helpers
[12:33] TG.Lucan tips cap
[12:33] Andrew Hellershanks: TG.Lucan, I have a Bluegrass Internet station playing in the background here so I've also been listening to music. :)
[12:35] Ubit Umarov: guess its all over now :)
[12:35] Bill.Blight I just don't get what he was saying
[12:35] Ubit Umarov: what?
[12:35] Bill.Blight if he is seeing that much loopback traffic , he has something setup wrong, IMHO
[12:35] Ubit Umarov: the loopback thing?
[12:35] Ubit Umarov: read firewall
[12:36] Bill.Blight yeah sounds like he could be redirecting all requests to the FW which is pushing it back to the loopback
[12:36] Bill.Blight I don't let my FW route, only block and allow
[12:36] Ubit Umarov: that will happen if using same port
[12:37] Ubit Umarov: and on a closed grid no need to send packets to the firewall at all
[12:37] Bill.Blight well, sounds like maybe since he has everything on one system he is talking on the loopback to the private port
[12:37] Ubit Umarov: so yeap private things can also have a internal private ip
[12:38] Bill.Blight I'm unclear if he means loopback interface or loopback traffic
[12:38] Ubit Umarov: well i stopped reading loopback and did read firewall :)
[12:38] Ubit Umarov: bc loopback nic makes no sense
[12:38] Bill.Blight nope
[12:39] Bill.Blight my FW actually does not allow any loopback traffic
[12:39] Bill.Blight it allows and blocks, but I don't run regions on my robust server
[12:40] Ubit Umarov: some do run dedicated firewalls :p
[12:40] Ubit Umarov: not iptable
[12:40] tx.0h normaly fw should be dedicated. but it's not a must.
[12:40] Ubit Umarov: cisco etc do sell those beasts
[12:40] Bill.Blight well even then you can store and forward if you have a good firewall,
[12:41] Bill.Blight I run a linux firewall, standalone here , have two Cisco PIX in the closet, never use them
[12:41] Ubit Umarov: yes but point is that on that case those "private" robust things can also have a private ip
[12:41] tx.0h ipfire
[12:41] Bill.Blight ipfire all the way
[12:41] Ubit Umarov: so nothing hits the firewall box
[12:41] tx.0h but i use fw on my router (openwrt) which also does nat loopback
[12:41] Ubit Umarov: regions connect directly to robust
[12:42] Ubit Umarov: ( can be a intranet ip also )
[12:42] Bill.Blight of course I'm a strange critter, I have many public IP's at home
[12:42] Bill.Blight but I don't run any servers from home anymore, just the occasional region
[12:43] Ubit Umarov: if there are regions outside the firewall allowed to connect also then no point on having a dif ip
[12:43] Misterblue Waves: I did speak with John Hurliman a few weeks ago... he's back in the US after a time in England (this being Hurliman plaza)
[12:43] tx.0h with 10mbit upload, what else should i do with that?
[12:43] Bill.Blight LOL
[12:43] Ubit Umarov: ohh nice
[12:44] Ubit Umarov: yeah it should be Umarov plaza by now.. but nahh
[12:44] Bill.Blight 10 mb can hold a few small regions or ONE big one
[12:44] Ubit Umarov: hihihihi
[12:45] Bill.Blight I should bring my OSG region back up just for giggles
[12:45] Bill.Blight I used to keep an attached region from one of my servers up all the time. but .... meh
[12:45] Bill.Blight Now I keep a hi-fi region up
[12:46] Ubit Umarov: hmm had no more news on those tp fails
[12:46] Bill.Blight nope, networking issue I bet
[12:47] Ubit Umarov: i did fly and tp on my testset even not waiting and no issues
[12:47] Ubit Umarov: but well empty regions.. same 100mbps net
[12:48] Bill.Blight I tested on my grid, multiple regions multiple servers, could not make it happen, broke other stuff , but not that
[12:48] Bill.Blight LOL
[12:48] Misterblue Waves: I'm looking to spin up regions in Docker containers.. it would be easier if there is always a known execution and networking environment
[12:48] Ubit Umarov: but on those reports client had not done lludp "connection"
[12:48] Bill.Blight there is a guy that made an opensim docker
[12:48] Ubit Umarov: dahlia uses docker
[12:49] Bill.Blight
[12:49] Misterblue Waves: I'm thinking that's a viable solution for the 'standard' running setup... especially now that there's Docker for Windows and Docker environments in all the clouds
[12:50] Ubit Umarov: dunno.. didn't played with that still
[12:50] Bill.Blight not a fan of virtual environments for opensim
[12:50] Bill.Blight but they have their uses
[12:50] Ubit Umarov: its a extra layer
[12:50] Bill.Blight yep
[12:51] Ubit Umarov: but more isolation also etc
[12:51] Ubit Umarov: ( and can do mono32bit on a 64bit box :p )
[12:51] Bill.Blight yes it is better for isolation, but I just don't like that extra software layer .. Opensimulator is touchy enough don't want to make it more so
[12:52] Misterblue Waves: the main advantage for us would be a single execution environment -- we wouldn't be trying to stay compatible with Windows 10, Ubuntu, CentOS, etc and all the different versions of same
[12:52] Ubit Umarov: yeap its extra overhead
[12:53] Misterblue Waves: for Docker, the extra layer is between the OS and the app... it uses cgroups and other Linux isolation stuff to create the container
[12:53] Bill.Blight well, from my old NOC days, stay updated or get left behind ... If staying updated is a problem , get out of the business ...  :P
[12:54] Bill.Blight which speaks to the current state of the dev box
[12:54] Misterblue Waves: OpenSim has always had the problem of being behind because there is always someone trying to run it on 32bit Windows95 or something
[12:54] Ubit Umarov: you had to remember that also :p
[12:54] Ubit Umarov: we should not run on top of tops but we stayed still 4 2 long
[12:55] Bill.Blight yeah but you can use cggroups and such without needing docker as well
[12:55] Ubit Umarov: even when moving to 4.0 the box stayed with mono 2.10
[12:55] Bill.Blight Docker just makes it so you don't need to know how to configure that stuff
[12:57] Ubit Umarov flüstert: and its something else ppl needs to install :)
[12:57] Bill.Blight hate to be blunt but staying that far behind is just plain LAZY
[12:57] Misterblue Waves: There needs to be some changes to OpenSim to make console-less installation possible (like all the queries about region groups) but it would be a good thing in the long run, I think
[12:57] Misterblue Waves: we're going to have to leave some people behind... like the people still running 8.2
[12:58] Ubit Umarov: think there was an old project on that
[12:58] Bill.Blight I just run in screen, gives me the best of both worlds
[12:58] Ubit Umarov: aurora also got a gui installer i think
[12:58] Misterblue Waves: fork and run your own grid or stay updated!!
[12:58] Bill.Blight yep
[12:58] Ubit Umarov: and of course there are the other distros
[12:58] Ubit Umarov: like fred's one
[12:59] Bill.Blight there is going to come a point where the current code is not at all backwards compatible from a communication standpoint and then it is either update or watch opensimulator die
[13:00] Bill.Blight why you got to use dirty words like that to me Ubit
[13:00] Bill.Blight I'm not a fan of that windows installer thing
[13:00] Misterblue Waves: to keep it alive, I do think we need to savage OpenSim -- get rid of old, unused modules, old stats gatherers (personal sore spot), all the different communication schedulers and options,...
[13:00] Ubit Umarov: yeap
[13:00] Bill.Blight All the "one click" installers do is create support nightmares because nobody actually knows what they are doing ..
[13:01] Ubit Umarov: but some of those still have fans.. grrrr
[13:01] Bill.Blight I totally agree, needs a good spring cleaning
[13:01] Ubit Umarov: and a damm clean region/robust API
[13:01] Ubit Umarov: what a mess we have now
[13:01] Bill.Blight OMG yes
[13:02] Ubit Umarov: and a damm BINARY serialized
[13:02] Ubit Umarov: damm folish xml..
[13:02] Bill.Blight hell some of the other core devs do not even know how the Robust/Region/Estate thing works
[13:03] Misterblue Waves: my personal wish is to recode the core of OS to be a distributed simulation engine and only keep the existing communication part since the LL protocol is what people really want
[13:03] Ubit Umarov: peek mem use of 400MB to send 36k inventory itens drives me mad
[13:03] Bill.Blight Not going to mention names, but one core dev said that you can't have centralized estate management, that it does not work ... umm been using it for a long time , I beg to differ
[13:04] Misterblue Waves: I shouldn't admit it, but I HATE the plugin system that Mel and Diva added for the robust stuff -- WAY too complicated with many little exceptions
[13:04] Misterblue Waves: it's the worst of all worlds: lots of moving parts can configurations without really creating a clean layering
[13:04] Ubit Umarov: and diva did not follow it on groups :p
[13:05] Bill.Blight well , when you have things that are "core" they should not be plugins, the ability to be overridden by a plugin yes, but they should be core not plugins
[13:06] Ubit Umarov: there was a "modules" fever
[13:06] Bill.Blight yep
[13:06] Bill.Blight everything became a module
[13:06] Ubit Umarov: things like temporaryattachment module
[13:06] Ubit Umarov: nonsense
[13:06] Bill.Blight yep
[13:06] Bill.Blight why the hell is that a module
[13:06] Ubit Umarov: BS
[13:07] Ubit Umarov: and why is lightshare a full script API ??
[13:07] Bill.Blight oh I like that
[13:07] Bill.Blight hehe
[13:07] Ubit Umarov: why was it not just part of ossl
[13:07] Ubit Umarov: etc etc etc
[13:07] Bill.Blight but should be core
[13:07] Bill.Blight yes
[13:08] Ubit Umarov: modules diarreia to be nasty :p
[13:08] Bill.Blight and it sure does smell bad
[13:08] Bill.Blight also .. umm if it is not enabled, it should not be loaded
[13:08] Ubit Umarov: that attachments is one im considering removing for ages
[13:08] Bill.Blight we need an ini that tells it what modules to load, and what not to load
[13:09] Bill.Blight so it does not load everything under the sun
[13:09] Bill.Blight no reason for every physics module to get loaded at runtime if you are/can only use the one
[13:09] Ubit Umarov: well that also means an explosion of dlls
[13:09] Misterblue Waves flüstert: I'm currently fighting that Mel added a hard dependence on MySQL in one of her recent additions <frowny-face/>
[13:09] Ubit Umarov: also not that good
[13:10] Ubit Umarov: what hard dep ?
[13:10] Bill.Blight MySQL for the win ...
[13:10] Bill.Blight loves my , MySQL
[13:10] Ubit Umarov: hm but what dep?
[13:10] Misterblue Waves: she added something for groups, I think, with the comment 'only works with MySQL' and she had coded it against the MySQL lib rather than the usual database access layer
[13:11] Misterblue Waves: I'll have to look... give me a sec
[13:11] Ubit Umarov: i don't remember that
[13:11] Bill.Blight not sure if that was Mel
[13:11] Ubit Umarov: Ahhh
[13:11] Ubit Umarov: she means no SQlite
[13:11] Ubit Umarov: i guess
[13:11] Ubit Umarov: SQlite does not have groups i think
[13:12] Ubit Umarov: it is not "mysql" runs on maria also
[13:12] Ubit Umarov: and possibly psg
[13:12] Ubit Umarov: just SQlite does not support all features
[13:13] Bill.Blight oh the mute module
[13:13] Ubit Umarov: i mean no one added support
[13:13] Misterblue Waves: it's the money module (has your name on it, Ubit): change SampleMoneyModule intialization, make it look for economymodule setting both in startup and economy sections; ch...skipping... drop. It's classic Avination, tied to MySQL and a direct database connection and I don't have the bandwidth to fix it to core methods. However, it has all the strange mute logic that OpenSim was missing and maybe someone can step up to make it work in an open grid context.
[13:13] Bill.Blight only works on mysql
[13:14] Ubit Umarov: well mutes have a full module now
[13:14] Ubit Umarov: but no sqlite also i think
[13:14] Ubit Umarov: btw i don't remember that commit
[13:15] Ubit Umarov: i was talking about the money or mutes ?
[13:15] Bill.Blight that is text above from two different commits
[13:16] Bill.Blight
[13:16] Bill.Blight that one and
[13:16] Bill.Blight
[13:16] Bill.Blight that one
[13:17] Ubit Umarov: that was the Xmutes
[13:17] Ubit Umarov: now removed
[13:18] Ubit Umarov: Xmutes was just a temporary thing
[13:18] Ubit Umarov: httptests now does have a full mutes module
[13:18] Ubit Umarov: with its damm 15 files needed for a robust module
[13:19] Ubit Umarov: it is also on master, but there under other name
[13:19] Ubit Umarov: MutesModuleTst if i remember
[13:19] Ubit Umarov: see it MB ?
[13:19] Ubit Umarov: xmutes are gone..
[13:20] Misterblue Waves: sounds like I have to move to httptests and get off master :)
[13:21] Ubit Umarov: well im using it as the dev branch now
[13:21] Ubit Umarov: master is the bugs fix for 0.9
[13:21] Ubit Umarov: that we should actually release
[13:22] Ubit Umarov: i didnt remove xmutes also out of master ?
[13:24] Ubit Umarov: hmm and mutes where a feature req from our Sheera.Khan here :)

Personal tools
About This Wiki