Chat log from the meeting on 2017-05-30

From OpenSimulator

Jump to: navigation, search

[11:05] Selby.Evans IMA Proposed Working Group: Test Management
[11:05] Selby.Evans
[11:06] SkyFlier.NorthStar i downloaded and compiled the master from last week and it worked fine
[11:06] Gavin.Hird great
[11:07] Gavin.Hird I diffed it with RC2 today and they now look almost identical
[11:07] SkyFlier.NorthStar a bit laggy at times but eventually everything rezzes'
[11:07] Gavin.Hird 3 additions and 1 removal since the RC2 was built, but that should be updated for the final version
[11:08] SkyFlier.NorthStar what is the date of the RC2?
[11:08] Gavin.Hird at the top of my head at least 6 weeks back
[11:09] SkyFlier.NorthStar ok
[11:09] Arielle Popstar: nods
[11:10] Arielle Popstar: looks like andrew and ubit are mia
[11:11] Bill.Blight Ubit is having nightmares about mono and DNS ....
[11:11] SkyFlier.NorthStar ubit might have been working late hours
[11:11] Gavin.Hird I have noticed
[11:11] Arielle Popstar: heh yes he was telling me that yesterday
[11:11] Gavin.Hird I have not dared to take the latest updates
[11:11] Arielle Popstar: assuming that is the new mono versions
[11:12] Bill.Blight Mono 5 appears to run fine, but still suffers from very high mem usuage
[11:12] Gavin.Hird does master compile on 5-0?
[11:12] Bill.Blight yes
[11:12] Gavin.Hird ok
[11:13] Bill.Blight Ubit has been testing on my linux box with 5
[11:13] Gavin.Hird ok
[11:13] Leighton.Marjoram quick question what is the issue with mono and DNS?
[11:14] Gavin.Hird glad you asked Leighton because that beats me too
[11:14] Bill.Blight well dns caching .. although Ubit just discovered that it may be linked to an old linux issue around resoved
[11:14] Arielle Popstar: something about a lack of caching the addresses
[11:14] Bill.Blight no too much caching
[11:14] Gavin.Hird It would be good to have a case scenario to test
[11:15] Leighton.Marjoram Is that the same issue raised by Shaun Emerald of Sanctuary grid
[11:15] Bill.Blight disabling the resolvd service on linux seems to mitigate the issue
[11:15] Arielle Popstar: he mentioned it on ops virtual yes
[11:16] Leighton.Marjoram thanks Arielle
[11:16] Bill.Blight the addresses are being cached for too long, so if dyndns is involved you end up with a discrepancy
[11:16] Gavin.Hird disabling resolvd will not the rest of the system lose name resolution?
[11:16] Selby.Evans hi Andrew
[11:16] Andrew.Hellershanks Hello, everyone. Sorry I'm late. I side tracked myself on something. :)
[11:16] Bill.Blight not if you have your dns-servers set in your networking config
[11:16] Bill.Blight resolved is the caching dns service
[11:16] Gavin.Hird ok
[11:17] Bill.Blight
[11:18] Gavin.Hird Has anyone run OpenSim in Docker?
[11:18] Bill.Blight there were some other command line tweaks and env variables we were testing with mono on linux that seems to give some better peformance
[11:19] Andrew.Hellershanks What is Docker?
[11:19] Bill.Blight most of them were posted to my mantis about high mem usage
[11:19] Gavin.Hird Docker is a light weight virtual machine of sorts
[11:19] Gavin.Hird it runs a linux at the core
[11:19] Andrew.Hellershanks ok, ty. It made me think of something else. I've used virtualbox for a VM
[11:19] Basil Sosides: light Whight is Healthy
[11:20] Gavin.Hird
[11:20] Andrew.Hellershanks BTW, If I disappear unexpectedly it is likely due to the thunderstorms in my area.
[11:21] SkyFlier.NorthStar where are you located Andrew?
[11:21] Arielle Popstar: oh thunderstorms coming?
[11:22] Arielle Popstar: so docker doesnt work for windows?
[11:22] Andrew.Hellershanks Arielle, yes.
[11:22] Gavin.Hird it runs on Linux, Windows and macOS
[11:22] Andrew.Hellershanks Yes, re: thunderstorms.
[11:22] Arielle Popstar: (sunny and blue skies here atm)
[11:22] Arielle Popstar: ok Gavin
[11:23] Arielle Popstar: is it free?
[11:23] Gavin.Hird you pack an application with all it needs into a docker container, that is deployed to one of those systems
[11:23] Gavin.Hird it is free
[11:26] Arielle Popstar: ok
[11:26] Arielle Popstar: so Bill, does that change for mono fix the issue or why is Ubit still having nightmares?
[11:26] Andrew.Hellershanks I received an offline IM that a Permissions Testing Workgroup has been initiated but hasn't started work yet.
[11:27] Arielle Popstar: Andrew, what about the automated testing config that Diva was working on?
[11:28] Arielle Popstar: any word on it?
[11:29] Andrew.Hellershanks I haven't heard anything recently from diva as to how the automated testing framework is progressing.
[11:30] Selby.Evans The WG is forming -- we have 3 members
[11:30] Selby.Evans IMA support not ready yet
[11:31] Gavin.Hird Pls remind us what the purpose of the group is?
[11:31] Selby.Evans I am posting general plans in my blog.
[11:31] Arielle Popstar: take testing load off of the developers
[11:32] Selby.Evans Just a min
[11:32] Selby.Evans WG: Test management
Objective: support development of open source software for user-built virtual worlds.
Need: testing needed to characterize bugs, evaluate alternatives, and validate betas.
Who: The OpenSim community can do this. (We, the users.)
Rationale: We need not and should not put the burden on the developers.
Support: IMA will undertake to organize and support an OpenSim Working Group for test management.
Lead: Selby Evans will undertake initial organization of this WG.
Selby hopes to turn the lead over so someone more capable in the future.
Objective: To design and implement a test capability to support the development and selection of open sim and related code.
[11:32] Selby.Evans Expected output (deliverables)
A document giving the categories of tests required for the cited development
Road-maps showing how to execute tests in each category
Approval of these items by developers and other users of the test results
Formation of a self-maintaining tester team intended to continue indefinitely.
Output may change by decisions of the WG
[11:34] Gavin.Hird thanks Selby
[11:34] Arielle Popstar: in what way will IMA support such an entity?
[11:36] Selby.Evans just a min
[11:37] Selby.Evans What will IMA do?

Publicize tests under development and tests needed
Archive and curate tests and test protocols ready for use.
Archive and curate tests results.
Seek to become the go-to place to find tests useful in OpenSim.
[11:38] Gavin.Hird code tests
[11:38] Gavin.Hird user tests
[11:38] Andrew.Hellershanks It will be a big help to take some of the repetetive testing load off the shoulders of the developers.
[11:38] Gavin.Hird performance tests
[11:39] Leighton.Marjoram will the curated and archived tests be publicly searchable
[11:39] Selby.Evans As needed, handle date collection for tests
[11:39] Gavin.Hird to test the code you have to work very close with the developers
[11:39] Leighton.Marjoram nods @Gavin
[11:40] Selby.Evans All IMA curated content will be public
[11:40] Leighton.Marjoram thank you @Selby
[11:40] Bill.Blight Will these tests be "laboratory" or "Real World" tests, many things will preform well in a "perfect" environment ... If the tests do not survive contact with the end users then they become a moot point ..
[11:41] Arielle Popstar: agreed Bill
[11:41] Leighton.Marjoram with real human controlled avatars (not bots)
[11:41] Gavin.Hird + much of the test needs a significant amount of kit and users to flesh out the issues
[11:41] Arielle Popstar: thinking especially for example between .8.2.1 and .9.0
[11:41] Leighton.Marjoram simulated use can be useful initially but real world use is much better data
[11:42] Selby.Evans for QA tests, the coders will supply test scripts. We will only run the tests and collect data.
[11:42] Bill.Blight Just my two cents, you cannot quantify the X factor, which is the end user, so spending too much time in the laboratory and not in the real(virtual)World, you can just end up spinning your wheels ..
[11:42] Gavin.Hird not sure how that takes the load off the developers?
[11:43] Gavin.Hird usually you build your unit tests and make it part of the automated build procedure
[11:43] Selby.Evans details of operation will be TBD by the test group. To offer advice, join the test group.
[11:43] Andrew Hellershanks: Gavin, the developers won't have to spend time running some of the tests every time code changes.
[11:44] Gavin.Hird it runs automatically if you set it up the right way
[11:44] Selby.Evans I understand Diva is writing test scripts for the permissions problems.
[11:45] Arielle Popstar: thing with Ubit is that he makes changes and tests immediately to see the results
[11:45] Gavin.Hird but complex tests where you involve mutliple versions of OpenSim in a real grid running on different operating systems and with users doing unexpected things, thats what help is needed on
[11:45] Andrew Hellershanks: Some tests of OS require many avatars to do the tests under load to see how the code holds up. Some tests are not impacted by a system load and either show something works or doesn't work as expected.
[11:45] Andrew Hellershanks hears thunder (again)
[11:46] Andrew Hellershanks: That thunderstorm has gotten a lot closer.
[11:46] Selby.Evans If the devs don't want tests we won't do them. There are comparison tests to do and problem analysis
[11:46] Arielle Popstar: bottom line is that we need cooperations from the devs
[11:47] Arielle Popstar: with some idea of expected behaviour on changes
[11:47] Gavin.Hird All I'm saying there is a whole class of tests that handing over to someone else is more hassle for the developers than doing it themselves
[11:47] Selby.Evans I have an informal test going now
[11:47] Andrew Hellershanks: Selby, Diva was writing some automated tests that will check for object related perms testing, IIRC.
[11:48] Bill.Blight Something Leighton and I are tossing around, is an Oar file with scripts, physics and perms testing objects in the oar, to test on TOP of the code, not as the code compiles ... Something you can just load up and have everything you need for "In-world" tests.. A little less code oriented and more end-user focused on how it actually behaves ..
[11:48] Selby.Evans If the devs dont want us to do tests, we won't do them
[11:49] Arielle Popstar: good idea Bill and Leighton
[11:49] Leighton.Marjoram smiles @Bill
[11:49] Andrew Hellershanks: That is a good idea, Bill.
[11:49] Gavin.Hird There is a lot ot end user tests that can be done including viewer behavior
[11:50] Selby.Evans .we are plannning to have OAR with test suites
[11:50] Arielle Popstar: i really like that oar idea thing
[11:50] Bill.Blight Yeah and that would be the focus of what we are doing .. End users don't care about the "code" they care about their experience .. and at the end of the day that is what is going to make or break opensim ..
[11:50] Gavin.Hird Developers are not always very good at testing end user cases as they have too intimate knowledge of the code, so they will test in different patterns and even avoid known issues
[11:51] Gavin.Hird so a test group detached from the developers could be good in that respect
[11:52] Bill.Blight I have that issue Gavin, I tend to accept bugs for what they are, and even avoid some without conscious thought ...
[11:52] Gavin.Hird of course giving feedback both in terms of recommendations, feature requests and bug reports
[11:52] Selby.Evans The Idea is to have OARs and test scripts archived and curated for future use.
[11:53] Selby.Evans Quality Assessment is never done by the developers
[11:53] Arielle Popstar: so the support of IMA is mostly in handling the data output
[11:54] Selby.Evans And any user can do tests for their own purposes
[11:56] Andrew Hellershanks: Developers can do basic QA but they can miss things.
[11:56] Bill.Blight Quality assessment is also never very useful if it is done from the 1000 mile view, it needs to be at ground level, with the people that actually use the product.
[11:56] Selby.Evans IM support is in making the test resources available (oars, test scripts, etc, managing data collection, and archiving results.
[11:57] Andrew Hellershanks: Having an OAR file will make it easier for others to do testing using a standard set of objects and scripts.
[11:57] Gavin.Hird when you say test scripts, do you mean in-world lsl/ossl scripts?
[11:58] Andrew Hellershanks: Gavin, I did.
[11:58] Bill.Blight That is what I am talking about Gavin
[11:58] Gavin.Hird ok
[11:59] Andrew Hellershanks: ok, Selby. Thanks for what you are doing with the workgroup.
[11:59] Bill.Blight As we know the code can be ugly and still work well for the end-user, so there needs to be someplace to meet in the middle, and without data from both sides you can end up with something that works great and people hate. .
[12:00] Andrew Hellershanks nods
[12:03] JayR Cela: I have been helping to develop a new grid running 0.9RC2 on Ubuntu server 16.04........
[12:03] Arielle Popstar: newer mono
[12:03] Bill.Blight (my entire grid runs on MASTER, mono 5)
[12:03] JayR Cela: when we try to log into the grid with the Singularity Viewer we keep getting Packet out of Order errors ...
[12:04] Gavin.Hird what version of singularity?
[12:04] JayR Cela: expected packet 10xxxx/ recieved packet 11xxxx
[12:04] JayR Cela: etc...
[12:04] Gavin.Hird cause it is getting seriously outdated in terms of functionality on master
[12:04] JayR Cela: all versions of Singularity
[12:04] JayR Cela: including the Alpha's
[12:04] JayR Cela: and the old 32 bit version as well
[12:05] Andrew Hellershanks: JayR, what version of mono are you using?
[12:05] Bill.Blight You may need to modify your sysctl.conf to optimize the networking .. by default it is really turned down in Ubuntu
[12:05] Gavin.Hird for all practical purpose any version of Singularity is old
[12:05] JayR Cela: you can Guest TP into the Grid and it works fine
[12:05] JayR Cela: just a native login is when We see the Packet out of Order errors
[12:07] Bill.Blight this may help some what ...
[12:07] JayR Cela: thaks Bill :_)
[12:07] Bill.Blight ubuntu 16 networking is a little anemic by default
[12:08] JayR Cela: :_)
[12:08] Bill.Blight this is how I have mine set ..
[12:08] Bill.Blight net.core.wmem_max=12582912
net.ipv4.tcp_rmem= 10240 87380 12582912
net.ipv4.tcp_wmem= 10240 87380 12582912
net.ipv4.tcp_window_scaling = 1
net.ipv4.tcp_timestamps = 1
net.ipv4.tcp_sack = 1
net.core.netdev_max_backlog = 5000
net.ipv4.tcp_keepalive_time = 3600
net.ipv4.udp_mem = 10240 87380 12582912
net.ipv4.udp_rmem_min = 32768
net.ipv4.udp_wmem_min = 32768
[12:09] JayR Cela: ok / thanks again Bill
[12:09] Bill.Blight No problem
[12:11] Andrew Hellershanks: Any other topics for today before everyone vanishes? :)

Personal tools
About This Wiki