Simulator User Group/Transcripts/2011.10.28

From Second Life Wiki
Jump to: navigation, search


Prev 2011.10.25 Next 2011.11.01

List of Speakers

DMC Zsigmond Falcon Linden Flip Idlemind
Imaze Rhiano Jonathan Yap Kadah Coba
Kallista Destiny Kelly Linden Latif Khalifa
Questar Utu Renae Daines Sahkolihaa Contepomi
Simon Linden Sniper Siemens Sopherian Yumako
Tankmaster Finesmith Tiny Pumpkin Vincent Nacon


[16:04] Sahkolihaa Contepomi: Hey Simon.

[16:04] Kadah Coba: Eh?

[16:04] Kallista Destiny: Hey Simon

[16:04] Simon Linden: HI there ... sorry to be late, I got stuck fighting the beta viewer updater

[16:05] Sahkolihaa Contepomi: Haha

[16:05] Vincent Nacon: whoa.... PG sim PG! Kadah

[16:05] Sopherian Yumako: HAI2U!!!1!

[16:05] Vincent Nacon: don't go humping the face of a Firestorm developer

[16:05] Sopherian Yumako: wow are u a cat?

[16:05] Questar Utu: Simon's a kitty. o.o

[16:05] Sopherian Yumako: thats cute i love cats

[16:05] Simon Linden: Right, I have a mesh cat AV. If you see a blob with a head, you should try a newer viewer

[16:06] Imaze Rhiano: cute black kitty!!!

[16:06] Sopherian Yumako: i see u fine

[16:06] Sahkolihaa Contepomi: Haha, the hat is awesome.

[16:06] Kallista Destiny: Cat in a hat

[16:06] Simon Linden: IT doesn't work very well with sitting or animations, but I'm sure you've all run into similar issues with odd AVs

[16:07] Simon Linden: Anyway, Andrew is out today, so I'm the ringleader

[16:07] Kallista Destiny: Yeah you need specailzed tiny/furry/tiny-furry animations

[16:07] Simon Linden: There isn't much news ... the RC updates went as planned this week and are detailed on the web

[16:07] Vincent Nacon: any sign of Falcon Linden?

[16:07] Imaze Rhiano: there are those tiny huds - but doesn't mesh allow you to change bone lenghts for tinies?

[16:08] Simon Linden: I've spent most of the week getting the next round of maintenance work together, which should hopefully hit an RC channel next week

[16:08] Simon Linden: That has Andrew's queued rezzing fixes in it, I believe

[16:09] Kallista Destiny: YEAH, dod you know which RC will get it>

[16:09] Vincent Nacon: wasn't it Magnum?

[16:09] Simon Linden: I can check, 1 sec...

[16:09] Kallista Destiny: I do hope it is.'

[16:10] Simon Linden: That's what went out on Wednesday :

[16:10] Latif Khalifa: Simon, was it you who made SVC-7379 change? (ER-1338)

[16:10] JIRA-helper:

[#SVC-7379] For group notices groupID is being sent in the AgentID field

[16:10] Simon Linden: Magnum is the most interesting, with some bug fixes, but those have been there before

[16:11] Kadah Coba: So is AgentID going to actually be the sending agent's ID?

[16:11] Simon Linden: That branch _should_ get promoted to the full grid next week, and another round of bug fixes into one RC channel on Wednesday

[16:11] Simon Linden: Yes Kadah, that notification will have the agentID in the sender

[16:11] Simon Linden: I forget the jira you mean, but it's some group notification message

[16:12] Simon Linden: Very old viewers may have a minor display bug ... they may not indicate a sender, and will just say "unknown" or something similar

[16:12] Kadah Coba: Awesome, a while ago I was running in to that problem of not having that in the notices.

[16:12] Jonathan Yap: The person who requested that (me) unrequested it a little while ago when I 1) found another way of doing the same thing and 2) it causes issues

[16:12] Tankmaster Finesmith: "an unknown group"

[16:12] Latif Khalifa: My take on is that the change should probably be backed out as it causes some side effects and is not used for fixing STORM-1653 for which it was originaly made

[16:12] JIRA-helper: [#SVC-7379] For group notices groupID is being sent in the AgentID field

[16:12] JIRA-helper:

[#STORM-1653] Group notices sent by muted residents are still displayed

[16:12] Kadah Coba: Will have it fixed for the next releae.

[16:13] Kallista Destiny: so the fix for SVC-5927 will be coming out on an RC next week, do you know which, and can I request Magnum if it's not assigned?

[16:13] JIRA-helper:

[#SVC-5927] Temp on Rezzed objects get queued

[16:13] Jonathan Yap: I am making a call to the name cache to get that ID

[16:14] Simon Linden: I'd have to dig deeper to make any opinion on it, but I'll mention it to Andrew

[16:14] Jonathan Yap: Let's wait to see if Simon will agree to backing out that change

[16:16] Vincent Nacon: I think that's more of a logical issue than a bug, Jonathan.

[16:16] Jonathan Yap: Vincent, that change breaks Radegast

[16:16] Vincent Nacon: I understand but the way how it's viewed on

[16:16] Jonathan Yap: There's no longer a need for it, too

[16:16] Simon Linden: I'll see if I can get 107 on Magnum. Otherwise, if you have some regions you own and want on a particular channel, I think Maestro or Oskar is the person to contact ... check the wiki page, there should be instrutions there

[16:17] Jonathan Yap: Vincent, not all viewers parse that packet the same way

[16:17] Vincent Nacon: might be hard opinion how it should handled when some notices could have been importance and related to the group

[16:17] Simon Linden: Next weeks fixes should help us with stability ... Kelly and others have been working on some of the back-end systems that seemed a bit worse off after the kernel upgrade

[16:18] Simon Linden: There were a bunch of fixes in those areas that should help restarts

[16:18] Jonathan Yap: I don't think Simon is watching chat

[16:18] Simon Linden: Teh llMinEventDelay bug has been fixed

[16:18] Vincent Nacon: guess not

[16:18] Simon Linden: I can't commit on backing that fix out myself ... I'll talk with Andrew about it

[16:19] Vincent Nacon: I'm not against it, just let you know.

[16:19] Simon Linden: He did the orignial work, I believe

[16:19] Jonathan Yap: Kadah, look at my change for storm-1653 to see how to get the AgentID

[16:20] Simon Linden: There's one performance fix in the next round that should help a bit with crowded regions

[16:20] Renae Daines: Simon, will this also help with Net Time sky rocketing with even just four people in the region?

[16:20] Simon Linden: Yes, that's exactly the area for the fix

[16:21] Kallista Destiny: wonderful

[16:21] Renae Daines: Perfect.. I think I have annoyed support with that this week asking for region moves and host changes

[16:21] Simon Linden: ... if those 4 people have a lot of scripts that take controls, the sim fps should be better

[16:21] Kallista Destiny: Oh, even better.

[16:21] Flip Idlemind: I was in a crowded region last night, and various things started failing. Couldn't save scripts in upload snapshots. Eventually got logged off because of a "stalled main region event poll". Related? Or something entirely different?

[16:21] Simon Linden: That's the basic problem now ... if there are a lot of scripts that take controls in use, and people are moving around a lot, the net time skyrockets

[16:22] Kallista Destiny: Like AO?

[16:22] Simon Linden: This now puts some sane limits in there, which should keep the sim fps higher

[16:23] Kallista Destiny: or scripts that are waiting for events to actually take control?

[16:23] Flip Idlemind: By "related?" I mean related to the stability fixes you're talking about

[16:23] Renae Daines: That is wonderful news... too bad we can't push that sooner rather than next week

[16:23] Simon Linden: Right Kallista, AOs and such are the worst offenders

[16:23] Kadah Coba: Jonathan, gCacheName->getUUID(legacy_name, agent_id); ?

[16:23] Simon Linden: As usual, I'm paranoid about existing content, but the limits aren't very tight

[16:24] Kallista Destiny: Renae there is a cycle of releases which gets volater at great cost (broken regions)

[16:24] Jonathan Yap: yes that's it Kadah

[16:25] Kadah Coba: Cool, I'll have to remember to make use of that later.

[16:25] Vincent Nacon: time to add AO into client-side support?

[16:25] Simon Linden: So Kelly - do you have any news?

[16:25] Imaze Rhiano: at least phoenix AO is working fine

[16:26] Simon Linden: I guess not ... so the table is open ... are there any issues to bring up?

[16:26] Simon Linden: fyi Falcon said he would try to make it later

[16:26] Kallista Destiny: Firestorm AO is even better

[16:27] Kelly Linden: I don't have anything exciting this week. Maybe next week.

[16:27] Imaze Rhiano: touch events are not allways working with attached meshes - anwyay to fix it?

[16:27] Flip Idlemind: For that matter, right-click-selecting attached meshes doesn't always work

[16:27] Tankmaster Finesmith: :D

[16:27] Simon Linden: If you have a repeatable case and a model you can share, it would be best to set up a jira

[16:27] Latif Khalifa: viewer is not capable of detecting touches to rigged meshes afaik

[16:28] Latif Khalifa: it's a viewer thing

[16:28] Simon Linden: It may have something to do with the mesh image vs. it's physical shape, but otherwise I don't think meshes are significanly different than other shapes when it comes to touches

[16:28] Kelly Linden: The shouldn't be.

[16:28] Simon Linden: The physics engine is used to detect it with an internal ray cast based on where you click

[16:28] Latif Khalifa: Simon, i think it's the viewer picking code that doesn't understand rigged meshes

[16:29] Flip Idlemind: I think that's the case, yes

[16:29] Flip Idlemind: Viewer picking

[16:29] Imaze Rhiano: maybe if mesh is covering SL default avatar mesh by using alpha clothing layer?

[16:29] Simon Linden: I don't know much about that end of it, but that sounds plausible

[16:29] Kallista Destiny: then there might be an issue with mesh with no physics shape?

[16:30] Latif Khalifa: i did some testing, the viewer would typically select whatever is behind the rigged mesh as if it wasn't there

[16:30] Meeter: Timecheck : User Group is half over

[16:30] Questar Utu: Would be nice to have a legitimate way to export your shape for use in mesh designing programs.

[16:30] Kelly Linden: I need to head out early today. Have a good weekend all.

[16:30] Jonathan Yap: Take care Kelly

[16:30] Latif Khalifa: take care Kelly :)

[16:30] Imaze Rhiano: bye kelly!

[16:30] Kallista Destiny: See you one Mondy Kerry

[16:30] Kallista Destiny: Monday

[16:30] Simon Linden: Good question, Kallista ... if that was drastically differnet, and my idea on how it works is correct (not a given) then it might cause that kind of problem

[16:30] Tankmaster Finesmith: tc kelly

[16:31] Simon Linden: Bye Kelly

[16:31] Sahkolihaa Contepomi: See you, Kelly.

[16:31] Renae Daines: Bye, Kelly

[16:31] Imaze Rhiano: Xentha - not sure if that is legimate - but at least phoenix was able to export your SL avatar mesh as OBJ

[16:31] Sniper Siemens: bye kelly

[16:31] Vincent Nacon: that was quick

[16:31] Questar Utu: Pretty much what I meant.

[16:31] Questar Utu: I'd rather not have to use a third party viewer to do it.

[16:33] Imaze Rhiano: ping

[16:33] Tankmaster Finesmith: pong

[16:33] Vincent Nacon: dong

[16:33] Kallista Destiny: nack

[16:33] Renae Daines: fizzle

[16:33] Simon Linden: lol

[16:33] Vincent Nacon: anyway

[16:33] Questar Utu: Baah.

[16:34] Simon Linden: I've spent most of today looking into some of the sad states our servers can get into when overloaded

[16:34] Sahkolihaa Contepomi: Fun. :p

[16:34] Tankmaster Finesmith: time to chase off the dust bunnies?

[16:34] Simon Linden: Their disk IO seems to hit a point when it all falls over ... just backs up and slows way down.

[16:34] Simon Linden: It may be the bunnies

[16:35] Jonathan Yap: Maybe some day you will be using SSDs

[16:35] Renae Daines: So what would fix that, Simon? Does it catch up eventually? (if for example, a region were to be restarted)

[16:35] Kallista Destiny: is that paging?

[16:35] Simon Linden: But we have some data pipelines that aren't very efficient ... some things might be moved to a side thread, for example, and that would keep the fps higher

[16:35] Vincent Nacon: or stacking sims less per server

[16:35] Renae Daines: It would, I would imagine, lessen the load on the I/O

[16:35] Simon Linden: Yes, it gets into a bad swap fight sometimes

[16:36] Simon Linden: We're looking at hte next generation of servers too ... and considering SSDs

[16:36] Jonathan Yap: Good news! I have an SSD and it made a huge difference in my pokey old computer

[16:36] Simon Linden: I with they were about twice as big as they are now for the cost, but taht's another rant

[16:36] Tankmaster Finesmith: PCI-e SSDs or 2.5"?

[16:36] DMC Zsigmond: Great News Simon, CCP Games in Iceland for EVE Online run SSDs

[16:36] DMC Zsigmond: Hi everyone :)

[16:37] Latif Khalifa: what is causing high disk io? lack of memory and swapping?

[16:37] Tankmaster Finesmith: you can get a 1TB SSD soon in 2.5" format

[16:37] Simon Linden: I'm not sure ... there are some conflicting trends, as we get optiosn for more cores and lots of ram, but the disk space shrinks if we look at SSDs

[16:37] Imaze Rhiano: CCP did do lot's of research about SSD - and published some of their results

[16:37] Sahkolihaa Contepomi: What file system are you using? ext2 or 3?

[16:37] Tankmaster Finesmith: what size of hard drives do you use now?

[16:38] Kallista Destiny: The way I see it, if a region has to start paging, it's hosed anyway

[16:38] Tankmaster Finesmith: howdy brett

[16:38] DMC Zsigmond: Koli, nice question

[16:38] Simon Linden: I forget exactly, but the size isn't a problem unless we let some old files hang around and take up too much space

[16:38] Tankmaster Finesmith: bye brett

[16:38] Kallista Destiny: Hi Brett

[16:38] Renae Daines: Any thoughts on dynamic ram disks? Not sure how volatile the data would be in that sort of scenario.

[16:39] Renae Daines: Or how much data/ram is being used for each region for that matter - no specifics on that data

[16:39] Simon Linden: There is plenty of space available for normal usage ... but we collect crash cores, for example, which are very large. If those aren't reguarly cleaned up, they can potentially fill a disk

[16:39] Tankmaster Finesmith: falcon is here now

[16:39] Kallista Destiny: I would suspect that the memory would be better put to use as memory for the region as opposed to paging.

[16:39] Simon Linden: Hey Falcon you have any news or things to share?

[16:40] Falcon Linden: Well, if everything goes to plan, Havok 2011.2 and llSetKeyframedMotion will go to RC next week

[16:40] Vincent Nacon: magnum?

[16:40] Simon Linden: Koli - I missed it, what was your question?

[16:40] Vincent Nacon: or all?

[16:40] Sahkolihaa Contepomi: What file system are you using on the servers? ext2 or 3 or something entirely different?

[16:40] Falcon Linden: Vincent: I don't know--I try to avoid that part of the process :)

[16:40] Falcon Linden: Andrew wrangles that sort of thing for me

[16:40] Vincent Nacon: muhaha!

[16:41] Vincent Nacon: k, I'll bug Andrew later

[16:41] Sopherian Yumako: lol

[16:41] Kallista Destiny: @simon, Un*x does that for apps , OMG you could quickly run a FS out of space. We used to go through aour user Filespace are clean out core files.

[16:41] Falcon Linden: Wait, no, it was in an e-mail! Magnum.

[16:41] Simon Linden: I think it's ext2, but I wouldn't bet lunch on it

[16:41] Vincent Nacon: knew it

[16:42] Sahkolihaa Contepomi: 2? If so, then ouch...

[16:42] Vincent Nacon: k

[16:42] DMC Zsigmond: Ah, Ext 2

[16:43] Kadah Coba: Its not like they update them often :v

[16:43] Simon Linden: We have not been really aggressive about keeping up-to-the-moment on system and supporting software versions. Our upgrades tend to be painful, as witnessed by this last one.

[16:43] Kadah Coba: You need a better deployment system.

[16:43] Simon Linden: If things work (that's a whole different discussion...) then we dont' want to fix it

[16:43] Falcon Linden: We need a better operating system :)

[16:44] Vincent Nacon: they need better everything!.... one step at a time, folks.

[16:44] Falcon Linden: in particular, something that is actually supported by a company rather than a bunch of random bearded folks on IRC :)

[16:44] Kallista Destiny: While I;m sur it was painful for y'all I think that the uses really didnlt notice much impact (except the lll reboot)

[16:44] Simon Linden: Well, deploying to thousands of servers that deal with so much data while tens of thousands of people want to use the system is going to be tough

[16:44] Renae Daines: cleared up the issues I was seeing - though it brought around a new problem ;)

[16:44] DMC Zsigmond: Sorry to come in late in the conversation, but Simon, in short, how is the Linden Server Hardware holding up these days. Is the need to update server spec pressing or not?

[16:45] Kallista Destiny: Better than Linux? AIX perhaps?

[16:45] Tankmaster Finesmith: mac server!

[16:45] Jonathan Yap: Was the last hardware update about 2 years ago?

[16:45] Imaze Rhiano: how many sims are run in single server these days anwyay?

[16:45] Simon Linden: We were talking earlier about us investigating new hardware ... hopefully some of the old ones will be phased out soon

[16:45] Sahkolihaa Contepomi: One sim per CPU core, isn't it?

[16:46] Sahkolihaa Contepomi: (For full ones)

[16:46] Simon Linden: For regular regions, yes

[16:46] DMC Zsigmond: What about Homesteads?

[16:46] Kallista Destiny: except the 8 core machines only have 7 regions, right?

[16:46] Kadah Coba: vSphere 5 has some awesome deployment features

[16:47] Simon Linden: I believe that's correct, Kallista

[16:47] Kallista Destiny: lets you have 1 core for all the OS overhead.

[16:48] Simon Linden: Right, and there are other back-end jobs that need CPU time that kick in when assets are uploaded, capabilies set up, that kind of thing

[16:48] Flip Idlemind: So...I'd love to get this looked at: SVC-7404

[16:48] JIRA-helper:

[#SVC-7404] Display Name Lookups Fail for Banned Accounts

[16:49] Vincent Nacon: who knows but he found a failure

[16:49] Simon Linden: Do you know if those accounts ever had display names set?

[16:50] Simon Linden: According to the jira notes, Maestro has done some testing and didn't see the problem. I wonder if the age of the ban is a factor

[16:50] Flip Idlemind: I don't think they ever did have display name's set

[16:51] Flip Idlemind: Timeline's fuzzy but they could have been banned (not banned, but "put on hold", and no, they're not my alts ;D) before display names actually came out

[16:51] Flip Idlemind: But still, it should at least return their legacy names right?

[16:51] Flip Idlemind: And also, Maestro tested on Aditi, which I mentioned seems to be unaffected

[16:51] Simon Linden: Assuming they are in the database, yes.

[16:51] Tiny Pumpkin: Your Meeroo is digging and seems to have found something!

[16:52] Simon Linden: That makes me think it's a database problem ... the Aditi database tends to be older

[16:52] Flip Idlemind: Sounds reasonable. This all started happening semi-recently

[16:52] Flip Idlemind: Maybe in the last month or so

[16:52] Simon Linden: I wonder if they were purged at some point ... I'm guessing here

[16:53] Jonathan Yap: Do accounts ever totally go away?

[16:53] Simon Linden: I don't think we purge old account records, but the display names system may be a snapshot or a sub-system

[16:53] Vincent Nacon: best to use username or key whenever possible

[16:53] Simon Linden: That system has an amazing about of load on it, and lots of caching at the viewer, server and back-end

[16:54] Flip Idlemind: Yeah I checked a "banned" profile against a non-existent profile. The banned one says "This name is unavailable" while the the non-existent one says "The page you're looking for doesn't exists"

[16:54] Flip Idlemind: So the accounts exist in some way

[16:54] Falcon Linden: Oh, by the way, we've significantly raised the llCastRay limits with the server version that will go grid-wide tuesday

[16:54] Flip Idlemind: Wooty

[16:54] Simon Linden: Right, they just don't seem to come back from that display name query with a different reponse

[16:54] Vincent Nacon: by how much?

[16:55] Meeter: Timecheck : User Group is almost over

[16:55] Simon Linden: You mean the throttles, Falcon?

[16:55] Falcon Linden: enough that maestro wasn't able to cause it to run out of time without using multiple scripts

[16:55] Falcon Linden: yes, Simon

[16:55] Falcon Linden: we increased the totals by 4x, I think, and the replenish rates by 2x

[16:55] Vincent Nacon: but how you know that's a good idea though?

[16:55] Falcon Linden: That said, small parcels on homesteads still may have problems

[16:56] Falcon Linden: vincent: what do you mean?

[16:56] Simon Linden: We never really know, Vincent, lol

[16:56] Vincent Nacon: doesn't tax the server's resource much?

[16:56] Falcon Linden: oh, right, because:

[16:56] Falcon Linden: the only reason llCastRay was given its own resource pool was so that people couldn't abuse it by having multiple scripts in the same object all calling it repeatedly

[16:57] Falcon Linden: llCastRay on its own is not very resource intensive (usually) and script time slicing should keep things well behaved as long as resis don't try to cheat

[16:57] Falcon Linden: but they will try to cheat, so the llCastRay throttle is a backup

[16:58] Vincent Nacon: just have feeling about some weapons using llCastRay for each bullets

[16:58] Vincent Nacon: or even try to make a sneaking target bullets in semi firing

[16:59] Vincent Nacon: tracking... sneaking... whatever they fancy calling it

[17:00] Meeter: Thank you for coming to the Server User Group

[17:00] Vincent Nacon: alrighty, thanks for coming out

[17:00] Simon Linden: Yeah, I think cheating is a whole diffenent topic taht gets into script permissions, and such


Prev 2011.10.25 Next 2011.11.01