Chat log from the meeting on 2014-08-12

From OpenSimulator

Jump to: navigation, search
[11:02]  Sasha Nae: sorry If I'm interupting
[11:02]  Freaky.Tech @freakytech.spdns.de:8002 is Online
[11:02]  Andrew Hellershanks is Online
[11:02]  Nebadon.Izumi @cc.opensimulator.org:8002 is Online
[11:03]  Shez Oyen: not at all
[11:03]  Andrew Hellershanks: I don't think the meeting has started yet, Sasha. All are welcome to attend.
[11:03]  Richardus Raymaker: hi shez
[11:03]  Shez Oyen: Hi Rich :)
[11:03]  Justin Clark-Casey: hi folks
[11:03]  Sasha Nae: what is the meeting about?
[11:03]  Andrew Hellershanks: Hey, justin
[11:03]  Nebadon Izumi: this is the Weekly OpenSimulator Developers office hours meeting
[11:03]  Shez Oyen: Olla Justin
[11:03]  Mata Hari: hi Justin
[11:03]  Richardus Raymaker: hi justin
[11:04]  Sasha Nae: Heya Dan Banner
[11:04]  dan banner: hello
[11:04]  Andrew Hellershanks: I filed two mantis reports this week.
[11:04]  Shez Oyen: Sasha the meeting is about development of Opensim/osgrid
[11:04]  Kayaker Magic: Yay!
[11:04]  Sasha Nae: aaah
[11:04]  Sasha Nae: okay
[11:04]  BlueWall Slade is Online
[11:05]  Andrew Hellershanks: Justin, Can I push the changes to llEmail?
[11:05]  Simulator Version v0.5 shouts: OpenSim 0.8.1 Dev          8738445: 2014-08-09 00:52:51 +0100 (Unix/Mono)
[11:05]  Justin Clark-Casey: andrew: did you make further changes to that patch on mantis@
[11:05]  Justin Clark-Casey: ?
[11:06]  Andrew Hellershanks: Justin, yes. I added an extra check to see if the destination was local or off-world. If local, the address is used as provided. My previous version would have affect object to object email.
[11:06]  Sasha Nae: sory
[11:06]  Shez Oyen: lol
[11:06]  Justin Clark-Casey: andrew: ok cool. I say go for it
[11:06]  Sasha Nae whispers: my avatar is lagging AGAIN!
[11:07]  Andrew Hellershanks: justin, ok. I'll do that later today.
[11:07]  Andrew Hellershanks: Does anyone have any sample code to test object to object email?
[11:07]  Richardus Raymaker: Sasha, check your draw distance first
[11:07]  Richardus Raymaker: http://wiki.secondlife.com/wiki/Llemail
[11:07]  Shez Oyen: Hi Robert
[11:07]  Nebadon Izumi: Andrew i dont think you can do object to object email
[11:07]  Nebadon Izumi: without some kind of pop3 module
[11:07]  Robert Adams: hello all
[11:08]  Justin Clark-Casey: hi robert
[11:08]  Richardus Raymaker: hi robert
[11:08]  Justin Clark-Casey: nebdadon: just fyi, changing maxconn didn't make any difference to my issue so I think we should leave it as it is now on oscc
[11:08]  Andrew Hellershanks: Nebadon, I wanted to test that. I haven't dug deep enough in to the internal mail system about that.
[11:08]  Nebadon Izumi: ok cool
[11:08]  Nebadon Izumi: ya without pop3 im not sure you can do object to object email
[11:09]  BlueWall Slade: object to object email only works within the same simulator
[11:09]  Nebadon Izumi: either pop3 service or some kind of custom module that acts as a delivery service
[11:09]  Robert Adams: what tangled threads we weave
[11:09]  Sasha Nae: OSgrid is a ver awesome place to be...seing that everything is free and all
[11:09]  Andrew Hellershanks: I'll have to look closely at how llGetNextMail (or whatever it is called) gets mail.
[11:09]  Justin Clark-Casey: I did observe a curious effect where outoing udp is very peaky but after 23 bots out of 100 timed out, the reamining 77 bots had vastly lower and more stable outgoing
[11:09]  Justin Clark-Casey: maybe I'#m hitting some capacity limit still on my local system
[11:10]  Nebadon Izumi: hmm could be
[11:11]  Robert Adams: queues on top of caches on top of queues... all of which have their own limits
[11:12]  Justin Clark-Casey: yes, it's a bit thicket of stuff
[11:12]  Sasha Nae: I want to go to lf grid
[11:12]  Justin Clark-Casey: not to mention that c# http server which has had custom things done to it....
[11:12]  Andrew Hellershanks: Glad I don't dig in to all that stuff. :)
[11:12]  Robert Adams: OS's http server has always been a weak spot
[11:12]  Justin Clark-Casey nods
[11:13]  Robert Adams: one reason we built Simian -- to use a real web server
[11:13]  Andrew Hellershanks: Could one of the other HTTP server projects been incorporated?
[11:13]  Justin Clark-Casey: seems there still aren't good c# webserver implementations, at leaston a casual look yesterday
[11:13]  Justin Clark-Casey: robert: yeah, real webservers are definitely the way to go for services
[11:13]  Justin Clark-Casey: but still need to deal with an embedded one for the http capabilities in the simulator
[11:14]  Justin Clark-Casey: andrew: there still don't seem to be any good ones in c#
[11:14]  Justin Clark-Casey: incorporating someting from native code would bring hcallenges of it's own. though we do it for physics after all...
[11:14]  Andrew Hellershanks: oh, right. Need it in C#
[11:14]  Mata Hari: how extensive an upheaval would it be to pull some of the proxying from the region and let viewer do stuff directly with the assets or inventory servers?
[11:14]  Robert Adams: since we need only a subset of web-serverness, building our own is an option
[11:14]  Nebadon Izumi: I thought there was a mono C# web server that supposedly out performed apache
[11:15]  Justin Clark-Casey: mata: asset stuff can already be done. Inventory is a bit more of a challenge I think because of the general need for security
[11:15]  Justin Clark-Casey: but somebody may have done that
[11:15]  Mata Hari nods
[11:15]  Justin Clark-Casey: robert: true - I always try and avoid that approach though - yet another thing to maintain
[11:16]  Mata Hari: it seems to be during that point that the inventory fetch gets locked.....Bluewall and I spent a chunk of Sunday data-gathering
[11:16]  Justin Clark-Casey: it's easy for somebody to make great performance claims for their stuff - I like to see it backed up with numbers :)
[11:16]  Justin Clark-Casey: mata: it depends where the problem is happening, there are unfortunately many points of failure
[11:16]  Robert Adams: very true -- then one discovers that something is realy using multi-part mime or some web essoterica... and then you have to build it :-(
[11:17]  BlueWall Slade: justin, we are fetching the same item over 60x in some cases
[11:17]  Andrew Hellershanks: eek
[11:17]  Justin Clark-Casey: bluewall: how do you know it's the same item?
[11:17]  BlueWall Slade: I have been digging through the packet traces and code
[11:18]  BlueWall Slade: at this point, I'm not sure whether the viewer is giving up and asking for the item again, or our fetch descendents is causing it
[11:18]  Robert Adams: there has been an on-going suspicion that the viewer caching has problems... but it has never been traced down or confirmed
[11:18]  Justin Clark-Casey: I've not observed that kind of thing - just a lot of requests becaus eht viewer appears to fetch practically all the inventory
[11:18]  Mata Hari: you'll see in the parse I posted from my test that the 146 items that hadn't yet been fetched from my inventory were getting called over and over gain dozens of times
[11:18]  BlueWall Slade: we get good returs, and withing a coul\ple of seconds
[11:18]  BlueWall Slade: but, the item seems to never reach the viewer
[11:19]  Justin Clark-Casey: it's not impossible that it's a problem with polling as well, either in the opensim part or the adjustments made to the c# http server
[11:19]  BlueWall Slade: the http looks very clean
[11:19]  Justin Clark-Casey: in what sense?
[11:19]  BlueWall Slade: if we are the problem, I suspect that a thread is crashing or soemthing
[11:20]  BlueWall Slade: well, we make a request and get it back - clean
[11:20]  BlueWall Slade: the packet traces
[11:20]  Justin Clark-Casey: to what, the services?
[11:20]  BlueWall Slade: yes
[11:20]  BlueWall Slade: from the simulator to the inventory service
[11:20]  BlueWall Slade: about ~2 sec round trip
[11:20]  Justin Clark-Casey: yes,I'm begining to think that's not the source of the problem, and it is something to do with actually getting the response back to the client
[11:21]  Mata Hari: my sends were consistently coming back with the corresponding data within 125ms
[11:21]  Justin Clark-Casey: I used to think that it was a spped issue and perhaps the client request was timing out (hence this working just fine on plazas)
[11:21]  BlueWall Slade: it seems like it might be in the fetch descentents module
[11:21]  Mata Hari: but then instead of passing it on to the viewer it was re-requesting the same data again
[11:21]  Justin Clark-Casey: but removing the inventory lock should have doubled the effective speed but I'm not sure that had an effect
[11:21]  Justin Clark-Casey: bluewall: that module hasn't changed for a long time
[11:21]  BlueWall Slade: that seems to take a list of items to fill the skeleton and pop it in a thread
[11:21]  BlueWall Slade: yeah
[11:22]  BlueWall Slade: but, maybe some underlieing dep.
[11:22]  Justin Clark-Casey: the polilng stuff is pretty complex. I may end up lloking yet more closely at it in the course of trying to find out why my local behaves oddly under high bot load
[11:22]  BlueWall Slade: anyways, that's where I am in my search atm.
[11:23]  Justin Clark-Casey: although there seem to be general issues both with http and udp, which is odd
[11:23]  BlueWall Slade: I will look for the point that the data is getting re-requested and try to isolate it a bit
[11:23]  Justin Clark-Casey: yeah,would be good to know the inveral there
[11:24]  Justin Clark-Casey: I didn't think the viewer actually did retry the request, so that is new to me
[11:24]  BlueWall Slade: each call is very clean
[11:24]  Justin Clark-Casey: interval
[11:24]  BlueWall Slade: between the simulator and the service
[11:24]  Justin Clark-Casey: yes, I suspect that part is not the real problem, even if it is quite slow
[11:24]  BlueWall Slade: ~2seconds
[11:24]  BlueWall Slade: that seems to be norm to OSGrid
[11:24]  Justin Clark-Casey: right, but you see how many requests go on
[11:24]  dan banner: actually when the inv fetch does work it seems pretty quick now
[11:25]  Justin Clark-Casey: it's still at least a 250ms round trip for austin, at least
[11:25]  Nebadon Izumi: Inventory almost always fetches fast for me
[11:25]  BlueWall Slade: as long as the user is in the sim, the request queue keeps rising
[11:25]  Nebadon Izumi: very odd
[11:25]  Mata Hari: thousands and thousands of requests for 150 items is the problem
[11:25]  Nebadon Izumi: considering how large my inventory is
[11:25]  Nebadon Izumi: though when i went to BlueWalls region
[11:25]  Justin Clark-Casey: bluewall; it is throttled to only carry out 2 requests at a time
[11:25]  Nebadon Izumi: i maybe only got 5 items downloaded ever
[11:25]  BlueWall Slade: when they leave, it will eventually clear out
[11:25]  Justin Clark-Casey: up until very recently it was only 1
[11:25]  Nebadon Izumi: but if i got to my regions or plazas its really fast for me
[11:26]  BlueWall Slade: but, the portion between the simulator and service seems to be working very well
[11:26]  BlueWall Slade: I thin kwe have an internal gremlin
[11:26]  Justin Clark-Casey: bluewall: but you see it's quite slow????
[11:26]  Mata Hari: between simulator and inventory service it isn't slow at all
[11:26]  BlueWall Slade: hmm, it should work as long as the request doesn't time out?
[11:26]  Justin Clark-Casey: yes it is, look at your logs
[11:27]  Justin Clark-Casey: bluewall: a particular WebFetch request can ask for contents of mlutiple folders
[11:27]  Dahlia Trimble is Online
[11:27]  Justin Clark-Casey: one such request does not translate to one inventory service request
[11:27]  BlueWall Slade: yes, there are plenty of "slow request" messages in the logs
[11:27]  Justin Clark-Casey: they also pop up on the osgrid inventory service logs
[11:27]  BlueWall Slade: hmm, each item is requested, in turn
[11:27]  Nebadon Izumi: that doesnt necesarrily mean slow network though
[11:27]  Nebadon Izumi: that could mean slow threading too
[11:27]  Justin Clark-Casey: for some reason, some of the requests take a considerable time to fulfill
[11:27]  Justin Clark-Casey: bluewall: right
[11:28]  Shez Oyen: Hi Key :)
[11:28]  BlueWall Slade: in the packet dump, it looks about ~2 seconds
[11:28]  Key Gruin: Hi Shez, hey All :)
[11:28]  Justin Clark-Casey: bluwall: so even if individual requests are quite quick, in aggregate it will take a long time to drag down a 10000 item inventory
[11:28]  Justin Clark-Casey: 250ms * 10000 = 2500000 ms
[11:28]  BlueWall Slade: >>> That is for the full request/response including all the sync/ack
[11:28]  Justin Clark-Casey: whcih is 2500 secs, which is about 40 mins
[11:28]  Nebadon Izumi: lol
[11:29]  Justin Clark-Casey: bluewall: what packet dump are you looking at?
[11:29]  BlueWall Slade: between my simulator and OSGrid inventory service
[11:29]  Justin Clark-Casey: where is your simulator?
[11:29]  Mata Hari: he recorded me logging into his sim
[11:29]  Richardus Raymaker: hi key
[11:29]  BlueWall Slade: Mata Hara was Guinna Pig
[11:29]  BlueWall Slade: Hari*
[11:30]  Key Gruin: hi Rich
[11:30]  BlueWall Slade: in my home network
[11:30]  Justin Clark-Casey: geographical location
[11:30]  BlueWall Slade: it is on a pretty slow dsl
[11:30]  Nebadon Izumi: its kind of perplexing that retreiving a list of uuids could have so much impact
[11:30]  BlueWall Slade: South Carolina USA
[11:30]  Justin Clark-Casey: and mata is eu right?
[11:30]  Mata Hari: no, Canada....Ontario
[11:30]  Mata Hari: ping to the inventory server is ~60ms on average for me
[11:31]  Justin Clark-Casey: ok, that's interesting
[11:31]  Justin Clark-Casey: yeah, for austin it's 250ms which is 4x
[11:31]  Nebadon Izumi: ya and also Pingdom who is constantly monitoring our asset server
[11:31]  Nebadon Izumi: for the entire past month
[11:31]  Nebadon Izumi: it had 100.00% uptime
[11:31]  Nebadon Izumi: literally 0 bad requests
[11:31]  Justin Clark-Casey: ok, so maybe it does make this more likely soemthing is going wrong in the simulator
[11:31]  Nebadon Izumi: they fetch an asset from our server every 60 seconds
[11:32]  Dahlia Trimble: hi
[11:32]  Justin Clark-Casey: but I don't know why we don't see more reports of this issue - perhaps it's just not reported
[11:32]  BlueWall Slade: I think there is some speed related trigger - but OSG services are good enough
[11:32]  Justin Clark-Casey: at 250ms I can definitely imagine issues
[11:32]  BlueWall Slade: it just happens in some cases
[11:32]  Shez Oyen: Hi Dahlia :)
[11:32]  Justin Clark-Casey: hi dahlia
[11:32]  Nebadon Izumi: http://stats.pingdom.com/rcoc14tnw05o/341147?BHT-8dae65d1-03c3-4d0c-b86c-2bede2922000.7
[11:32]  Mata Hari: when I was parsing through the log I made with debug at 5 I was able to match all outbound requests with a matching response with times in the range of 125-150ms
[11:32]  Nebadon Izumi: look
[11:32]  Key Gruin: I've experienced the fetch hang from my region on Neb's server
[11:32]  BlueWall Slade: Since I can faithfully reproduce it, I hope to isolate it to a specific area
[11:33]  Justin Clark-Casey: so even if one request gets dropped, if the viewer is retrying I'msurprised it's a consistent failure
[11:33]  Justin Clark-Casey: I did wonderif it could be connected with a corrupted inventory entry or something, but I think that's somewhat unlikely
[11:33]  BlueWall Slade: Also, it seems that if an avatar hangs it, then a new avatar gets hung at -0- items
[11:33]  Robert Adams: I've got to run... take care all. RL calls.
[11:34]  Nebadon Izumi: here is inventory server stats
[11:34]  Justin Clark-Casey: bluewall: that woudl be consistent with inventory processing stopping, which would be another kind of failure
[11:34]  Andrew Hellershanks: Bye, Robert
[11:34]  Nebadon Izumi: http://stats.pingdom.com/rcoc14tnw05o/341173
[11:34]  Justin Clark-Casey: bye robert
[11:34]  BlueWall Slade: nope, it's looking like something in the simulator
[11:34]  dan banner: if that was the case it wouldnt it also fail in the plazas?
[11:34]  Mata Hari: it's like a queue is filled and nothing new can get added
[11:34]  Justin Clark-Casey: bluewall: that's definitely the case? it's an interesting data point
[11:34]  Justin Clark-Casey: bluewall: the other possibility is that both inventory fetchrequests get tiedupwaiting for a response from the simulator
[11:34]  BlueWall Slade: anyways, I will continue looking at it until we can get a better idea
[11:34]  Justin Clark-Casey: but it sounds like that's notwhat is happening
[11:35]  Justin Clark-Casey: yeah, I've looked a lot at this to eliminate various possibilities and get more data but it's still elduing me
[11:35]  BlueWall Slade: the part between the sim and service are working ok
[11:35]  BlueWall Slade: something keeps making requests and it does make them
[11:36]  BlueWall Slade: but, the response is not getting back to the viewr < or the next stage of the reply in the sim
[11:36]  Justin Clark-Casey: ok
[11:36]  Justin Clark-Casey: good to know
[11:36]  BlueWall Slade: I want to add more logging <here> and possibly run in a debugger
[11:37]  BlueWall Slade hopes it's not in all the threading :S
[11:37]  Mata Hari: well, you know where to find your guinea pig when you're ready to test it
[11:37]  BlueWall Slade: pk, thanks :)
[11:38]  Justin Clark-Casey: It might ultimately be necessary to write a server side inventory data cache
[11:38]  Justin Clark-Casey: which I hope is possible along the same lines that the viewer is supposed to cache inventory data
[11:38]  Justin Clark-Casey: but that's not a 5 min job
[11:38]  BlueWall Slade: how much would that add to our memory footpring in a busy region?
[11:39]  BlueWall Slade: footprint*
[11:39]  Justin Clark-Casey: not a lot if it's just a disk based cache
[11:39]  Justin Clark-Casey: which does make it more complicated to write
[11:39]  BlueWall Slade: any security issues?
[11:39]  Justin Clark-Casey: seeing as simulators can access this data at any time anyway, I don't think so
[11:40]  BlueWall Slade: it passses through now, but if we wright it, it's there for all
[11:40]  Andrew Hellershanks: Could you leverage Flotsam cache code?
[11:40]  Justin Clark-Casey: all who?
[11:40]  BlueWall Slade: that is what strikes my mind
[11:40]  Justin Clark-Casey: andrew: I'm not fond of that code, it has unresolved issues
[11:40]  Andrew Hellershanks: ok
[11:40]  BlueWall Slade: haa, ok
[11:40]  BlueWall Slade: oh, all == non-coders
[11:41]  Mata Hari: sim operator sees it but could just as easily see it by turning on high level logging and then parsing it out
[11:41]  Justin Clark-Casey: yes. It would just be security by 'obscurity' and not even any great obscurity
[11:41]  BlueWall Slade: :D
[11:41]  BlueWall Slade: ok, I'll keep posting about what I find
[11:42]  BlueWall Slade: we may just need to dot and i or cross a t somewhere :/
[11:42]  Justin Clark-Casey: I'll talk to you too about what I look at/find when I get a chance to look at it further
[11:42]  BlueWall Slade: ok, great, thanks
[11:42]  Justin Clark-Casey: back in a min
[11:43]  Andrew Hellershanks: I have a couple of issues/questions this week.
[11:43]  Ai Austin is Online
[11:43]  Nebadon Izumi: k whats up?
[11:43]  BlueWall Slade: no more iussues, thtat queue is closed until further notice
[11:43]  BlueWall Slade: jeje
[11:43]  Mata Hari: lol
[11:44]  Andrew Hellershanks: Do region bans work? I looked at the code to check for references to the regionban table and only saw a reference in code relating to SQLite.
[11:44]  Nebadon Izumi: no
[11:44]  Nebadon Izumi: only estate
[11:44]  Nebadon Izumi: the parcel bans do not work
[11:44]  Nebadon Izumi: pretty certain they do not anyway
[11:44]  BlueWall Slade: hmm
[11:45]  Andrew Hellershanks: ok. I found it interesting there is a reference to the regionban table but only for one DB type.
[11:45]  BlueWall Slade: I think group access control works
[11:45]  Justin Clark-Casey: back
[11:45]  BlueWall Slade: would htat be similar?
[11:45]  Nebadon Izumi: ya group access whitelist might work
[11:45]  BlueWall Slade: wn jcc
[11:45]  Nebadon Izumi: that bans everyone
[11:45]  Nebadon Izumi: individual parcel bans are not effective at this time
[11:45]  Nebadon Izumi: no one ever wired that up
[11:45]  Andrew Hellershanks: ok, I can put that on my other list of things to look in to at some point.
[11:46]  Andrew Hellershanks: I noticed that when I was looking in to some stuff relating to the other mantis I filled.
[11:46]  Andrew Hellershanks: Has anyone else looked at mantis 7300 -> http://opensimulator.org/mantis/view.php?id=7300
[11:46]  BlueWall Slade: the email one?
[11:46]  Andrew Hellershanks: I'm wondering if anyone else has ever noticed that bug.
[11:46]  Justin Clark-Casey: I almost wonder if viewers are trying to fetch an inventory item that doesnt' exist and we don't know how to tell them that
[11:46]  Nebadon Izumi: i have never experiened that
[11:46]  Andrew Hellershanks: Blue, no. Duplicate entries in estate tables.
[11:47]  Justin Clark-Casey: but then it works on plazas....
[11:47]  Nebadon Izumi: oh I have seen that
[11:47]  Nebadon Izumi: duplicate estates anyway, but long time ago, but i dont ever recall seeing a duplicate entry for an estate manager
[11:47]  BlueWall Slade: never have seen it, but that is a lot of duplicates
[11:48]  Andrew Hellershanks: It happens once in a while. I get reports of problems in a region and that is what it turns out to be. I had to removed over 11,500 duplicates in one estate table the other day.
[11:48]  Richardus Raymaker: are the dub estate managers not because more sims ?
[11:48]  Andrew Hellershanks: I get duplicates in two different estate tables.
[11:49]  Andrew Hellershanks: Richardus, There are more duplicates than there are regions.
[11:50]  [2014/08/12 11:43]  Mata Hari: lol
[11:50]  [2014/08/12 11:44]  Andrew Hellershanks: Do region bans work? I looked at the code to check for references to the regionban table and only saw a reference in code relating to SQLite.
[11:50]  [2014/08/12 11:44]  Nebadon Izumi: no
[11:50]  [2014/08/12 11:44]  Nebadon Izumi: only estate
[11:50]  [2014/08/12 11:44]  Nebadon Izumi: the parcel bans do not work
[11:50]  [2014/08/12 11:44]  Nebadon Izumi: pretty certain they do not anyway
[11:50]  [2014/08/12 11:44]  BlueWall Slade: hmm
[11:50]  [2014/08/12 11:45]  Andrew Hellershanks: ok. I found it interesting there is a reference to the regionban table but only for one DB type.
[11:50]  [2014/08/12 11:45]  BlueWall Slade: I think group access control works
[11:50]  [2014/08/12 11:45]  Justin Clark-Casey: back
[11:50]  [2014/08/12 11:45]  BlueWall Slade: would htat be similar?
[11:50]  [2014/08/12 11:45]  Nebadon Izumi: ya group access whitelist might work
[11:50]  [2014/08/12 11:45]  BlueWall Slade: wn jcc
[11:50]  [2014/08/12 11:45]  Nebadon Izumi: that bans everyone
[11:50]  [2014/08/12 11:45]  Nebadon Izumi: individual parcel bans are not effective at this time
[11:50]  [2014/08/12 11:45]  Nebadon Izumi: no one ever wired that up
[11:50]  [2014/08/12 11:45]  Andrew Hellershanks: ok, I can put that on my other list of things to look in to at some point.
[11:50]  [2014/08/12 11:46]  Andrew Hellershanks: I noticed that when I was looking in to some stuff relating to the other mantis I filled.
[11:50]  [2014/08/12 11:46]  Andrew Hellershanks: Has anyone else looked at mantis 7300 -> http://opensimulator.org/mantis/view.php?id=7300
[11:50]  [2014/08/12 11:46]  BlueWall Slade: the email one?
[11:50]  [2014/08/12 11:46]  Andrew Hellershanks: I'm wondering if anyone else has ever noticed that bug.
[11:50]  [2014/08/12 11:46]  Justin Clark-Casey: I almost wonder if viewers are trying to fetch an inventory item that doesnt' exist and we don't know how to tell them that
[11:50]  [2014/08/12 11:46]  Nebadon Izumi: i have never experiened that
[11:50]  [2014/08/12 11:46]  Andrew Hellershanks: Blue, no. Duplicate entries in estate tables.
[11:50]  [2014/08/12 11:47]  Justin Clark-Casey: but then it works on plazas....
[11:50]  [2014/08/12 11:47]  Nebadon Izumi: oh I have seen that
[11:50]  [2014/08/12 11:47]  Nebadon Izumi: duplicate estates anyway, but long time ago, but i dont ever recall seeing a duplicate entry for an estate manager
[11:50]  [2014/08/12 11:47]  BlueWall Slade: never have seen it, but that is a lot of duplicates
[11:50]  [2014/08/12 11:48]  Andrew Hellershanks: It happens once in a while. I get reports of problems in a region and that is what it turns out to be. I had to removed over 11,500 duplicates in one estate table the other day.
[11:50]  [2014/08/12 11:48]  Richardus Raymaker: are the dub estate managers not because more sims ?
[11:50]  [2014/08/12 11:48]  Andrew Hellershanks: I get duplicates in two different estate tables.
[11:50]  [2014/08/12 11:49]  Andrew Hellershanks: Richardus, There are more duplicates than there are regions.
[11:51]  Justin Clark-Casey: I accidentally logged in the same av from another viewer
[11:51]  Nebadon Izumi: oh
[11:51]  Nebadon Izumi: i hate that
[11:51]  Mata Hari: lol
[11:51]  Richardus Raymaker: hehe common error
[11:51]  Andrew Hellershanks: yea, I've had that happen
[11:51]  Nebadon Izumi: i got all my 80 bots logged in already justin
[11:51]  Richardus Raymaker: calling that 'kicking myself out'
[11:51]  Andrew Hellershanks: always annoying
[11:52]  Justin Clark-Casey: amusingly the attachments disappear on the origianl
[11:52]  Justin Clark-Casey: my hair suddenly went missing and I thought, huh?
[11:52]  Nebadon Izumi: probably moves them to the newest connection
[11:52]  Richardus Raymaker: i heared that attachments on HG still disappeared, at least on the dev region i where running it failed for that person. hmm. need to test it again
[11:52]  Andrew Hellershanks: I tried to add an index using the two fields in the two tables but it complained of a duplicate index even though the only existing index was on a single field.
[11:53]  Nebadon Izumi: its been like 99% better at the weekly load tests Richardus
[11:53]  Nebadon Izumi: it was very very bad before
[11:53]  Richardus Raymaker: yes neb, i stuill run version short after it where fixt. so possible mine still buggy
[11:54]  Mata Hari: was fun today after a crash to log back in, go back to my region and still see it sending out friends offline status notices for 2 minutes :p
[11:54]  Richardus Raymaker: you hav eto many friends :P
[11:54]  Mata Hari: yeah...oughta purge them all
[11:54]  Nebadon Izumi: i did an experiment a long time ago
[11:54]  Mata Hari: unless they send cash
[11:54]  Nebadon Izumi: i used to have 100s of people on my friends list
[11:54]  Nebadon Izumi: i purged like 75% of them once
[11:54]  Nebadon Izumi: and my log ins were soo much faster after that
[11:55]  Nebadon Izumi: that was few years back though
[11:55]  Nebadon Izumi: i doubt it would be the same impact today
[11:55]  Andrew Hellershanks: As part of fixing 7300 I'm going to have another look at my mantis 4380. I've been running mysqltuner and finding it complains about joins being done with no indexes.
[11:55]  Justin Clark-Casey: I think we do practically no joins though
[11:56]  Justin Clark-Casey: except in groups
[11:56]  Andrew Hellershanks: I haven't seached the code for them. Yes, groups is one place where some big joins take place.
[11:56]  Justin Clark-Casey: andrew: are yo uusing xmlrpc groups still?
Personal tools
General
About This Wiki