[Sugar-devel] Jabber Testing Monday Aug 24

[Martin Langhoff]

Hi Caroline,

sorry about the delay -- catching up on XS things now -- enormous
thanks for the report.

So controlling for Gadget and reg'd numbers, it seems that the
ejabberd version (or perhaps other XS compoment) version makes a
difference.

The ejabberd upgrade was very significant, and the rework of the
'online' patch was also fairly important (I assume you are using the
'online' SRG as it's the 'default' config).

In terms of Sugar "preferring" salut, and having to restart...

 - Maybe there are connectivity issues (high latency?) to the XS?

 - If connectivity to the XS is good, it should be enough to wait, as
I've seen Sugar (0.82 at least) try Gabble every once in a while...

 - If waiting is not enough, it should be enough to reconnect to the
AP. That definitely triggers a Gabble connection attempt

IOWs if you do have to restart, there maybe a regression there...

Overall, I find the relatively fast Telepathy switching a bit
frustrating as a programmer, but I've seen how it's useful in real
world usage.

All I would ask for -- should a genius-in-a-lamp ask for my wishes --
would be some subtle UI hint. Maybe a variation on the quater-arcs
that mark the connected AP or something like it...

cheers,



m

On Tue, Aug 25, 2009 at 5:18 AM, Caroline
Meeks<caroline at solutiongrove.com> wrote:
> Again we are working with


> XS1 - responding to jabber.sugarlabs.org. Installed a few months ago and
> currently with >1000 registered users, it tends to have about 20 people on
> it these days.
> XS2 - Installed a few weeks ago, responding to
> schoolserver.xsdev.solutiongrove.com with about 100 registered users
> First i connected to XS1 and joined a Turtle Art instance.  CPU usage rose
> to 100%.  When we stopped sharing the TA it dropped back down.
> We then shared a chat and CPU again rose to 100%.
> Next we switched to testing on XS2.  Sharing Turtle Art and Chat did not
> spike the CPU.
> We felt there were 4 differences between XS1 and XS2
>
> The number of registered users
> XS 1 is an older version of the XS 0.5.2. XS2 is 0.6beta
> XS 2 has Gadget
> XS1 had about 20 users logged in XS2 had 4 or 5
>
> Next we turned off Gadget on XS2 and repeated our test. No CPU spike.  Thus
> we think #3 is unlikely to be the important difference.
> Next we deleted the registered users on XS1 and switched our testing back to
> XS1.  When we shared TA and chat there was still spikes up to 100% CPU.
>  Thus we think #1 is unlikely to be the important difference.
> We will work on this again tomorrow so let us know your ideas
> A separate issue is that it hard to get the computers to connect up the the
> XS. They tended to stay on Salut, sometimes they would goto Jabber on
> resstart, other times not.  We could really use better control and the
> ability to force one or the other.
> Here is the log of our conversation.
>
> Josely is Caroline at the GPA School
> 12:06 < joasely-2260> the jabber server is back up to 100% CPU usage.
> 12:06 < joasely-2260> 22 people are online. over 1000 are registered
> 12:06 < joasely-2260> one turtle Art is being shared.
> 12:06 < joasely-2260> the load did not seem to change when the GPA computers
> I am controlling stopped sharing turtle art
> 12:07 < joasely-2260> current hypothisis - its th 1000 registered users that
> is making the difference between decent performance and 100% CPU
> 12:08 < daveb> jabber is back to 0%cpu
> 12:09 < joasely-2260> Interesting note. The Red Turtle Art activity seems to
> have been started by David who is not online anymore.
> 12:09 < joasely-2260> ah very very interesting.
> 12:09 < joasely-2260> let me join the turtle art again.
> 12:09 < daveb> joasely-2260: ok
> 12:09 < daveb> we have cpu logs every minute back until august 15
> 2:09 < joasely-2260> ok I have joined the Turtle Art
>
> 12:10 < joasely-2260> lets give it 5 minutes with just one person sharing
> it. then I'll add another peron
>
> 12:10 < daveb> up to 100%
> 12:10 < joasely-2260> so note, we can scratch they hypothisis that its the
> 1000+ registered users.
> 12:10 < daveb> yes
> 12:11 < joasely-2260> ok very cool.
>
> 12:11 < joasely-2260> so sharing a turtle art activity causes 100% CPU
>
> 12:11 < joasely-2260> I'm gong to quit TA
> 12:11 < tomeu> erikos: yes
> 12:11 < satellit_> daveb: that TA has been up for several days
>
> 12:11 < joasely-2260> let me know when its back down and I'll try a chat.
> 12:12 < joasely-2260> thanks satelit_
>
> 12:12 < satellit_> I think it is an orphan
>
> 12:13 < daveb> satellit_: yes but when someone joins te cpu spikes
> 12:13 < daveb> joasely-2260: should we quit that and try a different
> activity?
> 12:13 < daveb> unfortunatly i tink debug is off on jabber
>
> 12:13 < daveb> ok cpu back to 0%
>
> 12:14 < satellit_> daveb: is there a way to kill orphan shared items?
> 12:14 < daveb> satellit_: i have no idea
>
> 12:14 < joasely-2260> started chat on one computer
>
> 12:14 < daveb> cpu 100%
>
> 12:15 < joasely-2260> wow before I even joined it anywhere else
> 12:15 < daveb> yes
> 12:15 < joasely-2260> still at 100%?
> 12:15 < daveb> yes
> 12:15 < daveb> now
> 12:15 < daveb> did we ever get this on the 0.6 server?
> 12:15 < joasely-2260> still at 100%?
> 12:15 < daveb> 60% now
> 12:15 < daveb> down to 0
> 1
> 12:15 < joasely-2260> i'm not sure I thought both XS1 and XS2 was 0.6
>
> 12:15 < joasely-2260> ok now going to join with another person
> 12:16 < daveb> joasely-2260: no XS1 is 0.5.2
> 12:16 < joasely-2260> ok second person joined
> 12:16 < tomeu> erikos: fortunately is not big nor invasive, so won't be such
> a big deal
> 12:16 < daveb> cpu between 0 and 3 %
> 12:17 < joasely-2260> hmm ok I'll chat with myself
>
> 12:17 < joasely-2260> anything?
> 12:17 < daveb> cpu up to 100
>
> 12:17 < joasely-2260> ok so doing anythign with collaboration pops the CPU
> to 100%
> 12:18 -!- iwikiwi-1f30 (n=urk at 61.17.45.12) has quit: Read error: 60
> (Operation timed out)
>
> 12:18 < joasely-2260> should I try setting these to XS2?
> 12:18 < joasely-2260> and doing the same test?
> 12:18 < daveb> sure
>
> 12:18 < daveb> let me login over there
> .
> 12:19 < daveb> joasely-2260: ok i am ready
> .
> 12:20 < joasely-2260> schoolserver.xsdev.solutiongrove.com right?
> 12:20 < silbe> tomeu: not logging.exception("message") ?
> 12:20 < joasely-2260> I'm back to salut :(
> 12:20 < phitoo> sdziallas: Remind me where to get the rpm. That's not part
> of the distribution, right.
> 12:20 < satellit_> no it hangs on boot. do I have to do an interactive
> command? on boot
> 12:20 < daveb> joasely-2260: correct
>
> 12:21 < joasely-2260> hmm maybe I should reboot
>
> 12:22 < daveb> joasely-2260: it should work without restarting but sometimes
> that helps to get telepaty working again
> 1
> 12:24 -!- CarolineGPA-2260 (n=urk at 216.163.216.2) has joined #sugar
> 12:24 < CarolineGPA-2260> nope still salut
> 12:24 < daveb> oh i bet its misconfigured
> 12:24 < daveb> when you change te domain name it messed up ejabberd
> 12:25 < CarolineGPA-2260> how?
> 12:25 < tomeu> erikos: it should conflict
>
> 12:25 < daveb> CarolineGPA-2260: ejabberd is bound to the host name
> 12:25 < daveb> kind of annoying
> 12:25 < CarolineGPA-2260> ah maybe if I delete the keys it'll work again?
> 1
> 12:25 < CarolineGPA-2260> so right now you can't just change your host name
> and connect to a new jabber server?
>
> 12:26 < CarolineGPA-2260> should I try deleting the owner keys and restart?
> 12:27 < daveb> no
> 12:27 < daveb> one sec
> 12:27 < daveb> its the server side
>
> 12:32 < daveb> somethinbg is screy
> 12:32 < daveb> the ejabberd scripts are WAY too picky you need to restart
> the whole server
> 12:32 < daveb> i thought I fixed that but must have missed something.
> 12:32 < CarolineGPA-2260> ok
> 12:32 < CarolineGPA-2260> how long do you think the fix will take?
> 12:33 < daveb> a minute or two
> 12:33 < CarolineGPA-2260> oh ok :)
>
> 12:37 < daveb> CarolineGPA-2260: well the server did not restart and I can't
> login to the host machine either
> 12:38 < daveb> jabber.sl.o is back to 100%
> 12:38 < daveb> ls
> 12:41 -!- teefal (n=teefal at 24.229.118.195) has quit: "The computer fell
> asleep"
> 12:41 < daveb> CarolineGPA-2260: it looks like jabber is stll broken on XS2
> 12:41 < daveb> it does not wish to behave
> 12:42 < CIA-73> sugar: sayamindu * rddfc2879a308 / (bin/sugar-session
> extensions/cpsection/keyboard/view.py): Using logging.exception
> 12:44 < daveb> CarolineGPA-2260: nevermind i think its working now :)
>
> 12:46 < CarolineGPA-2260> back I had a call
> 12:46 < CarolineGPA-2260> should I give up for today?
> 12:46 < daveb> CarolineGPA-2260: it looks like te server is up
> 12:46 < daveb> 1 user online
> 12:46 -!- tamiel (n=tamiel at 213.30.183.226) has quit: "leaving"
> 12:46 < CarolineGPA-2260> are you able to connect?
> 12:47 < daveb> i did not try yet
> 12:47 < CarolineGPA-2260> ok I'm restarting gpa2
> 12:48 -!- kristianpaul (n=kristian at 190.165.132.161) has joined #sugar
>
> 12:49 < daveb> i am on xs2
> 12:49 < daveb> i see caroline1
> 12:50 < CarolineGPA-2260> great i see you and caroline2
> 12:50 < CarolineGPA-2260> what is the CPU?
> 12:50 < daveb> 0 now
> 12:50 < daveb> i don't see you anymore
> 12:50 < CarolineGPA-2260> ok starting a chat
> 1
> 12:50 < CarolineGPA-2260> chat is starting, one user
>
> 12:50 < daveb> CarolineGPA-2260: nevermind i had typed in seach
> 12:51 < daveb> CarolineGPA-2260: cpu 0
> 12:51 < daveb> CarolineGPA-2260: 1%
> 12:51 < CarolineGPA-2260> add a person
>
> 12:52 < daveb> it goes to 1% for a second then back to 0
> 12:52 < daveb> CarolineGPA-2260: want to start turtle art?
> 12:52 < daveb> CarolineGPA-2260: also not gadget is configured on here
>
> 12:57 < CarolineGPA-2260> dabeb hows the cpu doing?
>
> 12:57 < daveb> CarolineGPA-2260: 0%
> 12:57 < daveb> i shared tutleart
>
> 12:58 < CarolineGPA-2260> joining TA
> 12
> 12:58 < daveb> CarolineGPA-2260: ok. cpu went to 1% for a second
>
> 12:59 < CarolineGPA-2260> joining TA with 2
>
> 12:59 < daveb> still at 0
> 12:59 < daveb> goes up when you actually do something then back to 0
> 12:59 < daveb> of cours eonly 3 users online
>
> 12:59 < daveb> need 20 or so to compare better
> 12:59 < tomeu> silbe: ok, I guess you can recover it from git if you want
> 12:59 < silbe> tomeu: but good to know it's buried in git somewhere and i
> just need to revive it :)
> 13:00 < silbe> tomeu: similar to the data store test suite :)
> 1
> 13:00 < CarolineGPA-2260> ok so we have 3 differences. 1- 3 users vs 22
> users.  2 - less then 100 registered users vs 1000+ registered users and 3.
> XS .52 vs .60
> 13:00 < CarolineGPA-2260> correct?
>
> 13:00 < daveb> CarolineGPA-2260: gadget is on XS2
>
> 13:01 < CarolineGPA-2260> ok does that make difference #4?
>
> 13:01 < CarolineGPA-2260> does it make sense to turn off gadget and see if
> we see a differnece?
> 13:02 < daveb> CarolineGPA-2260: sure let me see if I can actually do that
>
> 13:02 < CarolineGPA-2260> ok I'm gong to run downstairs, they just annouced
> free food ;)
>
> 13:03 < CarolineGPA-2260> I need to chat up the teachers too so I maybe a
> while if I'm lucky
>
> 13:03 < daveb> ok
> 13:03 < daveb> CarolineGPA-2260: ok i restarted without gadget
>
> 13:09 < CarolineGPA-2260> looks like we have 5 connected. me 3 times you
> twice
> 13:09 < CarolineGPA-2260> hows the cpu daveb?
>
> 13:18 < CarolineGPA-2260> I'm adding some more computers here. they were all
> unplugged
> 13:18 < daveb> ok
> 13:18 < daveb> i see tem
>
> 13:35 < CarolineGPA-2260> daveb, how is the CPU?
>
> 13:35 < daveb> CarolineGPA-2260: <1%
> 13:36 < daveb> CarolineGPA-2260: i shared a new turtle art with 2 people
> 13:36 < CarolineGPA-2260> we are now running without gadget?
>
> 13:36 < CarolineGPA-2260> i'll join in
> 13:36 < daveb> CarolineGPA-2260: without gadget, correct
>
>
> 13:38 < CarolineGPA-2260> GPA1 fell back to salut
>
> 13:38 < CarolineGPA-2260> GPA2 and 3 are in the TA I think.
>
> 13:38 < CarolineGPA-2260> pretty messed up sharing though
>
> 13:40 < CarolineGPA-2260> hows the cpu?
>
> 13:41 < daveb> CarolineGPA-2260: 0 still
>
> 13:42 < daveb> CarolineGPA-2260: its not clear exactly what is different
> 1
> 13:42 < CarolineGPA-2260> daveb sent you a video of how it looks on the
> computers here.
> 13:43 < CarolineGPA-2260> lets do a chat too
> 13:43 < daveb> ok
> 13:43 < daveb> CarolineGPA-2260: shared chat
>
> 13:44 < CarolineGPA-2260> joined
> 13:44 < CarolineGPA-2260> hows it looking
> 13:45 < daveb> cpu went to 9.6% then right back to 0
> 13:45 < CarolineGPA-2260> GPA2 has now lost connection to the jabber server
> 13:45 < CarolineGPA-2260> ok so the experiement was, was Gadget what was
> making the difference
> 13:45 < daveb> CarolineGPA-2260: weird. still 0%
> 13:46 < daveb> CarolineGPA-2260: right it did not appeat to affect with a
> small number of online users
> 13:46 < CarolineGPA-2260> and the result was no. It works fine with Gadget
> off or on.
> 13:46 < CarolineGPA-2260> under these conditions
>
> 13:46 < CarolineGPA-2260> so we have 3 other difference to look at 1.
> registered users, 2. .52 vs 6 3. number of total users logged on
> 13:46 < CarolineGPA-2260> does that sound right?
> 13:46 < daveb> CarolineGPA-2260: correct.
> 13:46 < daveb> so there could be some weird interaction wit
> 13:47 < daveb> collab wit a large number of registered users
> 13:47 < daveb> like it has to loop through them or someting weird.
>
> 13:47 < CarolineGPA-2260> ok so one option is   to switch the DNS so
> jabber.sl.org points to XS2
> 13:47 < CarolineGPA-2260> the other option is to delete the registered users
> on XS1
>
> 13:49 < CarolineGPA-2260> so the log has it and incase anyone wants to ee
> here is the video of how TA sharing looked:
> http://www.youtube.com/watch?v=APt0J4ln28w
> 13:50 < daveb> i tink that is a TA artifact. I saw that here also. One had a
> square and te other te turtle had not moved.
> 13:50 < CarolineGPA-2260> yeah I think that is a separate issue, which I
> will report later.
>
> 13:50 < daveb> CarolineGPA-2260: ok right, we need to see if alot of
> registered users on XS2 works with collab.
> 1
> 13:50 < daveb> i can do hyperactivity and add alot fo random users quickly
>
> 13:51 < CarolineGPA-2260> seems like the one thing we can do right now is
> delete the users on XS1 and see if it works any better
> 13:51 < daveb> ok
> 1
> 13:51 < daveb> makes sense.
> 13:51 < CarolineGPA-2260> ok shall I swich the gPA back to
> jabber.sugarlabs.org?
> 13:52 < daveb> wait until i restart it
> 13:52 < CarolineGPA-2260> ok
>
> 13:54 < daveb> CarolineGPA-2260: ok to switch
> 13:58 < CarolineGPA-2260> ok all switched
> 13:58 < CarolineGPA-2260> but only GPA 1 seems to be on jabber and I don't
> see anyone else
> 13:58 < CarolineGPA-2260> are you connected?"
>
> 14:00 < daveb> CarolineGPA-2260: me too!!
> 14:00 < daveb> it says
> 14:00 < daveb> 23 people online
> 14:00 *** erikos is out for today
> 14:00 < daveb> and new registered users
> 14:00 < daveb> wait
> 14:00 < daveb> duh
> 14:00 < daveb> one sec
> 14:00 < daveb> i broke it.
> 14:00 -!- erikos (n=erikos at g225093143.adsl.alicedsl.de) has quit: "Leaving"
> 14:00 < daveb> again.
> 14:00 < daveb> :)
> 14:00 < daveb> CarolineGPA-2260: need to redo the shared roster
> 14:01 < CarolineGPA-2260> ok
> 14:01 < daveb> CarolineGPA-2260: should fix itself in a minute i hope
> 14:02 -!- blendmaster1024 (n=blendmas at 75-169-206-152.slkc.qwest.net) has
> joined #sugar
> 14:05 < dfarning> sdziallas: good morning!  sorry for the late response I am
> still a irc/fedora/pidgin newbie.  I need to figure out how to get the
> screen blink.
> 14:06 < daveb> CarolineGPA-2260: restarting ejabberd
> 14:06 < daveb> wait!
> 14:06 < daveb> its working finally.
> 14:06 < daveb> wait restart will be better
> 14:06 < CarolineGPA-2260> I see Jack
> 14:06 < sdziallas> dfarning: hey :) (heh. not sure about pidgin, I'm using
> xchat-gnome - which blinks)
> 14:06 < CarolineGPA-2260> are you restarting again?
> 14:06 < daveb> yeah you can only see people
> 14:07 < daveb> who were online AFTER i create the group
> 14:07 < daveb> restarted
> 14:07 < daveb> should pop up now
> 14:07 < daveb> there they go
> 14:07 < CarolineGPA-2260> yup hows the cpu?
> 14:08 < daveb> 129%!
> 14:08 < CarolineGPA-2260> nope, now all my computers are local
> 14:08 < daveb> i see alot of people now
> 14:08 < daveb> yours will have to reset telepathy
> 14:08 < CarolineGPA-2260> ok so I think we can eliminate the number of
> registered users as the issue
> 14:08 < daveb> takes a few mins
> 14:08 < daveb> wait its down
> 14:08 < daveb> to 6
> 14:08 < CarolineGPA-2260> I'll restart everything but 1
> 14:09 < daveb> so it looks like its a high load to update 20 people's
> neighborhood though
> 14:09 < daveb> i will try turtle art
> 14:10 < CarolineGPA-2260> ok but if its transient thats not really that bad.
> 14:10 < daveb> CarolineGPA-2260: correct.
> 14:10 < CarolineGPA-2260> lets see if we can peg it again sharing TA and
> chat with a few people.
> 14:10 < daveb> CarolineGPA-2260: that is only on restart when everyone
> updates at the same tme. it'll be the same
> 14:10 < daveb> when alot of people
> 14:10 < CarolineGPA-2260> right
> 14:10 < daveb> in the same calss for example
> 14:11 < daveb> CarolineGPA-2260: two people in turtle art
> 14:11 < daveb> 0%
> 14:12 < CarolineGPA-2260> gpa 2 joining
> 14:12 < daveb> i stopped it and started a new one
> 14:12 < CarolineGPA-2260> lets strat with chat
> 14:13 < CarolineGPA-2260> ok I'm with you n0w
> 14:13 < daveb> ok then
> 14:13 < CarolineGPA-2260> GPA2 joined
> 14:13 < CarolineGPA-2260> anything on the CPU?
> 14:13 < daveb> nope
> 14:13 < CarolineGPA-2260> ok I'll join some others
> 14:14 -!- blendmaster1024 (n=blendmas at 75-169-206-152.slkc.qwest.net) has
> quit: "gonna JB my iphone now"
> 14:14 < CarolineGPA-2260> hmm only GPA2 joined jabber after restart :(
> 14:14 < CarolineGPA-2260> i wonder what is going on with that.
> 14:14 < daveb> yeah it might take a few more mins to get telepathy to reset
> 14:14 < CarolineGPA-2260> ack I have to leave the buidling they are locking
> it
> 14:15 < CarolineGPA-2260> more tomorrow!
> 14:15 < CarolineGPA-2260> bye
>
>
> _______________________________________________
> Sugar-devel mailing list
> Sugar-devel at lists.sugarlabs.org
> http://lists.sugarlabs.org/listinfo/sugar-devel
>
>



-- 
 martin.langhoff at gmail.com
 martin at laptop.org -- School Server Architect
 - ask interesting questions
 - don't get distracted with shiny stuff  - working code first
 - http://wiki.laptop.org/go/User:Martinlanghoff