Web   ·   Wiki   ·   Activities   ·   Blog   ·   Lists   ·   Chat   ·   Meeting   ·   Bugs   ·   Git   ·   Translate   ·   Archive   ·   People   ·   Donate

#treehouse, 2010-05-04

Index | Today     Channels | Search | Join

All times shown according to UTC.

Time Nick Message
16:07 dogi #info Weekly Infrastructure meeting:
16:07 #info Volunteer Infrastructure Gang (http://olpcorps.org/ ),
16:07 #info Sugarlabs Infrastructure Team (http://sugarlabs.org/ ),
16:07 #info and TreeHousers (http://me.etin.gs/treehouse/ )
16:07 #info Date: 2010-05-04
16:07 #info Time: 20:00 UTC (16:00 EST, 22:00 CET)
16:07 #info Agenda: http://openetherpad.org/8PZrY7QXeR
16:07 #info Location: #treehouse on irc.oftc.net
16:07 #link http://embed.mibbit.com/?serve[…]nnel=%23treehouse
16:07 #topic last meeting
16:07 #info Usefull Links:
16:07 #info LastLog:
16:07 #link http://me.etin.gs/treehouse/tr[…]0100427_1601.html
16:07 #info LastMinutes:
16:07 #link http://me.etin.gs/treehouse/tr[…]0100427_1601.html
16:07 #info NextAgenda:
16:07 #link http://openetherpad.org/VEBg0R9bXZ
16:07 #info actioncheck
16:07 #done dogi added 2gb ram to rtVM
16:07 #done dogi/silbe set up openehterpadVM @ housetree
16:07 #done dogi/adam move aslocluster to RIT
16:07 #done bernie fixed munin problems
16:07 are there other done action item to report?
16:08 ping hhardy silbe _bernie lfaraone ?
16:08 silbe is there
16:08 dogi #topic agenda
16:08 #info * aslocluster @ RIT update (http://www.rit.edu/academicaff[…]tudentinnovation/ )
16:08 #info * fix lightwave backup
16:08 #info * fix karmic template
16:08 #info * create lucid template for VM
16:08 #info * access for olpc-nz
16:09 lfaraone dogi: semipong.
16:09 dogi #info * ubuntu-lucid-buildbot
16:09 #info * rtVM update
16:09 #info * homework for all treehousers
16:09 #info * access fairy
16:09 hi lfaraone
16:09 is there a topic to add?
16:11 #topic aslocluster @ RIT update
16:11 #link http://www.rit.edu/academicaff[…]tudentinnovation/
16:11 lfaraone dogi: no.
16:12 dogi friday and saturday walter adam and dogi was visiting RIT
16:12 by that we moved finally the servers ...
16:13 unfortunatly the server still are not racked appropriately
16:14 but I think RIT will soon find a place for them
16:15 right now jlew cdeslandes decause have root access to the servers
16:15 and will add wwdillingham as soon as they will meet him
16:15 hhardy hey
16:15 aorry tod rop in I am at work and can't stay
16:16 dogi jlew, cdeslandes did I report the situation accurately
16:16 hhardy O thought I should mention that Ed is going to shut down thinker to "reorganize"
16:16 dogi hi hhardy
16:16 hhardy I assume it will be coming back up but not sure when
16:16 I am copying my home directory off of there now
16:16 cdeslandes dogi: sound good to me
16:17 dogi cdeslandes, thx
16:17 hhardy I will try to touch bases with dogi maybe saturday
16:17 dogi ok hhardy
16:18 #action cdeslandes jlew decause rack aslocluster ...
16:19 by the way djbclark we did not move ippower ...
16:20 #link
16:21 questions?
16:21 next topic?
16:22 #topic fix lightwave backup
16:22 #info backup is too big with all the keys ...
16:22 #link http://monitoring.sugarlabs.or[…].org-libvirt.html
16:23 jlew dogi, back
16:23 dogi that is the purple spot there
16:24 _bernie arrives
16:24 silbe what is the purple spot?
16:24 dogi silbe, this is lightwave backup
16:24 silbe wow
16:25 does it change so much every day?
16:25 _bernie finished reading backlog
16:25 dogi _bernie, silbe how can we exclude the sks mount in the backup?
16:26 silbe dogi: I don't know the backup software used. But however you do it, just exclude /var/sks from the backup.
16:27 dogi silbe, don t thing that this is the problem it is more the amount of file which gpg has to check
16:27 s/thing/think/
16:27 silbe ah, ok.
16:28 dogi #action _bernie silbe exclude sks partition from lightwave backup
16:28 by the way should we delete the old backup there?
16:28 silbe what old backup?
16:29 dogi since the backup are incremental and we should try to not consume to much space on swan.l.o
16:30 ups +s
16:30 #link http://wiki.sugarlabs.org/go/Machine/lightwave
16:31 silbe add to notes: be careful what to run as root :-P
16:32 dogi :)
16:32 next topic?
16:32 #topic fix karmic template
16:32 #info bootproblems by migration ... grub was booting in graphic mode
16:32 #info will we have to fix all VMs?
16:32 _bernie dogi: that's easy, just add an exclude line in the backup script
16:32 dogi: there are many already
16:32 silbe: can you do that?
16:33 dogi _bernie, where is the configfile for this script ...
16:33 _bernie dogi: can we just drop the karmic template and use the lucid template from now on?
16:33 silbe _bernie: where? I would have expected something like /etc/duplicity/, but that doesn't exist...
16:33 _bernie dogi: the backup script has a config file, but not for excludes. let's just hack the script for now :-)
16:34 dogi: unless silbe or you want to implement it the proper way and add a new option
16:34 silbe: the config file is /etc/system-full-backup.conf iirc
16:34 silbe d'oh. can we at least document how the backup works? :)
16:35 _bernie: ok, that's the minimal config. and where's the script? :)
16:36 dogi only using lucid +1 _bernie ... but for that we should create a really new template not the one we used lucid lynx alpha 2 to install
16:36 -> next topic ...
16:37 for me is more important the question do we have to fix all VMs ...
16:37 silbe which are the grub modifictions we have to do there?
16:38 <silbe> serial --unit=0 --speed=115200
16:39 but i don t know where to add this ...
16:39 silbe /boot/grub/grub.conf or menu.lst
16:39 dogi silbe, thx
16:40 silbe for grub1 at least - for grub2 i'd need to look at my new VM
16:40 dogi ok
16:40 _bernie silbe: it's a stupid simple script come on
16:41 dogi: can't we just "aptitude dist-upgrade" the old template?
16:41 dogi: it's pretty much the same
16:41 dogi: and it takes a few minutes instead of a few hours
16:41 dogi _bernie, ok :)
16:41 silbe I've installed that one on an unpartitioned disk image, BTW - grub2 complains loudly, but works fine. Will be much easier to resize than the partitioned ones; maybe even online-resizable.
16:41 _bernie: but where does it live.
16:42 (the backup script)
16:42 _bernie silbe: the "serial --unit=0 --speed=115200" should be in all VMs already. it's in the template VM
16:42 silbe: /usr/src/devtools/sysadm/backup
16:42 silbe: it's on all our machines, including sunjammer
16:42 silbe _bernie: it wasn't on the meetings VM; maybe it was added to the template after meetings got cloned?
16:42 dogi _bernie, thx for /usr/src/devtools/sysadm/backup
16:43 _bernie silbe: yes, the meetings vm was installed way before we had a template
16:43 dogi _bernie, no meetingsVM is not meetingVM and is not a month old
16:44 -rw-r--r-- 1 dogi dogi 1692 2010-02-12 21:42 addmeetings
16:44 silbe _bernie: ok, /var/sks added to the script. I suppose I neither need nor should do any commit or anything else?
16:44 dogi ups 2and ahalf months
16:45 will jump to the next topic, ok?
16:45 silbe go on
16:46 dogi #topic access for olpc-nz
16:46 #info email ... tomeu and timclicks:
16:46 >> Do you really need physical access, or would be enough root access to
16:46 >> a virtual machine? That should be pretty easy to get.
16:46 >
16:46 > No, ssh access to something will be fine :) Ubuntu or other Debian
16:46 > derivative would be most familiar.
16:46 > Thanks Tomeu, who do I ask for help?
16:46 silbe what exactly are they asking for?
16:46 dogi _bernie, did we figure out what olpc-nz wants?
16:46 +1 silbe
16:47 _bernie silbe: please leave it uncommitted
16:47 silbe: devtools is shared with develer and is on svn because... because... I can't think of anything nice to say
16:47 silbe ;)
16:47 _bernie silbe: maybe we should fork it and put it on git.sugarlabs.org
16:48 silbe: their current sysadmins often break our scripts with careless commits which hardcode develer specific stuff all over the place
16:49 dogi hmm then maybe we should fork ...
16:49 silbe _bernie: unless there's anything private in it, that sounds like a good idea.
16:50 BTW, do we have any kind of policy on what may live on git.sugarlabs.org?
16:50 dogi hi timClicks
16:50 silbe timClicks: welcome!
16:51 timClicks hi there :)
16:51 dogi timClicks, backlog http://me.etin.gs/treehouse/tr[…]0504_1607.html#40
16:51 timClicks ty
16:51 hhardy Dogi is it ok for Ed to shutdown thinker now?
16:52 dogi hhardy, since you are the maintainer of thinker ... tell me if it is ok with you ....
16:53 timClicks, search for the topic access for olpc-nz
16:54 trusts hhardy s decisions
16:55 timClicks, can you tell us what do you (or olpc-nz) needs?
16:56 timClicks it would be great if I could have access to a low spec VM, e.g. 256mb RAM, so that we can test the usefulness of a LAMP stack application
16:56 called "Testlink"
16:56 the site seems to be down atm, Google Cache http://webcache.googleusercont[…]log.testlink.org/
16:56 hhardy well the agreement in writing was I could use till I got a job... which I have... so I guess he can do what he wants. I will "encourage" him to allow it to continue being used to support VIG and SL projects.
16:57 dogi #link http://www.flickr.com/photos/3[…]27@N00/134434138/
16:57 timClicks the project will be to create uniform test cases, to add some validity to the test process
16:58 dogi hhardy, +1 :)
16:58 timClicks and to enable different teams around the world to act as a single unit
16:58 as teams can pull down the latest test in the queue
16:59 dogi timClicks, is this a self coded lamp project?
16:59 silbe timClicks: what parts will you be hacking on? e.g. do you need a special web server?
16:59 timClicks dogi: I believe so
17:00 silbe: we'll be using testlink to test xo builds, sugar & activities
17:00 silbe timClicks: ah, ok.
17:00 though the build slaves might be fine for that :)
17:01 timClicks well, we'll see what we can achieve - if it seems worthwhile, I imagine that anyone from olpc/sl that would like manual testing can add a testcase
17:02 silbe in the long run I'd prefer automatic testing for most things, but that's a separate matter...
17:03 timClicks naturally, me too
17:03 dogi silbe, _bernie hhardy for me it sound reasonable to give timClicks a VM for this ... what is your opinion?
17:04 ups +s
17:04 timClicks, who will have access to the vm?
17:05 silbe dogi: depending on what exactly is going to get tested (or how), and what parts of LAMP need hacking, sunjammer + build slaves might do.
17:05 (just trying to keep the number of VMs we have to maintain manageable)
17:06 timClicks silbe: that will probably be fine - it may simply be easier if someone else installed the software & we just ran it
17:07 silbe timClicks: then let's try that way for now. If it doesn't work out, we can still go the VM way.
17:07 timClicks sure
17:07 dogi timClicks, for your information ... all users have to follow
17:07 silbe timClicks: just send a mail to systems@lists... explaining what exactly you need, and we'll set you up.
17:08 dogi #link http://www.sage.org/ethics/
17:09 silbe timClicks: see http://wiki.sugarlabs.org/go/S[…]l_account_request for how to get user accounts set up.
17:09 timClicks nods
17:09 silbe if the mail is GPG-signed, I'll do it, otherwise you'll need to wait for bernie ;)
17:10 dogi thx silbe :)
17:10 timClicks am on a locked down windows at uni today
17:10 I'll send a detailed noted in about 8-10h
17:10 dogi timClicks, ;(
17:10 silbe BTW, bernie: something like https://certifi.ca/ would be awesome for Sugar Labs.
17:10 dogi :P
17:12 timClicks, do you have more questions?
17:13 #action silbe help timclicks with testlink install
17:13 timClicks dogi: no
17:13 dogi silbe, ok?
17:14 silbe dogi: I'll try to, though I'm rather busy at least for the next few days (Bine has an oral exam on thursday).
17:14 timClicks #action timClicks to email systems@lists.sl.o
17:14 #action timClicks to request shell acct
17:14 dogi +1 timClicks
17:15 #topic ubuntu-lucid-buildbot
17:15 #info reduce subuntuVM to 3 vcpu?
17:15 #link http://monitoring.sugarlabs.or[…]labs.org-cpu.html
17:15 _bernie silbe: re: devtools having private stuff: yes, there's a directory that really doesn't belong to devtools, I'll drop it during the conversion to git
17:15 dogi: re: vm for timClicks: +1
17:16 silbe _bernie: ok. Just wanted to be sure no passwords etc. are accidently made public.
17:16 wow, there's really something running on that machine that utilizes several CPUs? :)
17:17 dogi yes
17:17 but in the wrong time ...
17:17 silbe though I'm wondering why it's _constantly_ doing something.
17:17 _bernie silbe: we have https://id.sugarlabs.org/ too
17:17 silbe 88% CPU average
17:17 _bernie silbe: we used to have passwords in the scripts many years ago, but they've been changed
17:18 silbe _bernie: but that's with passwords, not certificates, right?
17:18 _bernie silbe, dogi: yes. munin is a bitch
17:18 silbe _bernie: the cool thing about certifi.ca is that it works fully automatic and users can't be tricked into entering their password somewhere else.
17:18 _bernie silbe, dogi: if you want it to use little cpu power, you have to turn off the cron-based graph generation and use the cgi
17:18 dogi _bernie, munin right now is my best friend ...
17:19 _bernie silbe, dogi: which is hard to setup... the ubutu package is broken in several ways. there was also a very hard to find bug in munin's perl code
17:20 silbe, dogi: it has to do with a function being called as an argument to another function, which in perl changes its context to list context and makes it return a different value :-)
17:20 silbe one more reason for me to dislike Perl ;)
17:21 _bernie silbe: perl is the only langauge I know with type-base polymorphism on function *results* rather than *arguments* (like C++ does)
17:21 silbe: and java
17:21 silbe: oh, it's actually a very cool idea when it's used wisely.
17:22 silbe I suppose the number of people using Perl wisely is negligible :-P
17:22 but let's get back on topic
17:22 dogi hhardy, silbe _bernie are you ok with the reduction of subuntuVM to 3 vcpus?
17:23 silbe so the constant CPU usage is caused by munin?
17:23 dogi: why do you want to reduce it?
17:23 dogi cause they are running the build process during the also=web high time
17:24 ups also-web ...
17:24 silbe given that we have 4 CPUs, I wouldn't assign more than 2 CPUs to most VMs, but I'd like to know the reason for reducing just the build slaves.
17:24 *slave
17:25 dogi +1 silbe
17:25 silbe dogi: ah, OK. can we nice the build slave (in addition to reducing the number of CPUs, maybe even to 2)?
17:26 OTOH most CPUs in treehouse are idle almost all the time, so maybe just nice the slave and keep the number of CPUs assigned to it at 4?
17:27 dogi I was never ok with the rule to give all VMs already in the beginning 4 vcpus ... _bernie convinced me ...
17:27 _bernie dogi: what was subuntuVM for? and why does it need to reduce its cpus?
17:27 dogi: ah a build slave? also 1 cpu then... it does not need to be fast
17:28 dogi :)
17:28 no would reduce it to 3 ...
17:28 and or nice it
17:29 silbe _bernie: if possible, I'd prefer it to be able to use up to 4 CPUs, but nice it a lot. So it can use all the spare power most of the time, but will be second-class when other VMs need the CPU.
17:29 dogi the only problem is had was the wrong timing of the build action
17:29 _bernie dogi, silbe: for munin, you can copy the config from sunjammer. and watch out for file permissions in /var/something and for this tiny bugfix
17:29 dogi, silbe: for the bugfix, grep for "scalar" in the cgi source code
17:30 silbe: k... I dunno how to nice VMs in libvirt. there sure is a way
17:30 dogi silbe, ... same here
17:31 silbe dogi: what timezone do sunjammer and treehouse use, respectively?
17:31 dogi utc?
17:31 at least i hope
17:32 silbe because the Lucid build slave is scheduled to run at 03:05.
17:32 and that doesn't match with the load increase.
17:32 djbclark dogi: k, find something better?
17:33 dogi so it was a manual run ... silbe?
17:33 djbclark, no ...
17:33 jelkner dogi, ping me when you have a few moments to talk about the ST site.
17:34 silbe dogi: maybe a run by dfarning? Haven't spied on his crontab.
17:34 dogi ok
17:34 silbe dogi: but ideally the build slave should just be slowed down anytime more important stuff needs the CPU.
17:35 _bernie silbe: both ae in CET
17:35 silbe: sorry, both are in *EST*
17:35 silbe we can easily do that because it's an entire VM, whereas we couldn't nice SKS properly because it runs inside an important VM (lightwave)
17:35 _bernie djbclark: hola
17:36 silbe _bernie: what offset does EST have from UTC? maybe the configs are in UTC after all...
17:36 dogi thx _bernie
17:36 silbe, +5
17:36 jelkner, sure
17:36 _bernie silbe: EST is currently GMT+4
17:36 dogi: not now
17:37 dogi ok _bernie :)
17:37 silbe hmm, still no match, that would be 08h then, whereas the load increase starts at 05h.
17:37 hhardy dogi: sorry have not been reading but I don't know what is on subuntuVM atm so no opinion yet
17:37 dogi it was at 7 in the morning
17:38 silbe hhardy: it's running sugar-jhbuild (see https://buildbot.sugarlabs.org/) and some stuff for Ubuntu Sugar Remix.
17:38 hhardy EST = -5 UTC DST which is now = -4 UTC
17:39 how long does a build take now?
17:40 dogi hhardy, some minutes at 4 vcpu
17:41 would have reduced to 3 ...
17:41 silbe dogi: ah, wait: the 07:00 was a "manual" one - I restarted sugar-jhbuild on the slave today - it wasn't running for something reason.
17:42 it was also a full rebuild instead of compiling just the changed files (i.e. almost nothing)
17:42 dogi cause for exactly the same minutes I was not able to do something in treehouse
17:42 silbe I didn't think of it because it was afternoon here :)
17:43 dogi silbe, lets try the renice of the vm ...
17:43 silbe hhardy: I haven't checked lately, but when they all ran in parallel they were finished in less than an hour.
17:43 dogi: +1, especially ionice.
17:44 dogi jelkner, ping
17:44 #topic schooltool update
17:46 silbe dogi: we might also want to try the anticipatory scheduler - right now we use the deadline one. I've recently improved my knowledge about IO schedulers and it seems anticipatory should be much better for us.
17:46 dogi: it might even be all the change you need :)
17:46 but nice'ing the build slave is always a good idea.
17:47 dogi jupp
17:49 #topic rtVM update
17:49 #info this is the place to manage all sugarlabs first contacts
17:49 #link http://wiki.sugarlabs.org/go/S[…]/Getting_Involved
17:50 i would like to start to use rt with the getting involved page
17:50 changing this line ...
17:50 If you are an people person who would like to help with Sugar, please contact dfarning at sugarlabs dot org.
17:51 peopleperson@sugarlabs.org
17:52 i have asked dfarning already
17:52 and if we forward form rt to dfarning at sl o he is fine with that
17:53 lfaraone, silbe hhardy _bernie is this ok?
17:53 lfaraone dogi: sure, but you'll have to set up the queue :)
17:53 silbe dogi: sounds fine to me
17:53 as long as I don't have to do it ;)
17:54 dogi #action dogi set up the peoples person queue
17:55 lfaraone, are there other updates there?
17:55 ssl cert?
17:55 _bernie dogi: I'd prefer join@sugarlabs.org or something likethat
17:56 dogi: or volunteer@sugarlabs.org (but join is better)
17:56 dogi _bernie, hmm
17:56 lfaraone dogi: no.
17:59 dogi _bernie +/-0 ... we should the join@sugarlabs.org anyway ... but maybe not there ...
17:59 like the getting involved page
17:59 ups +s
17:59 _bernie dogi: I love it too
17:59 dogi: the http://join.sugarlabs.org page
18:00 dogi yes that is a the page for ... :P
18:01 #action treehousers brainstorm what to put on http
18:02 http://join.sugarlabs.org/
18:02 #info join@sugarlabs.org
18:03 #topic homework for every treehouser
18:03 #info "Solve an action per week"
18:03 #info is the main ritual in #treehouse channel
18:03 #info by doing so you get the right to vote ...
18:03 #action treehouser since an idea as action for getting vote right
18:03 #link http://idea.sugarlabs.org
18:03 <- this is the usual prelast topic ...
18:03 #topic other ...
18:06 thx for joining today the weekly infrastructure meeting
18:06 #endmeeting

Index | Today     Channels | Search | Join

Powered by ilbot/Modified.