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

#treehouse, 2010-03-09

Index | Today     Channels | Search | Join

All times shown according to UTC.

Time Nick Message
16:09 dogi #topic agenda
16:09 #info * last meeting
16:09 #info * aslocamp at RIT - RIT/dfarning (http://aslo-web.sugarlabs.org )
16:09 #info * aslo node update - alsroot/dfarning (http://aslo1.sugarlabs.org )
16:09 #info * puppet update - dfarning
16:09 #info * lightwaveVM update - silbe/bernie
16:09 #info * cleaning of old VMs on treehouse/housetree - dogi
16:09 #info * sugar content sharing - alsroot/tonyf (http://idea.olpcorps.net/drupa[…]atorrent/idea/19/ )
16:09 #info * launchpadVM update - lfaraone (http://pad.sugarlabs.org )
16:09 #info * test fairy - treehousers (http://idea.sugarlabs.org )
16:10 #info * access fairy - treehousers
16:10 #info * discussion: build farm next steps
16:10 #info * discussion: new server for 2k
16:10 #topic last meeting
16:10 #link http://me.etin.gs/treehouse/tr[…]0100302_1606.html
16:10 #info action check:
16:10 #info action dogi/_bernie help dfarning with nfs for aslo-web - done
16:10 #link http://aslo-web.sugarlabs.org/en-US/sugar/
16:10 #info action dogi rename trac to puppet - done
16:10 #info - puppetVM was then recycled during the spring cleaning
16:10 #info action dogi delete aslo-proxy, aslo-db on treehouse - done
16:10 #link http://sunjammer.sugarlabs.org[…]irtual%20machines
16:10 #info action dogi delete aslo from housetree too - done
16:10 #link http://sunjammer.sugarlabs.org[…]irtual%20machines
16:10 #info action dogi delete beamrider on treehouse - not done
16:10 #info - bernie wants to use beamrider to test upgrade from jaunty to karmic
16:10 #info   as an example for sunjammer
16:10 #info action delete backup from aslo-proxy and aslo-db - done
16:10 #info action dogi send meeting to  #ubuntu-us-dc on freenode - done
16:10 #link http://me.etin.gs/ubuntu-us-dc[…]0100306_2037.html
16:10 #info open actions
16:10 #action dfarning move wiki recipes for aslo-machines to oldmachines
16:10 #action dogi create schooltoolVM for lfaraone
16:10 #action dogi create monitoringVM for hhardy
16:10 #action decause/cdeslandes give systems@list.sl.o update about result for CSI meeting
16:10 #action bernie pootle migration to housetree
16:10 #action dogi brainstorm hardware donation site for wikipediaservers
16:10 #action bernie/dogi ask on fedora cloud list for VMs
16:11 #action dogi research into spanify idea.sl.o
16:11 #action dogi update http://wiki.sugarlabs.org/go/I[…]ure_Team/Meetings and all the other meetings pages
16:11 #topic aslocamp at RIT
16:11 #info it is time to set up the next date for aslocamp
16:11 #info in the meantime the backup plan of dfarning is taking most of the traffic
16:11 #link http://aslo-web.sugarlabs.org
16:11 ping dfarning_laptop decause cdeslandes
16:11 dfarning_laptop pong
16:12 dogi how do we want to going on in this matter
16:12 what is the outcome of the meeting at RIT?
16:13 dfarning_laptop dogi, I don't know yet.  RIT was on spring break last week and the new semester started this week. So i have been leaving them alone to get settled.
16:14 dogi #info the private link between sunjammer was fixed by fsfadmins
16:14 dfarning_laptop I will contact decause this after noon unless hi chimes in durning the meeting
16:15 dogi #link http://sunjammer.sugarlabs.org[…].org-if_eth1.html
16:16 #idea should set up aslocamp when walter is visiting RIT
16:16 dfarning_laptop #link http://lists.sugarlabs.org/pri[…]March/001520.html
16:16 dogi wants to have at least a vague date
16:17 thx dfarning_laptop
16:17 dfarning_laptop ok, we will try
16:18 dogi +1
16:18 #agreed to try to set up aslocamp the same time as walter visits aslocamp
16:19 #topic aslo node update
16:19 #info eu[2-4] was renamed to aslo[1-3].sugarlabs.org
16:19 #info we are nearly finished with aslo node out of puppet
16:19 #link http://aslo1.sugarlabs.org
16:19 #link http://aslo2.sugarlabs.org
16:19 congrats dfarning
16:19 good work there :)
16:19 dfarning_laptop working with a university can be interesting.  You can wait and wait and wait for a single signature before you can proced.
16:20 thanks I already broke both nodes:)
16:20 dogi that is fine ...
16:20 dfarning_laptop this after noon will work on load balancer and database
16:21 dogi, are you going to try to set up aslo3:)
16:21 dogi jupp
16:21 dfarning_laptop great!
16:21 dogi actions?
16:21 dfarning, but we have to turn down at least one of the other 2 aslos
16:22 dfarning_laptop #action dfarning set up loadbalancer, database and backups.
16:22 let's turn off aslo2.  we need aslo1 as the puppetmaster
16:22 dogi #action transform aslo3 to an aslo-node
16:22 +1
16:23 #topic puppet update
16:23 #info puppet was moved to lightwaveVM
16:24 dfarning_laptop we have to do a bit more work, because commits are stepping on a simlink on lightwave
16:24 #action dfarning silbe fix symlink issue on lightwave for puppet.
16:25 I am happy with the work flow
16:25 we can do all tests on aslo1, the test puppetmaster.
16:25 dogi dfarning, silbe: how is the upgrade process for puppet working now? I m right that there is no account on lightwave needed for doing this
16:26 dfarning_laptop then when we are happy, we can push the commits to lightwave for production.
16:26 dogi #action dfarning silbe fix symlink issue on lightwave for puppet
16:26 dfarning_laptop I belive that only you, bernie, silbe, and I can push to lightwave.
16:27 anyone with access aslo1 can hack on puppet
16:27 dogi hmm ok that sound reasonable
16:27 silbe dogi: upgrade?
16:27 dogi silbe, sorry push ...
16:28 silbe ah, OK.
16:28 dfarning_laptop but they must ask for a code review for you bernie of silbe to push
16:28 dogi interesting workflow ...
16:29 dfarning_laptop puppet is a strange best.  If you control puppet you can control any machine which pulls from puppet.
16:29 s/ best/beast/
16:30 dogi :)
16:30 dfarning_laptop currently _no_ machines automatically pull
16:30 dogi is that possible
16:30 cron job?
16:31 or is there a better way to do this
16:31 dfarning_laptop yes, typically puppet is either run from a dameon or a cron job.
16:32 there are memory leak issues in ruby and rail so most people prefer to use a cron job.
16:32 dogi hi blake ... we are in a meeting ... backlog: http://me.etin.gs/treehouse/tr[…]0100309_1609.html
16:33 next topic?
16:33 silbe?
16:34 silbe dogi: yes?
16:34 dogi #topic lightwaveVM update - silbe/bernie
16:34 #info silbe fixed some of the problems with git and nameserver update
16:34 #idea lightwave on housetree?
16:34 silbe nameserver should be working fine, including automatic propagation after pushes
16:34 dogi silbe tell us about the changes made on lightwave ...
16:35 dfarning_laptop thinks dogi runs a tight ship:)
16:35 silbe we can remove the OSU-OSL slaves now as well
16:35 dogi likes the automatic emails when something is changed there
16:36 silbe haven't worked on converting the zone files to tinydns-data format yet (so we don't need to manage serial numbers and reverse zones manually)
16:36 dogi #idea remove OSU-OSL slave from the nameservices
16:37 silbe it's rather an action - bernie already did the first step, but we need to finish it
16:37 will do it later
16:37 dogi what do you think about a second ns on housetree?
16:38 silbe I'd tried installing SKS (PGP keyserver) on lightwave so we don't cause load on public keyservers once we use monkeysphere. Unfortunately I ran out of space...
16:38 dogi ohh cool
16:39 #action bernie/dogi add more harddrive space on lightwave
16:39 how much do we need?
16:39 silbe dogi: we already have name servers on lightwave and sunjammer. housetree is redundant to one of them, though I don't know to which one :)
16:40 dogi: I'd save +20G so we have enough room to grow. On my old server it took 14G.
16:40 *say
16:40 it's really too late already...
16:41 dogi thx silbe :)
16:42 should i write down an action item for next week, silbe?
16:42 silbe do as you please :)
16:43 dogi #action silbe keyserver for monkeyshere on lightwave
16:43 #topic cleaning of old VMs on treehouse/housetree - dogi
16:43 #info spring cleaning was successful
16:43 #link http://sunjammer.sugarlabs.org[…]irtual%20machines
16:43 #link http://sunjammer.sugarlabs.org[…]irtual%20machines
16:43 #info by doing this I found an error between mdadm and lvm:
16:43 #info "vgdisplay treehouse"  is showing more harddrivespace then "cat /proc/mdstat"
16:43 #info 1.78TB instead of 0.9xy TB
16:43 #action dogi/bernie fix md1/lvm on treehouse
16:43 don t know how to solve this problem ...
16:44 _bernie> dogi: I have no time to look into the treehouse problem now... but I thought of a possible solution
16:44 <_bernie> dogi: it involves removing one drive from md1, format it as md2, add it to the vg, then move all filesystems to md2, then remove md1, and then add all the drives to md2 :-)
16:44 <_bernie> dogi: easy, eh? :-)
16:44 but this was bernie answer to it ...
16:45 _bernie hey
16:45 dogi and we both agree that it was cause by my first install as a raid5 system with lvm on top
16:45 hi _bernie
16:45 silbe bad dogi, you woke _bernie ;)
16:45 _bernie sorry guys for missing the meeting I was super busy
16:45 dogi _bernie, you are right in time
16:46 we are speaking about the harddrive issue on treehouse
16:47 _bernie dogi: ah yeah. can we work on it this sat/sun when I'm back from caacupe?
16:47 dogi: do you know already how to add disk space on lightwave?
16:48 dogi: you can use either virsh or qemu-img
16:48 dogi _bernie, good question ...
16:48 I don t remember the 2 commands
16:48 _bernie dogi: if it's really a qcow, it can be grown on the fly without rebooting the machine. but I found out that some images are *named* .qcow but they are in fact vulgar images.
16:49 dogi oh right lightwave is not lvm
16:50 hmm should we move lightwave to lvm then? bernie?
16:50 silbe lightwave wants a reboot anyway
16:50 dogi silbe, kernel upgrade?
16:51 silbe so just take her down and move to LVM
16:51 dogi: no idea, motd complains about it
16:51 gives no reason AFAIR
16:51 dogi then it is normally a kernelupgrade ...
16:52 silbe OTOH motd is incorrect anyway (e.g. number of users doesn't match "w" output)
16:53 dogi the other thing we could do is to mount some lvm space into lightwave ...
16:53 likes this idea -> quickfix
16:53 silbe dogi: that would work equally well as the space is only needed for /var/lib/sks
16:54 _bernie dogi: I vote for lvm
16:54 dogi :) _bernie
16:54 silbe, ?
16:54 silbe _bernie: additional or converting?
16:54 _bernie dogi: to move it, we have a procedure somewhere...
16:55 dogi _bernie, but I don t feel confident do to that alone ...
16:55 _bernie hmm not really: http://wiki.sugarlabs.org/go/S[…]e_virtual_machine
16:55 dogi screen meeting?
16:55 _bernie this is to migrate guests from one host to another using lvm
16:55 I can only work on it this week-end...
16:55 I have too much urgent stuff to do now
16:55 dogi thats fine
16:56 _bernie silbe: would it bother you if we delayed the gpg keyserver thing a few more days?
16:56 silbe then let's just add an LV to it, that's low-risk
16:56 _bernie: not really, it's been stalling for months now
16:57 dogi think that the one is not blocking the other ...
16:57 _bernie silbe: do you think mail that goes through @sugarlabs.org is being spam-filtered by google?
16:58 dogi I would start with the additional lvm space and then move the hole thing to lvm ... do we agree?
16:58 _bernie dogi: yes and no... if you're going to create the lv, format it, and add it to the virtual machine... then you're almost there!
16:59 dogi: all you'd need to do to complete the migration is:
16:59 dogi #link http://wiki.sugarlabs.org/go/S[…]e_virtual_machine
16:59 _bernie 1) find a way to loopback-mount the old qcow partition (I think it can be done with some crap libvirt tools packaged for ubuntu)
16:59 2) use rsync -PHAXphax --delete /foo/ /bar/ to copy everything over
17:00 3) change the libvirt config to use the "external kernel", like the other VMs based on lvm
17:02 silbe _bernie: "all but", yes - those are the complicated steps :)
17:02 _bernie: having SKS in a separate LV isn't bad even in the long run - it ensures the keyserver doesn't fill up /
17:03 dogi and blocks hole treehouse with a load of 13-14 ...
17:04 silbe dogi: what blocks treehouse?
17:04 _bernie silbe: ok yes
17:05 I see
17:05 dogi #action add lvm to lightwaveVM for keyserver
17:06 #topic test fairy
17:06 #info eat your own dogfood
17:06 #info since web2.0 has the same problems a party has to being good
17:06 #info we have to produce some content for having this service starting
17:06 #idea every treehouser has to produce an idea for the next week
17:06 hope we all agree there :P
17:07 #link http://idea.sugarlabs.org/
17:07 silbe dogi: don't think so :-P
17:07 is anti-web 2.0 :D
17:07 dogi :P
17:07 #topic access fairy
17:07 #info - dogi need wikiadmin rights
17:07 #info - jasg wants an account
17:07 #info   Jorge Saldivar - jorgesaldivar at gmail dot com
17:07 _bernie dogi: btw, idea.sugarlabs.org redirects to idea.olpcorps.net... is it possible to keep it on the same domain?
17:07 dogi: also, is it possible to update to drupal6?
17:08 dogi yes and no
17:08 the first is possible
17:08 the second unfortuantly no
17:08 _bernie dogi: for wikiadmin rights, I can move the wiki to sunjammer, but I'd like someone to be a full-time maintainer for it before I do.
17:09 dogi: I asked cjl and fgrose if they would want to do it... but I had no time to check my email yet so I don't know if they have answered.
17:09 dogi nono bernie
17:09 wnated to delete a page there
17:10 but have not the right needed
17:10 _bernie dogi: ah... I think I could add you...
17:10 dogi thx bernie
17:11 _bernie dogi: also cjl and fgrose could... and they probably also remember how :-)
17:11 dogi: I said "I think I could" not because I'm uncertain if you should be admin, but because I'm unsure how it's done :-
17:11 :-)
17:11 also on the access fairy thing, I'd like to give lfaraone admin access to google apps to do various things. since there's no granular privilege control, I'd ask him to not edit other settings.
17:12 lfaraone _bernie: so you don't want me snooping in walter's email account? :P
17:12 dogi :)
17:12 silbe lfaraone: just ask walter himself - I already got his data store ;)
17:13 (for testing version support)
17:13 dogi agree to give lfaraone access to google apps
17:13 silbe abstains - no idea what Google Apps access includes
17:14 lfaraone silbe: if you've been talking to Eric Schmidt, Walter shouldn't have a problem, since he shouldn't be doing anything he doesn't want to be known :)
17:14 dogi does somebody know jorge?
17:14 silbe lfaraone: who is Eric Schmidt?
17:14 dogi is curious ...
17:14 silbe dogi: IIRC he develops an activity for blind users
17:15 dogi hmm thanks silbe
17:15 _bernie lfaraone: I think not even admins can do that :-)
17:15 lfaraone silbe: CEO of Google. http://www.eff.org/deeplinks/2[…]dismisses-privacy
17:15 silbe dogi: he has low bandwidth and needs to update and upload a large activity on a.sl.o
17:15 _bernie lfaraone: oh, well, unless one sets up a redirect
17:16 silbe _bernie: DNS hijacking anyone? :)
17:16 lfaraone _bernie: or changes a user's password.
17:16 _bernie silbe: jorge works in my same room. I know him well. he's not very good with the shell (yet), but he won't do any harm either.
17:16 silbe: we should sign our zone btw:
17:17 dogi #idea dogi will change idea.olpcorps.net to idea.sugarlabs.org if there are 40 ideas there ...
17:17 _bernie http://www.nlnetlabs.nl/public[…]_howto/index.html
17:17 dfarning_laptop dogi, I added you to wiki administrators and bureaucrats
17:17 _bernie dfarning_laptop: thanks! how is it done btw?
17:17 dogi thx dfarning_laptop :)
17:17 the is a speical page
17:18 s/the/there/
17:18 silbe _bernie: ah, cool! Can you check&sign his PGP key then, please?
17:18 dogi should we move to the discussions?
17:18 dfarning_laptop if you are a bureaucrat there is a user management special page
17:19 silbe lfaraone: thx for the link, interesting read
17:19 dogi #topic discussion: build farm next steps
17:20 silbe (and only alarming if one didn't already suspect Google has that kind of attitude)
17:20 dfarning_laptop #link http://lists.sugarlabs.org/pri[…]March/001518.html
17:20 background reading on buildfarm
17:21 is there enough resources on bender to move everything there
17:22 Hardware
17:22    * Crappy minitower
17:22    * Core 2 Duo 1.8GHz
17:22    * 4GB RAM
17:22    * 2x1TB RAID1, 2x320GB RAID1
17:22    * Fedora 11 x86_64
17:22 [edit]
17:22 1TB is room for alot of vms
17:23 dogi 4gb ram is not much ...
17:23 dfarning_laptop _bernie,   can/should we add ram?
17:23 silbe the RAID doesn't have any VM on it
17:24 *1TB RAID
17:24 _bernie dfarning_laptop: for the build bots, I think it's fine...
17:24 dfarning_laptop silbe, what else is bender used for?
17:24 dogi but on the other side buildmachines don t need to run the hole time ... or am I wrong? silbe?
17:25 _bernie silbe: btw, bender is now running Fedora 12
17:25 dogi _bernie, :)
17:25 _bernie dfarning_laptop: oh, we have 6GB of ram now
17:25 silbe RAM is 6GB already (thanks _bernie!), Fedora got upgraded to 12 (thx again)
17:25 _bernie dfarning_laptop: can you update the wiki pls?
17:25 dfarning_laptop dogi, the vm only need to run a few minutes a day.
17:25 _bernie, yes
17:26 _bernie dfarning_laptop: I also *think* the develer sysadmins were kind enough to move it into a nice 2U rack case... but I need to ask them again
17:26 dogi #info bender runs now F12 and has 6gb of ram
17:26 dfarning_laptop dogi, but it is a total pain in the ass to dynamically turn vms on and off!
17:26 silbe dogi: they don't strictly need all to run all the time (except for the Fedora ones that are used for SoaS building etc.), but it simplifies managing them a lot
17:27 dogi dfarning_laptop, I know
17:27 dfarning_laptop much easier to just let them sit in memory.
17:27 what else runs on bender?
17:28 silbe dfarning_laptop: nothing AFAIK
17:28 dogi This leads us to:
17:28 1. What releases should we build?
17:28 2. How can we make the system more maintainable?
17:28 3. What features do we want to add to the build machines?
17:28 silbe there was asterisk running on it some time, but not anymore
17:29 dfarning_laptop Is there room on the raid for a nice lvm to hold vms?
17:30 silbe #1: "We" don't build any release on there currently - it's others that use our slaves to build SoaS, F11-on-XO1 etc. image
17:30 dfarning_laptop if nothing else runs on bender is it worth switching to ubuntu?
17:30 silbe buildbot is for testing whether stuff compiles fine and whether the test suite finishes successfully
17:31 In order to maintain sugar-jhbuild, I'd like at least one VM per supported distro release, so I can try it out myself.
17:31 _bernie #info I added lfaraone to group admins of google apps
17:32 dogi thx _bernie
17:32 _bernie #info I removed fgrose and mchua from the admin group because I could not remember what they were supposed to do with admin privileges. if they really needed it, we'll hear their complaints
17:33 silbe dfarning_laptop: all storage on bender is managed via LVM
17:33 _bernie #info I enabled the "user managed groups" feature in google apps, but it seems to behave in an unwanted way, I'll turn it off again
17:33 silbe (on top of RAID1)
17:34 dfarning_laptop silbe,  so there is lots of room?
17:35 dogi dfarning_laptop, at least it sounds like
17:35 silbe dfarning_laptop: as for Ubuntu, I'd prefer it over Fedora, but it's probably not worth the trouble. The interesting stuff happens inside the VMs, not on the host.
17:35 dogi +1 silbe
17:35 silbe dfarning_laptop: just checked, we still have ~190G unallocated, that should be enough for now
17:36 dogi 19 or more VMs
17:36 dfarning_laptop ok can you give me access to bender and we can figure out a why to build the vm using puppet.?
17:36 dogi silbe, where are you saving the builds?
17:37 dfarning_laptop s/why/way/
17:37 silbe dogi: saving which builds?
17:37 _bernie oh, wait: can someone who is not a google apps admin please look at this page and tell me which groups appear here: https://groups.google.com/a/sugarlabs.org
17:37 silbe _bernie: none, it's a login page
17:37 dfarning_laptop: sure. can you remind me tomorrow?
17:38 dfarning_laptop sure.
17:38 are the vm qemu or xen?
17:38 dogi hopes kvm ...
17:38 or some chroot stuff
17:38 _bernie silbe: login with your sugarlabs.org account please
17:39 silbe dfarning_laptop: KVM
17:39 dogi :)
17:39 _bernie dfarning_laptop: bender uses kvm
17:39 dfarning_laptop ok good that should be straight forward
17:39 _bernie dfarning_laptop: I can give you access
17:39 silbe _bernie: give google my sl.o LDAP password? sure?
17:39 _bernie dfarning_laptop: I think you have it already, no?
17:39 dfarning_laptop: try dfarning@bender.sugarlabs.org
17:39 silbe: no, not that one
17:40 silbe: don't you have a google apps account too?
17:40 dogi silbe, _bernie for me too
17:40 silbe _bernie: does everyone with access on sunjammer have access to bender as well?
17:40 _bernie: I hope not :-P
17:40 dfarning_laptop yes I have access.
17:41 I'll poke around and post a plan on the systems ml
17:41 _bernie silbe: yes and no. they are unix users on bender, but with no ssh keys, home dir...
17:41 silbe _bernie: ah, I see.
17:41 dfarning_laptop silbe, how about apt-catcher and a rpm proxy?
17:42 _bernie silbe: did you create dfarning's home? or was it created on the fly on first login?
17:42 silbe dfarning_laptop: we need to follow up with Develer on apt-cacher. Bernie, can you do that?
17:42 _bernie: not that I remember at least. certainly not today.
17:42 _bernie silbe: ok
17:42 silbe _bernie: thx
17:42 _bernie silbe: as soon as I can get hold of an admin
17:43 silbe: feel free to ping them if you find the original thread
17:43 dfarning_laptop silbe, devel is running their own proxy which we can use?
17:43 _bernie silbe, do you need a google apps account?
17:43 dfarning_laptop: uh? what proxy?
17:44 silbe dfarning_laptop: they have an apt-cacher instance we can use, yep. for rpm we need to setup something on bender
17:44 _bernie dfarning_laptop: we'd need squid running on bender imhi
17:44 silbe _bernie: what use is it?
17:44 _bernie imho
17:44 silbe _bernie: (which probably means no)
17:44 _bernie so it would do both debs and rpms...
17:44 dogi _bernie, silbe since there is kvm on bender I would like to have access there too ...
17:44 _bernie I really do not see the motivation behind a specialized tool such as apt-cacher.
17:45 silbe _bernie: the motivation is develer maintains it already and squid doesn't work perfectly well for this task.
17:45 _bernie silbe: none... I was actually thinking that maye we could stop using it. it's only used for aliases, which we can do easier on sunjammer.
17:45 silbe _bernie: I still need to figure out why it's handing out stale Packages files here
17:45 _bernie silbe: develer does not want to maintain squid because rasky (one of the bosses) does not understand it :-)
17:46 silbe _bernie: then no, I don't need an account :)
17:46 _bernie silbe: I have a trick for squid... wait...
17:46 silbe _bernie: squid would be maintained by us (for use with rpm and misc http stuff), though I cannot claim to really understand it either
17:46 _bernie silbe: the default config is BROKEN
17:47 silbe _bernie: I set refresh-ims which should fix exactly my issue, but it doesn't make a difference :(
17:48 _bernie === Fix in squid.conf ===
17:48 #bernie: fuck, this makes squid serve stale objects such as repomd.xml!
17:48 #refresh_pattern .              0       20%     4320
17:48 refresh_pattern .               0       0%      4320
17:48 silbe: this one
17:48 dfarning_laptop dogi, should we move onto the next topic?
17:48 dogi jupp
17:49 _bernie silbe: if we need squid to cache rpms, then we should not work harder to prevent it from caching also debs :-)
17:49 dogi #topic discussion: new server for 2k
17:50 dfarning_laptop, was only waiting for somebody to tell me to move to the next and last ome
17:50 silbe _bernie: hmm, wouldn't that cause squid to check back to the original server every time? need to read the description again...
17:50 dogi one
17:50 dfarning_laptop _bernie, new server?
17:51 dogi which is the time frame there?
17:53 _bernie we don't have one, but... as fast as possible!
17:53 dogi: when are you leaving from boston btw? we'd better move before this
17:53 dogi ok good to know
17:53 _bernie btw, I talked with Ivan last night. He had in mind to buy a $2300 HP machine for himself.
17:53 $2300 with no ram and drives
17:54 he offered to use it for SL too, at the condition that we buy the ram and drives.
17:54 I mentioned that we need admin accounts besides me and him
17:55 ivan is very paranoid for security, and is also very reluctant to use virtualization.
17:55 dogi I know ... to buy servers was cheaper last year ... ram costs now 30% more ...
17:56 _bernie but he would agree to use kvm provided it does not create stability or security problems
17:56 dogi: :-(((
17:56 dogi: because the $$$ went 30% down, I suppose
17:56 dogi not only
17:56 ram is short on the market right now
17:57 that means the price is higher
17:57 silbe _bernie: in case we'd go for that option, who would get access and what would we host on it?
17:57 dfarning_laptop sorry if this has been asked and answered... but with the load on sunjammer below 20% what is the new server going to be used for?
17:57 silbe certainly not a.sl.o if neuralis is even more paranoid than I am
17:58 dogi lets think about the base design the new server should have
17:58 a dual processor quadcore at least?
17:59 _bernie silbe: the VM(s) would host  the same stuff that is currently on solarsail: wiki, lists, planet... trac could come back... also aslo if we want to..
17:59 silbe: and we could give root to few trusted people only. but it's not necessary to admin any of those things: mediawiki, trac, planet and mailman...
18:00 dfarning_laptop: oh, sunjammer is so low on cpu usage now?
18:00 dfarning_laptop: there does not seem to be the same amount of traffic to aslo as before.
18:00 dfarning_laptop: maybe latu are redirecting kids to their drupal site?
18:00 dogi _bernie, most is now on treehouse ....
18:01 silbe _bernie: and who could get shell access?
18:01 _bernie silbe: yeah, neuralis told me it would be crazy to run something as heavyweight as aslo alongside critical services such as email, web, wiki...
18:02 dogi #link http://sunjammer.sugarlabs.org[…].org-libvirt.html
18:02 _bernie, +1
18:02 silbe _bernie: I tend to agree :)
18:03 dogi and treehouse is byos
18:03 build your own server ...
18:03 so even less stable
18:03 _bernie silbe: I'd go back to the original (as in beginning of SL) plan: a people's machine for shell accounts and less important services (sunjammer) and one only for the core services
18:04 silbe: in this picture aslo could be spread out across multiple machines (at rit, or treehouse + sunjammer), or even run on the primary machine
18:04 silbe: the primary machine would be 8 cores, with each core 2 times faster than our current sunjammer.
18:05 silbe: so... who gives a damn if it uses some cpu? :)
18:05 silbe _bernie: as long as the people who are maintaining the primary services (wiki, Trac, etc.) have no trouble getting shell access, I won't argue against that plan.
18:05 _bernie silbe: if we confine aslo in its own VM, it will be scheduled as a normal process by the guest, so if it uses too much cpu time its priority will simply sink.
18:06 silbe _bernie: it's not just CPU - causing the machine to swap hard can impact other stuff pretty bad
18:06 _bernie silbe: so, there was a misunderstanding on this: ivan has always been ok with giving normal user accounts on solarsail to few trusted individuals. just not root.
18:06 silbe _bernie: ok
18:06 (shell accounts)
18:07 _bernie silbe: and we could of course give root to others on the VMs.
18:07 silbe _bernie: ah, even better.
18:07 _bernie: then I'm perfectly fine with the plan in general
18:07 _bernie silbe: you're right on swap, although it's the VM swapping, not the guest.
18:07 silbe I hope his planned server can use ECC RAM? :)
18:07 _bernie silbe: I/O bandwidth can be somewhat confined, but... not very effectively imho.
18:08 silbe: I hope too.
18:08 dogi: how about you?
18:08 dfarning_laptop: and you?
18:09 silbe _bernie: yes, the SKS keydump import on lightwave even caused ping RTTs (from outside) to increase by over an order of magnitude - even though ioniced and niced to the max :(
18:09 dogi is he relucant or willing to give VMs ...
18:10 dfarning_laptop _bernie, based on past experience with solarsail I don't think this is a good idea.
18:10 _bernie silbe: did anything happen to the other VMs? I'd consider it a serious problem in kvm if it did.
18:10 dogi: he said ok only on the condition that they don't cause instability and security holes
18:11 dfarning_laptop sugar labs is better off with a smaller machine which it owns 100%
18:11 silbe _bernie: haven't checked the Ubuntu buildslave which is the only other VM I have access to on housetree/treehouse
18:11 _bernie dogi: I think he's skeptical, but I know for sure that kvm is stable and relatively secure because we've been using it for a long time
18:12 dfarning_laptop: I agree, but with $2k I can only buy a 4 core machine with no monitoring, no redundant fans and only 8GB of ram :-(
18:12 dfarning_laptop: did I send out the proposals?
18:12 dogi I don t have past experience but I don t like machines where we have always to ask permission ... on the other side I do understand him ... it is his machine ...
18:12 _bernie dfarning_laptop: oh yes I did. they were kind of depressing...
18:12 dfarning_laptop: another possiblity would be buying crap hardware from Supermicro
18:13 silbe dogi: this is only for primary services, so not so bad
18:13 dogi ala treehouse/housetree they are asus crap :P
18:13 _bernie dogi: I'm also a little wary... I worked with ivan for over 1 year now without any incident. he's been a good "guest" for us.
18:14 dogi but _bernie you are right it is the best plan we have for now ...
18:14 _bernie dogi: on the other hand, he's really paraonid for security :-)
18:14 silbe _bernie: well, with BIND and all that stuff running at sugarlabs at bit of paranoia doesn't hurt <gdr>
18:15 dogi _bernie, did I told you that with some $$$ the wikipediaserver could be transformend in really useful machines
18:15 silbe ok, getting serious again
18:15 _bernie silbe: lol
18:15 dogi: but they are only 2 cores, right?
18:15 dogi: and no redundant power supply, no management...
18:16 dogi: good enough as web slaves, but not as our #1 machine.
18:16 dfarning_laptop going to eat dinner.  I have expressed my reservations... but luckily I am not responsible for anything and am happy playing in my own sand box.
18:16 dogi sure but how much you want to spend for ram and drives?
18:17 silbe FWIW I tend to agree with bernie about the usefulness of a high-end server for our primary services
18:18 it's fine to have build slaves, a.sl.o etc. on PC crap
18:18 but for Trac etc. we should do better
18:18 dogi with 50$ per cpu they would be dual-core, so every machine would have 4 cores and could have 32 gb ram
18:18 silbe (for a.sl.o it's only OK because we're making it redundant over several servers, so a single dead machine doesn't hurt)
18:19 dogi and ipmi is buyable too
18:19 silbe dogi: but not redundant power supplies I guess
18:19 dogi: and power supplies are dying rather often
18:19 dogi silbe, thats then an other 100$
18:20 silbe dogi: just 100$ for two redundancy-capable, hot-swappable power supplies? sure?
18:20 dogi wanted only say this once
18:20 hot-swappable???
18:20 silbe in germany I pay that much for a regular power supply
18:20 dogi for what
18:21 I understand dualpower
18:21 silbe dogi: for fixing a broken power supply without major downtime
18:21 dogi and as long you have enough machines to move the services in the meantime
18:22 _bernie dfarning, silbe, dogi: I'll check with the other slobs
18:22 dogi +1 _bernie
18:22 #info next meeting agenda
18:22 silbe dogi: that's exactly the point: for stuff like a.sl.o that scales easily we can use several PCs
18:22 dogi #link http://openetherpad.org/S4ogRbK1YK
18:23 silbe dogi: but for Trac etc. that are high-priority and not easily distributable, we need a high-end server
18:23 i think there was something else I wanted to discuss...
18:23 dogi: do you remember what that was?
18:23 dogi silbe, wanted only to say if we only have to but ram and harddrives I would not spend all money for this one machine
18:23 silbe dogi: you told me to ask again here in the meeting
18:24 dogi: no, buying the new machine is about reality, not additional computing power
18:24 *reliability
18:24 is getting really tired :(
18:24 dogi lfaraone> something else to add to the agenda: mails to feedback@sugarlabs.org are rejected if the user does not subscribe to IAEP, and held in moderation if they do, since the message "Reason:   Message has implicit destination                     Spam? 0  "
18:24 nearly forgot this ...
18:25 _bernie, is this wanted?
18:26 #endmeeting

Index | Today     Channels | Search | Join

Powered by ilbot/Modified.
Webmaster