Re: [Server-devel] [UKids] XO-1 classrooms don't reliably connect to many/most Wifi AP's
I've seen XO-1s randomly drop off the AP in Paraguay. I *think* we solved it by disabling 802.11s (the mesh network). This can be done at boot time by setting a parameter of the libertas kernel module. Sorry for being vague, it's been a long time ago. On 02/07/2014 10:10 AM, Adam Holt wrote: /[Terry Gillett summarizes his weeks of testing, ///with this very revealing report below/. That's tgill...@gmail.com mailto:tgill...@gmail.com of the Village Telco project: can we/Nepal/Lesotho/others help him add the key takeaways to http://wiki.laptop.org/go/Wifi_Connectivity so the almost 2 million XO-1s worldwide can benefit? Spoiler Alert: XO-1 deployments must carefully buy the correct Wifi Access Point, EG Linksys WRT54GL or Billion 7404VGP appear to solve most all problems. Likewise we've had a lot of success in Haiti with the TP-Link TL-MR3020.]/ SUMMARY The core problem is that XO-1 laptops will not reliably connect to a range of wifi Access Points (AP). By comparison, XO-1.5 and later laptops will successfully connect to the same APs. The behaviour of a group of XO-1s is different from that when they are tested individually. A single XO-1 may connect quite reliably, but when used in a group of ten or more, many individual XO-1s will fail to connect to the AP. Note that this issue is just about connecting to the AP, it is not about whether the AP can sustain a large number of connections or handle the associated data throughput requirements. A number of routers configured as APs have been tested to establish a baseline. The test process used requires 10 XO-1 laptops and is as follows: 1. Set up the AP on an unoccupied wifi channel, at least two and preferably three channels away from unoccupied channels. 2. Connect each XO-1 individually to the AP and check that it is operating correctly and has adequate signal strength. 3. Power off all the XOs 4. Start up one XO and allow it to connect successfully. 5. Start up the other none XOs 6. When the last XO has completed its boot up sequence, check the connection status of each XO. The result of a connection test for each XO is one of the following: 1. Successful automatic connection 2. No connection, but AP icon shows in Network Neighbourhood (NN) window 3. The AP icon does not appear. Typically there will be a mix of XOs in each of the three states. A Pass requires that all ten XO-1s are successfully connected at the end of the test without manual intervention. The proportion of XOs in each state will typically vary from 20 to 80% in a Failed test. The proportion of successful connections seems to vary by router type, but changes in repeated tests. Individual XOs will typically be in different states in repeated tests. A range of routers has been tested with this procedure and the results appear in the table below. The only two routers that passed the test were the Billion and the Linksys. Interestingly both these routers date from the same vintage as the XO-1. Note that testing with less than five XO-1s results in a much greater likelihood of a Pass result, and if the same AP is tested with ten XO-1s it will likely fail. A Pass result with ten XO-1s is considered (at this point) to be a reasonable indication of likely success in a real world deployment with greater numbers of XOs. The working hypothesis is that modern APs have implemented the wifi specs and/or default configurations in a way that has resulted in an interoperability problem with the wifi implementation in the XO-1. ROUTER TEST RESULTS Billion 7404VGP(old, Star Int, proprietary OS) Pass Linksys WRT54GL (old, Broadcom, DD-WRT) Pass Netgear FWG114P (old, proprietary OS) Fail TP Link WR710n (new, proprietary OS) Fail TP Link WR703(Atheros AR9331, OpenWrt) Fail TP Link WR842 (Atheros AR9287, OpenWrt) Fail TP Link MR3020 (Atheros AR9330, OpenWrt) Fail TP Link WDR4300(Atheros AR9341, OpenWrt) Fail VT MP01 (Atheros AR23xx, OpenWrt) Fail VT MP02 (Atheros AR9331, OpenWrt) Fail -- Unsung Heroes of OLPC, interviewed live @ http://unleashkids.org ! -- Unsung Heroes of OLPC, interviewed live @ http://unleashkids.org ! --- You received this message because you are subscribed to the Google Groups Unleash Kids group. To unsubscribe from this group and stop receiving emails from it, send an email to unleashkids+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out. -- _ // Bernie Innocenti \X/ http://codewiz.org ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
[Server-devel] ejabberd eats up 100% cpu time when it should be idle
Context: bernie alsroot: beam on jita is eating up cputime :-((( bernie alsroot: are we already running the latest version with all the patches? martin_xsa says he has fixed some bugs in the fedora 14 rpm. alsroot bernie: yup, this is fedora packages. also ejabberd /var is being recreated to the initial state every day bernie alsroot: we should tell this to martin_xsa alsroot bernie: thats why I was for something else for school server bernie alsroot: me too... my experience with ejabberd has been terrible so far. but let's ask martin first... he seems confident that it can be made to behave well. bernie alsroot: i'll forward this conversation to him This is ejabberd-2.1.6-4.fc14.x86_64 from Feb 24 2011. Is there a newer release, perhaps? -- Bernie Innocenti Sugar Labs Infrastructure Team http://wiki.sugarlabs.org/go/Infrastructure_Team ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
Re: [Server-devel] [Dextrose] Regarding my OLPC XS Wishlist
On Thu, 2011-06-02 at 11:57 -0400, Martin Langhoff wrote: On Thu, Jun 2, 2011 at 11:56 AM, Aleksey Lim alsr...@activitycentral.org wrote: Have you spent any time learning how to configure ejabberd? Diagnosing your problem? Discussing it on the ejabberd mailing list? Well, I assume OLPC people did it many times before me, I just reused their experience tryinhg to follow wiki.l.o docs and using native packages from fedora. Yes -- everytime we saw a perf problem we diagnosed. Right now we don't see performance problems when load testing the XS. What's the exact binary package of ejabberd and configuration that works well? How many users has it been tested with? I've had similar an experience similar to Aleksey with all versions of the ejabberd I tried, and so did the Collabora people I spoke with. I tried tweaking the configuration a bit, but the impression I got is that ejabberd is over-engineered for our needs (only 1 server, about 1000 users). If you see perf problems in your specific setup, I can only suggest you diagnose -- perhaps with the help from the ejabberd developers via their mailing list. Thanks. Send me your public ssh key, I'll give you access to the machine hosting jabber.sugarlabs.org. If you make it work, I'll buy you a green beer at EduJam 2012 :-) -- Bernie Innocenti Sugar Labs Infrastructure Team http://wiki.sugarlabs.org/go/Infrastructure_Team ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
Re: [Server-devel] [PATCH] Use LSB functions for initscript
El Mon, 23-08-2010 a las 15:09 -0500, Jerry Vonau escribió: Be careful with that one, check the Requires of redhat-lsb first, there are lots of rpms that will get pulled in because of it. Indeed :-( I've always been looking for a good cross-distro replacement for start-daemon. The LSB init-functions are really a half baked solution that doesn't even include a portable way to print OK/FAIL after starting something. -- // Bernie Innocenti - http://codewiz.org/ \X/ Sugar Labs - http://sugarlabs.org/ ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
Re: [Server-devel] [Tecnologia] Schoolserver development in Uruguay
El Thu, 19-08-2010 a las 22:02 -0600, Daniel Drake escribió: Actually, no cleanup was being done on that particular schoolserver. Are you sure? Yes, and now I know why: [r...@schoolserver ~]# /usr/bin/ds-cleanup.sh Traceback (most recent call last): File /usr/bin/ds-cleanup.py, line 42, in module import syck ImportError: No module named syck The package is also missing on our xs 0.6. It's simply a missing dependency in ds-backup-server, but it broke all schoolservers worldwide ;-) After installing syck-python, ds-cleanup.py still doesn't seem to do anything. I'll figure out why another day, now it's 2am again. There's more than that. It's a tool that makes you really think through the changes that your making. It helps you centralize everything. It also results in a configuration that results in the ability to upgrade from any point in time to the latest version. It would be less error prone in many ways. That's true. The problem with this sort of tools is that it's very hard to diagnose what went wrong. On a puppetized machine, a rule containing rpm -i package; do_something failed after the first command for some obscure reason. Then, the first command would also keep failing because the package was already installed. This time we could blame it on the incautious rule author who forgot --force, but you see what I mean: error paths become complex with puppet because the flow control is non-linear (like in make) and rule execution happens asynchronously and in thebackground (even more obscure than make). And if you have a mix of offline and online servers, its a no-brainer. The puppet benefits (vs shell script) for connected servers are very significant. And if you can just take a few easy steps to share the configuration with your offline servers, you save a lot of time. That is true. OTOH, one could also share a script to be executed off a USB stick or over the net with Puppet. As ugly as it may sound, humans would probably find it easier to maintain linear code that could be debugged interactively from the command line, independently of an asynchronous client/server system based on interdependent rules. [takes breath] -- // Bernie Innocenti - http://codewiz.org/ \X/ Sugar Labs - http://sugarlabs.org/ ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
[Server-devel] Initial port of idmgr to Debian
The following patch series makes idmgr work on the Plan Ceibal schoolserver out of the box, after configuring a different network address. [PATCH] create_registration: Directly create a v3 format database [PATCH] Use LSB functions for initscript [PATCH] Make users home directory configurable In order to get this accepted in Debian (and in Fedora), we'll have to relocate the package home from /home/idmgr to /var/lib/idmgr and probably also rename the package to olpc-idmgr or schoolserver-idmgr. ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
[Server-devel] [PATCH] create_registration: Directly create a v3 format database
Signed-off-by: Bernie Innocenti ber...@codewiz.org --- idmgr.spec.in |2 +- scripts/create_registration |4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/idmgr.spec.in b/idmgr.spec.in index bbdbcd5..9f8d09e 100644 --- a/idmgr.spec.in +++ b/idmgr.spec.in @@ -52,7 +52,7 @@ getent group xousers /dev/null 21 || groupadd xousers # Create the identity database, if there is no pre-existing one # and set the current rev number if [ ! -r /home/idmgr/identity.db ] ; then - # creates a v2 format file + # creates a v3 format file /home/idmgr/create_registration fi diff --git a/scripts/create_registration b/scripts/create_registration index b72a950..190b0f9 100755 --- a/scripts/create_registration +++ b/scripts/create_registration @@ -20,6 +20,6 @@ # create_registration # This script creates a new database for the registration server # -sqlite3 /home/idmgr/identity.db CREATE TABLE laptops ( serial VARCHAR(20) NOT NULL, nickname VARCHAR(200) NOT NULL, full_name VARCHAR(100) NOT NULL, pubkey TEXT NOT NULL, uuid VARCHAR(100), lastmodified TEXT DEFAULT '1970-11-12 12:34:56', PRIMARY KEY (serial) ) +sqlite3 /home/idmgr/identity.db CREATE TABLE laptops ( serial VARCHAR(20) NOT NULL, nickname VARCHAR(200) NOT NULL, full_name VARCHAR(100) NOT NULL, pubkey TEXT NOT NULL, uuid VARCHAR(100), lastmodified TEXT DEFAULT '1970-11-12 12:34:56', class_group INTEGER, PRIMARY KEY (serial) ) -[ -x /home/idmgr/storage_format_version ] || echo 2 /home/idmgr/storage_format_version \ No newline at end of file +[ -x /home/idmgr/storage_format_version ] || echo 3 /home/idmgr/storage_format_version -- 1.5.6.5 ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
[Server-devel] Intitial port of idmgr to Debian
For some reason, the cover letter of the previous patch series was not sent even though I had the --cover option in git-send-email (I blame Perl). ---cut--- The following patch series makes idmgr work on the Plan Ceibal schoolserver out of the box, after configuring a different network address. [PATCH] create_registration: Directly create a v3 format database [PATCH] Use LSB functions for initscript [PATCH] Make users home directory configurable In order to get this accepted in Debian (and in Fedora), we'll have to relocate the package home from /home/idmgr to /var/lib/idmgr and probably also rename the package to olpc-idmgr or schoolserver-idmgr. -- // Bernie Innocenti - http://codewiz.org/ \X/ Sugar Labs - http://sugarlabs.org/ ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
Re: [Server-devel] Looking for additional hands on XS
El Thu, 19-08-2010 a las 15:49 -0400, Martin Langhoff escribió: http://www.laptop.org/en/utility/people/opportunities.shtml Ha! I was just about to write you about some XS requirements for Uruguay :-) -- // Bernie Innocenti - http://codewiz.org/ \X/ Sugar Labs - http://sugarlabs.org/ ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
[Server-devel] Schoolserver development in Uruguay
calendars, lesson plans and even student records right inside the wiki. I have a dream that one day each school will evaluate and choose their favorite tools autonomously... but this is still a few years into the future. For the time being, we have to make a choice that would fit everyone and requires minimal remote management. If we make an impopular choice, teachers will simply start using Google Docs and other online tools. == Server management tools == Paraguay uses Puppet. We're very happy with it. Uruguay uses CFengine. They seem to be very happy with it as well. Both employ a flat hierarchy with one puppet master controlling all the schools, which is simple and straightforward, but requires excellent connectivity. Needless to say, comments are very welcome. Especially criticism. But no distro advocacy, please... they're all good, ok? :-) -- // Bernie Innocenti - http://codewiz.org/ \X/ Sugar Labs - http://sugarlabs.org/ ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
Re: [Server-devel] [Tecnologia] Schoolserver development in Uruguay
an issue that once with access you could delete your co-workers files, but they used it anyways A wiki has the same issues, but with versioning you can easily revert any vandalism, so it could be opened even to children. When a CMS was proposed, and later plone was accepted, i recomended dokuwiki, but for personal reasons i have to say, i have experience with dokuwiki, and deployment, updating, backup and other stuff is super fast (because it doesn't use a database, uses text files and folder schemes), but a being a wiki, is not as friendly to the newbie, because of the special syntax and all Except for Mediawiki, all the wikis I used use plain files or a VCS as their backend. Some wikis come with WYSIWYG ajax editors, but I find it simpler not to teach users much about the wiki syntax, so they mostly write plain text. The minimalistic wikis excel in not tempting users to waste too much time on syntax and concentrate on content instead. [...] If i think about what deplying in rwanda, uganda or even tibet, where the laptop is intended to go. There is no internet there, so their experience should point us to some key features that need to be included in the schoolserver Agreed. I'm sorry this got so long, but there was plenty to be said.. :-) Thanks for taking the time to summarize your first-hand experience, it's much appreciated. -- // Bernie Innocenti - http://codewiz.org/ \X/ Sugar Labs - http://sugarlabs.org/ ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
Re: [Server-devel] Schoolserver development in Uruguay
El Fri, 20-08-2010 a las 00:51 -0300, Bernie Innocenti escribió: Heh, these are good questions, but answering them all would take quite some time, and it's 1AM over here :-) Meanwhile, my du run to find out the size of current backups completed: # du -sh --exclude datastore-200* /library/backup 92G/library/backup So, backing up the last versions of all journals would take just 92GB, which would take more that 4 days on a 2mbit link for the initial backup. -- // Bernie Innocenti - http://codewiz.org/ \X/ Sugar Labs - http://sugarlabs.org/ ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel
[Server-devel] Plans for a new xs release?
Martin, I was wondering if there are any plans to rebase the schoolserver onto a more recent version of Fedora. We seem to have stability and performance issues with ejabberd bundled with the XS 0.5. Probably also on 0.6, since it's also based on Fedora 9. I've also been running ejabberd on Karmic for a while. Stability has been good for me, but it remains an arcane and hard to manage piece of software. Robert McQueen today proposed switching from ejabberd to Prosody, which supposedly has a more active and helpful upstream. For more info, see http://prosody.im/ . Thoughts? -- // Bernie Innocenti - http://codewiz.org/ \X/ Sugar Labs - http://sugarlabs.org/ ___ Server-devel mailing list Server-devel@lists.laptop.org http://lists.laptop.org/listinfo/server-devel