[Server-devel] Networking options
Tony Pearson
tpearson at us.ibm.com
Sat Feb 9 00:23:24 EST 2008
Team,
Here are my thoughts on networking.
In the current School configuration:
http://wiki.laptop.org/go/XS_Configuration_Management#School_Specific_Configuration
it has the XS being connected directly to the ISP modem. All access is
through the primary school server. If it fails, all is lost.
1--I suggest instead (modem)--(WiFi)--(XS)--(Mesh)
A simple four-port hub could support 4 direct-connect items (an XS server,
a Library server, a printer, etc.) and over 200 Wireless.
2--For normal operations, a WEP key can prevent any XO laptop from using
the Wireless directly. Instead, they use the XS server,
which has all the squid cache, library server cache, etc. The WEP key can
be provided to teachers to access the system directly
via WiFi from their laptops.
3--In the even the XS is down, or the Mesh Active Antenna is down, a
teacher can turn off WEP (by accessing the Wifi hub from their laptop),
and open the WiFi up to all XO laptops. This would also mean no content
filtering, squid caching, etc.
4--Alternatively, leave the WEP key in place, and if a student needs to
update an activity during the time the XS server is down, the
teacher can either download the XO file on their behalf and send it to
them via mesh, or enter the WEP key on the student's XO for
that exception. In this mode, everyone just uses their XO and meshes with
each other, but has no access to the outside internet,
moodle or the library cache.
A step up from this would be:
(modem)--(Wifi 1)--(XS)--(Wifi 2 + Mesh)
In this environment, WiFi 1 would be WEP-protected, teachers only. WiFi 2
would be open, and complement the Mesh of the XS.
In the event that the Mesh Active Antenna failed, students could use WiFi2
without any significant changes. In the event XS fails,
Wifi2 could be cabled to Wifi 1, allowing all students to access the
internet, unfiltered, uncached.
This approach has the advantage that if we don't know how many XO laptops
each antenna can handle, the WiFi 2 can certainly
handle 200 or more IP addresses. Kind of like an insurance policy to
ensure success.
Just a thought.
--- Tony
Tony Pearson
Senior Storage Consultant, IBM System Storage?
Telephone: +1 520-799-4309 | tie 321-4309 | Cell: +1 520 990-8669
email: tpearson at us.ibm.com | GSA: http://tucgsa.ibm.com/~tpearson
Blog: http://www.ibm.com/developerworks/blogs/page/InsideSystemStorage
AKA: 990tony Paravane, eightbar specialist
server-devel-request at lists.laptop.org
Sent by: server-devel-bounces at lists.laptop.org
02/08/2008 11:34 AM
Please respond to
server-devel at lists.laptop.org
To
server-devel at lists.laptop.org
cc
Subject
Server-devel Digest, Vol 10, Issue 14
Send Server-devel mailing list submissions to
server-devel at lists.laptop.org
To subscribe or unsubscribe via the World Wide Web, visit
http://lists.laptop.org/listinfo/server-devel
or, via email, send a message with subject or body 'help' to
server-devel-request at lists.laptop.org
You can reach the person managing the list at
server-devel-owner at lists.laptop.org
When replying, please edit your Subject line so it is more specific
than "Re: Contents of Server-devel digest..."
Today's Topics:
1. Re: Nepal project plan and requirements (Bryan Berry)
2. Re: Server-devel Digest, Vol 10, Issue 12 (Tony Pearson)
3. Re: ejabberd crash (Dafydd Harries)
4. Re: ejabberd crash (Robert McQueen)
5. Re: ejabberd crash ( J?r?me Sautret )
6. Re: ejabberd crash (Robert McQueen)
----------------------------------------------------------------------
Message: 1
Date: Fri, 08 Feb 2008 12:50:57 -0500
From: Bryan Berry <bryan.berry at gmail.com>
Subject: Re: [Server-devel] Nepal project plan and requirements
To: server-devel at lists.laptop.org
Cc: "Greg Smith \(gregmsmi\)" <gregmsmi at cisco.com>, sulochan
acharya
<sulochan at olenepal.org>
Message-ID: <1202493057.6442.43.camel at dell.linuxdev.us.dell.com>
Content-Type: text/plain
Greg and Wad,
> The network is the biggest headache right now. Are you planning to
> deploy in any schools larger than 150 students ?
We will have about 150-170 total laptops, which included students+teachers
> After we lock down SSO strategy, what do you want to focus on next? I
> > think we should switch from design to implementation in less than two
> > weeks. Hot topics we could try to flush out include:
> > - Moodle design and usage
> > - Backup and reliability, including storage and network aspects
The # 1 thing is backup and reliability
-
#2 Students need to be able to be able to "archive" their files to the
server
once they run out of local space, or at least able to browse the regular
backup of their files to the server. This could conceivably be done using
Moodle.
#3 Moodle
Localization -- We are taking care of this. Don't worry about it.
Sulochan, any other thoughts?
> Message: 2
> Date: Fri, 8 Feb 2008 01:02:17 -0500
> From: John Watlington <wad at laptop.org>
> Subject: Re: [Server-devel] Nepal project plan and requirements
> updated
> To: "Greg Smith (gregmsmi)" <gregmsmi at cisco.com>
> Cc: server-devel <server-devel at lists.laptop.org>
> Message-ID: <3D76415A-971C-46B6-86E4-406A217E4B3F at laptop.org>
> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed
>
>
> On Feb 7, 2008, at 5:04 PM, Greg Smith (gregmsmi) wrote:
>
> > Hi All,
> >
> > The Nepal project plan and school server specification Wiki pages have
> > been updated.
> > See:
> > http://wiki.laptop.org/go/Nepal#Planning
> >
> > We need to pick the XS server image we will deploy ASAP. Teacher
> > training starts April 1!
> >
> > Will there be another XS server image blessed by OLPC beyond
> > OLPC_XS_150.iso in the next few weeks?
> >
> Yes. We are hoping to get one including some critical new
> services (activation and update) out in the next week.
>
> > One open question is the status of single sign on.
> >
> > I see three options:
> > 1 - Don't use it in phase 1
> > 2 - Use one of the proposed work arounds. See Single Sign On
> > section at:
> > http://wiki.laptop.org/go/
> > Nepal:School_Server_Specification#Requirements
> > _Not_Specific_to_a_Core_Software_Module_
> > 3 - Use some other OLPC developed solution. (Bitfrost?)
> >
> > Let me know if option 3 has a chance to make a new build or if we
> > should
> > focus on the first two for now.
>
> Focus on one of the first two. I'm curious what you mean by
> "BitFrost", cause
> the BitFrost document didn't really extend past the laptop. The
> current ID scheme
> (described in earlier mail to Martin) IS the BitFrost ID scheme. A
> private/public
> key pair is used to identify the student. As this is kind of
> awkward for a username,
> two derivatives are in use: the laptop serial number and a hash of
> the public key.
> This is probably changing somewhat in the future.
>
> > Any other comments on the which of the listed requirements are
> > addressed
> > in which XS server build are appreciated.
> >
> > Also, clarifications, additions and suggestions are welcome. I will
> > try
> > to improve the formatting and add detail regularly.
> >
> > For the Nepal team,
> >
> > After we lock down SSO strategy, what do you want to focus on next? I
> > think we should switch from design to implementation in less than two
> > weeks. Hot topics we could try to flush out include:
> > - Moodle design and usage
> > - Backup and reliability, including storage and network aspects
>
> The network is the biggest headache right now. Are you planning to
> deploy in any schools larger than 150 students ?
>
> > - Caching and filtering
>
> Squid is working fine, although I'm not sure the logs get deleted as
> well as rotated... Update to the testing repository (edit /etc/
> yum.repos.olpc.d/testing
> and /etc/yum.repos.olpc.d/stable) --- make sure you get xs-config 2.5-8.
> Restart the httpd server, and now you should be able to start and
> stop web caching with
> /etc/sysconfig/olpc-scripts/TURN_SQUID_ON and TURN_SQUID_OFF.
> These persist across reboots.
>
> I'm working on getting Dansguardian packaged and configured.
>
> > - Localization
> > Something else?
> >
> > Pick an area and we'll try to suggest and help implement a design
> > ASAP.
> >
> > Thanks,
> >
> > Greg Smith
> >
> > _______________________________________________
> > Server-devel mailing list
> > Server-devel at lists.laptop.org
> > http://lists.laptop.org/listinfo/server-devel
>
>
>
> ------------------------------
>
> Message: 3
> Date: Fri, 8 Feb 2008 11:18:50 -0500
> From: John Watlington <wad at laptop.org>
> Subject: Re: [Server-devel] ejabberd crash
> To: "J?r?me Sautret" <jerome.sautret at process-one.net>
> Cc: olpc at collabora.co.uk, server-devel at lists.laptop.org, Robert
> McQueen <robert.mcqueen at collabora.co.uk>
> Message-ID: <FADC65D5-F6EA-44DE-BBFD-BFCF041754E5 at laptop.org>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
>
> We shouldn't have a ejabberd node named schoolserver1.
> It might have happened as part of a recent upgrade.
>
> I'll look into it.
> wad
>
> On Feb 8, 2008, at 11:01 AM, J?r?me Sautret wrote:
>
> > I can see that you have 2 nodes in your ejabberd cluster :
> > ejabberd at schoolserver1 and ejabberd at schoolserver (the former is
> > stopped). What is this ejabberd at schoolserver node, why it is stopped
> > and where it is supposed to run ?
>
>
>
> ------------------------------
>
> Message: 4
> Date: Fri, 8 Feb 2008 11:29:17 -0500
> From: John Watlington <wad at laptop.org>
> Subject: Re: [Server-devel] ejabberd crash
> To: "J?r?me Sautret" <jerome.sautret at process-one.net>
> Cc: olpc at collabora.co.uk, server-devel at lists.laptop.org, Robert
> McQueen <robert.mcqueen at collabora.co.uk>
> Message-ID: <90277CD9-371A-4BD8-8EFE-4FA89E231853 at laptop.org>
> Content-Type: text/plain; charset=ISO-8859-1; delsp=yes; format=flowed
>
>
> The init.d script grabs the node name from the hostname.
> My problem is that a server many have serveral names....
>
> schoolserver, school always resolves to the principal server in a
> school.
> schoolserver1, schoolserver2, etc... resolve to particular servers
> (any of which may become the principal server).
> I want the ejabberd node to be school.domainname, not
> schoolserver1.domainname.
>
> Is there going to be a problem if I just change /etc/init.d/ejabberd to
> hardwire a node name ?
>
> On Feb 8, 2008, at 11:01 AM, J?r?me Sautret wrote:
>
> > I can see that you have 2 nodes in your ejabberd cluster :
> > ejabberd at schoolserver1 and ejabberd at schoolserver (the former is
> > stopped). What is this ejabberd at schoolserver node, why it is stopped
> > and where it is supposed to run ?
>
>
>
> ------------------------------
>
> _______________________________________________
> Server-devel mailing list
> Server-devel at lists.laptop.org
> http://lists.laptop.org/listinfo/server-devel
>
>
> End of Server-devel Digest, Vol 10, Issue 13
> ********************************************
------------------------------
Message: 2
Date: Fri, 8 Feb 2008 11:33:40 -0700
From: Tony Pearson <tpearson at us.ibm.com>
Subject: Re: [Server-devel] Server-devel Digest, Vol 10, Issue 12
To: server-devel at lists.laptop.org
Message-ID:
<OFE928B81B.891581FE-ON872573E9.0065E57E-072573E9.0065FA87 at us.ibm.com>
Content-Type: text/plain; charset="utf-8"
Greg,
I am glad we are making progress. Here is what I posted yesterday to
summarize my side so far.
http://www.ibm.com/developerworks/blogs/page/InsideSystemStorage?entry=understanding_lamp_platform_for_web
Tony Pearson
Senior Storage Consultant, IBM System Storage?
Telephone: +1 520-799-4309 | tie 321-4309 | Cell: +1 520 990-8669
email: tpearson at us.ibm.com | GSA: http://tucgsa.ibm.com/~tpearson
Blog: http://www.ibm.com/developerworks/blogs/page/InsideSystemStorage
AKA: 990tony Paravane, eightbar specialist
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.laptop.org/pipermail/server-devel/attachments/20080208/ce8c5729/attachment-0001.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 166 bytes
Desc: not available
Url :
http://lists.laptop.org/pipermail/server-devel/attachments/20080208/ce8c5729/attachment-0002.gif
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/jpeg
Size: 3808 bytes
Desc: not available
Url :
http://lists.laptop.org/pipermail/server-devel/attachments/20080208/ce8c5729/attachment-0001.jpeg
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 166 bytes
Desc: not available
Url :
http://lists.laptop.org/pipermail/server-devel/attachments/20080208/ce8c5729/attachment-0003.gif
------------------------------
Message: 3
Date: Fri, 8 Feb 2008 08:40:25 +0000
From: Dafydd Harries <dafydd.harries at collabora.co.uk>
Subject: Re: [Server-devel] ejabberd crash
To: John Watlington <wad at laptop.org>
Cc: olpc at collabora.co.uk, server-devel at lists.laptop.org, Robert
McQueen <robert.mcqueen at collabora.co.uk>
Message-ID: <20080208084025.GY939 at localhost>
Content-Type: text/plain; charset=us-ascii
Ar 07/02/2008 am 21:26, ysgrifennodd John Watlington:
>
> ejabberd decided to eat its database again, on schoolserver.laptop.org
> (http://dev.laptop.org/ticket/6365)
>
> The error (from sasl.log) is:
> Error reading Mnesia database
>
> A number of laptops (25 or so) were registered with
> schoolserver.laptop.org today.
>
> Any recommendations on how to recover without nuking the
> database ? Any Mnesia experts out there ?
Ouch. Perhaps we should back up the ejabberd data directory so that
somebody
can debug it later?
> And a relevant question: what will telepathy do in this case ? Will
> it automatically
> re-register if a laptop's jabber account is deleted ?
No, but there is a patch that will do that:
http://dev.laptop.org/ticket/6295
--
Dafydd
------------------------------
Message: 4
Date: Fri, 08 Feb 2008 10:19:20 +0000
From: Robert McQueen <robert.mcqueen at collabora.co.uk>
Subject: Re: [Server-devel] ejabberd crash
To: John Watlington <wad at laptop.org>
Cc: olpc at collabora.co.uk, server-devel at lists.laptop.org, J?r? , me
Sautret <jerome.sautret at process-one.net>
Message-ID: <47AC2CA8.8030401 at collabora.co.uk>
Content-Type: text/plain; charset=UTF-8
Hi John,
John Watlington wrote:
> ejabberd decided to eat its database again, on schoolserver.laptop.org
> (http://dev.laptop.org/ticket/6365)
>
> The error (from sasl.log) is:
> Error reading Mnesia database
>
> A number of laptops (25 or so) were registered with
> schoolserver.laptop.org today.
>
> Any recommendations on how to recover without nuking the
> database ? Any Mnesia experts out there ?
The contract between OLPC and Process One is in place to look at our
ejabberd issues. I've CC'd J?r?me at P1 so he might consider having
somebody look at this issue for us if there's time.
J?r?me, would you or one of your staff like access to either of our
servers on jabber.laptop.org (where most development laptop images
connect to by default, so 100s of users), and schoolserver.laptop.org (a
test school server at the OLPC office) for debugging?
> And a relevant question: what will telepathy do in this case ? Will it
> automatically
> re-register if a laptop's jabber account is deleted ?
Currently no, as Daf says we have a patch for it in #6295. Might it be
worth raising the priority of this patch for Update 1? dwmw2 said there
was also a concern in Mongolia with laptops (particularly teachers)
moving between identically-named but distinct servers (which is pretty
broken, because it means we can never connect those servers together,
but that's an issue for later I guess...).
> wad
Regards,
Rob
------------------------------
Message: 5
Date: Fri, 8 Feb 2008 17:01:25 +0100
From: " J?r?me Sautret " <jerome.sautret at process-one.net>
Subject: Re: [Server-devel] ejabberd crash
To: "Robert McQueen" <robert.mcqueen at collabora.co.uk>
Cc: olpc at collabora.co.uk, server-devel at lists.laptop.org
Message-ID:
<8397da240802080801q3f53ffcfgd82e16c00db70796 at mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
I can see that you have 2 nodes in your ejabberd cluster :
ejabberd at schoolserver1 and ejabberd at schoolserver (the former is
stopped). What is this ejabberd at schoolserver node, why it is stopped
and where it is supposed to run ?
------------------------------
Message: 6
Date: Fri, 08 Feb 2008 16:03:31 +0000
From: Robert McQueen <robert.mcqueen at collabora.co.uk>
Subject: Re: [Server-devel] ejabberd crash
To: J?r?me Sautret <jerome.sautret at process-one.net>
Cc: olpc at collabora.co.uk, server-devel at lists.laptop.org
Message-ID: <47AC7D53.7090209 at collabora.co.uk>
Content-Type: text/plain; charset=ISO-8859-1
J?r?me Sautret wrote:
> I can see that you have 2 nodes in your ejabberd cluster :
> ejabberd at schoolserver1 and ejabberd at schoolserver (the former is
> stopped). What is this ejabberd at schoolserver node, why it is stopped
> and where it is supposed to run ?
I'm not sure we've ever tried to make a cluster, maybe wad has... Could
this result from changing the hostname of the machine?
Regards,
Rob
------------------------------
_______________________________________________
Server-devel mailing list
Server-devel at lists.laptop.org
http://lists.laptop.org/listinfo/server-devel
End of Server-devel Digest, Vol 10, Issue 14
********************************************
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.laptop.org/pipermail/server-devel/attachments/20080208/4713825c/attachment-0001.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 166 bytes
Desc: not available
Url : http://lists.laptop.org/pipermail/server-devel/attachments/20080208/4713825c/attachment-0002.gif
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/jpeg
Size: 3808 bytes
Desc: not available
Url : http://lists.laptop.org/pipermail/server-devel/attachments/20080208/4713825c/attachment-0001.jpeg
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 166 bytes
Desc: not available
Url : http://lists.laptop.org/pipermail/server-devel/attachments/20080208/4713825c/attachment-0003.gif
More information about the Server-devel
mailing list