[Server-devel] XSCE Sprint

[David Farning]

Thanks for this thread....

Creating a place to share and record this knowledge is as important as the code!

On Thu, Jul 11, 2013 at 2:49 AM, Tony Anderson <tony at olenepal.org> wrote:
> Hi,
>
> Thanks again. The laptops are good to go for this school year (12.1.0). I
> will be able to work the collection stick problem when I return to the
> schools (probably in December).
> I'll double check the flash time to check for variability between units.
>
> At these schools all the laptops are XO-1 or XO1.5. However, I think a
> Nandblast facility working across all the models would be very useful for
> the start-of-year update.
>
> Yours,
>
> Tony
>
>
> On 07/11/2013 09:32 AM, James Cameron wrote:
>>
>> On Wed, Jul 10, 2013 at 11:07:04AM +0200, Tony Anderson wrote:
>>>
>>> Hi,
>>>
>>> Thanks again for this!
>>>
>>> What I gather is that we should use Nandblast from an XO for
>>> reflash. For a time it was not supported for XO-1.5, but my current
>>> understanding is that it supported for all versions of XO.
>>
>> Up to 13.1.0 it is supported on XO-1, XO-1.5 and XO-1.75.
>>
>> It is fast for XO-1, and quite slow for XO-1.5 and XO-1.75.  The
>> breakeven point, where I would switch from USB drive to NANDblaster,
>> is about five XO-1, and about twenty for XO-1.5 and XO-1.75.
>>
>> In 13.2.0 it is broken, and I am working on that in ticket #12726,
>> hoping to get fixes in before release.  Fixes are available for XO-1
>> (Q2F19) and XO-1.5 (Q3C16).  XO-1.75 is still a problem.
>>
>> It was not intended to be supported on XO-4 with the new 802.11n
>> wireless card, but so far it looks possible.
>>
>>> In Lesotho, the flash was taking 15min from boot to reboot for
>>> registration. These laptops (XO-1) date from the first G1G1 and so
>>> there is no telling about endurance.
>>
>> That time of 15 minutes is far too long, and should be investigated
>> ... if the internal storage is three times slower than when the
>> laptops were produced, you will have performance problems.  I have
>> some old XO-1 units here that have been used by children, and they are
>> not showing that symptom.
>>
>>> Naturally, reload of the Journal occurs via the file system after
>>> the flash. Sadly, this is not a current issue because none of the
>>> deployments actually use the Journal (e.g. in Lesotho the laptops
>>> are shared among several students).
>>>
>>> The laptops (XO-1.5s) at Saint Jacobs were sponsored by a group in
>>> Stuttgart and are not part of the Rwanda purchase. In any case, I
>>> believe the information needed for the collection stick is available
>>> (serial number and uuid).
>>
>> If it is already available, get it to me.
>>
>>> Yours,
>>>
>>> Tony
>>>
>>>
>>> On 07/10/2013 10:55 AM, James Cameron wrote:
>>>>
>>>> On Wed, Jul 10, 2013 at 10:11:14AM +0200, Tony Anderson wrote:
>>>>>
>>>>> The 'locked' XO problem derives from XOs distributed in the minimum
>>>>> 100 XO purchase - many of these are locked. Also, in Rwanda the
>>>>> policy is to keep the laptops locked even though they have
>>>>> indefinite leases.
>>>>
>>>> Rwanda probably has a deployment key and should be able to sign builds
>>>> with it.  My guess is that the laptops would also have the deployment
>>>> keys injected already.  You will need to work with the people who have
>>>> the keys.
>>>>
>>>>> My current plans are to visit these schools in December and so I may
>>>>> be able to get them unlocked then.
>>>>
>>>> http://wiki.laptop.org/go/Collection_stick is perfect for this,
>>>> provided the deployment did not remove the OLPC keys.
>>>>
>>>> (I don't have records of what deployments have done).
>>>>
>>>>> What I would really love is a 'Nandblast' capability in the firmware
>>>>> that gets it's image from the schoolserver. That probably would work
>>>>> (how does an XO know an image is coming over wifi from an XO or a
>>>>> school server?).
>>>>
>>>> There is no implementation of NANDblaster for the school server,
>>>> because it requires special support in the wireless device.  A typical
>>>> access point will not work.  It requires an XO as the sender.
>>>>
>>>> (NANDblaster is implemented in the firmware, not the operating system.
>>>> An alternate design could be engineered, but that doesn't seem likely
>>>> to be attempted.)
>>>>
>>>>> The normal flash problem is that several XOs need to be reflashed at
>>>>> one time, so the usb key approach is time-consuming. My experience
>>>>> is that a reflash from usb key takes 15min. Naturally, one key to
>>>>> this process is the ability to reload the backed up (hopefully)
>>>>> Journal.
>>>>
>>>> Reloading the backed up journal is costing you 10min?  Reflashing an
>>>> XO-1 from USB drive 13.2.0-12 costs only 5min.  If the reflash is
>>>> taking much longer than that, there may be an endurance or end of life
>>>> problem with the internal storage of the XO.
>>>>
>



-- 
David Farning
Activity Central: http://www.activitycentral.com