[Planetlab-devel] Inconsistencies in records

[Larry Peterson]

It's entirely possible things aren't in sync. It's a long story,
recorded in the devel mailing list archive, but you've started
using sfi at a transition point in the underlying PLC (from 4.2
to 4.3).

Larry

On Thu, Jun 18, 2009 at 5:30 PM, Jeannie
Albrecht<jeannie at cs.williams.edu> wrote:
> As an aside, the fact that Danny can't authenticate with his private
> key is a little strange.  I have verified that he is using the correct
> keys (public and private), and he is able to use these keys to login
> to PlanetLab nodes.  However sfi keeps reporting a key mismatch error.
>  Is it possible that the database that we connect to via sfi is not
> entirely in sync with the information we see when we log in to PLC
> through the website?  That would explain some of our current issues...
>
> Thanks,
> Jeannie
>
> On Thu, Jun 18, 2009 at 4:58 PM, Danny Y. Huang<yh1 at williams.edu> wrote:
>> Hi,
>>
>> Having been poking around the sfi.py utility recently, I am getting
>> inconsistent records from the PLC. Could someone please explain what I have
>> done wrong?
>>
>> First, I have observed inconsistencies between SFI and the Planet-Lab.org
>> web interface. My advisor, a PI, added a new user (David, for the sake of
>> anonymity) to Williams in Planet-Lab.org yesterday. But when I do "./sfi.py
>> list plc.williams" today, the user is not listed. And when I do "./sfi.py
>> show plc.williams.david", sfi returns no record. Furthermore, we have a
>> slice called "gush", but it, too, fails to show up when I do "./sfi.py list
>> plc.williams". The Gush slice has been in existence for a long time, but why
>> doesn't it appear in the result?
>>
>> Then, I have also noticed inconsistences within the SFI. For the purpose of
>> testing the resolve() method, I added the following method to sfi.py:
>>
>>> def moo(opts, args):
>>>    global registry
>>>    user_cred = get_user_cred()
>>>    results = registry.resolve(user_cred, args[0])
>>>    print results
>>
>> In order for sfi.py to accept this new command, I added moo() to the list of
>> accepted commands in create_cmd_parser(). When I run "./sfi.py moo
>> plc.williams", both "plc.williams.gush" and "plc.williams.david" show up in
>> the result. Why are they here in the result of moo() yet not in "./sfi.py
>> list plc.williams"? In addition, my moo() method yields inconsistent
>> results. Although both David and Gush appear in "./sfi.py moo plc.williams",
>> their records are blank when I do "./sfi.py moo plc.williams.gush" and
>> "./sfi.py moo plc.williams.david" respectively.
>>
>> Have I understood something incorrectly? Could someone please point out my
>> mistakes? Thank you!
>>
>> Best,
>> Danny Y. Huang
>> http://sysnet1.cs.williams.edu
>>
>> PS: I'm testing with my advisor's private key because I cannot get my RSA
>> key pairs to work. She's PI. This is my configSfi.sh file:
>>
>>> export SFI_AUTH=plc.williams
>>> export SFI_USER=$SFI_AUTH.jeannie
>>> export SFI_REGISTRY=http://128.112.139.90:12345/
>>> export SFI_SM=http://www.planet-lab.org:12347/
>>
>>
>>
>> _______________________________________________
>> Devel mailing list
>> Devel at lists.planet-lab.org
>> https://lists.planet-lab.org/mailman/listinfo/devel
>>
>>
>
> _______________________________________________
> Devel mailing list
> Devel at lists.planet-lab.org
> https://lists.planet-lab.org/mailman/listinfo/devel
>