Changes between Version 4 and Version 5 of GENIRacksHome/ExogeniRacks/AggregateSurvey


Ignore:
Timestamp:
04/09/12 14:47:50 (12 years ago)
Author:
lnevers@bbn.com
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • GENIRacksHome/ExogeniRacks/AggregateSurvey

    v4 v5  
    753753}}}
    754754
     755=== GPO Dev Team Analysis ===
     756
     757The GPO Software development team analyzed the results from this test run and came up with the following list of Known issues with ExoGENI's AM API implementation.
     758
     759Priorities (from Trac): blocker, critical, major, minor, trivial [[BR]]
     760Status: where things stand - open, closed, etc. [[BR]]
     761Categories: AM API, RSpec, Acceptance Test, Stitching, Question [[BR]]
     762Issues are ordered by priority, and by relative importance within priority.
     763
     764==== Critical Issues ====
     765|| ID || Priority || Status || Category || Description ||
     766|| 1 || critical ||  || AM API || !ListResources at ExoSM returns multiple advertisement RSpecs. This is neither valid XML nor valid RSpec. ||
     767|| 2 || critical ||  || RSpec || ExoGENI advertises RSpec GENI v3 support in !GetVersion, but the RSpecs (ad & manifest) are ProtoGENI v2. Make the RSpecs use the GENI v3 schema. (As a fallback, change the !GetVersion return).||
     768|| 3 || critical ||  || RSpec || missing schemaLocation declarations for all used schemas ||
     769|| 4|| critical || || RSpec || All Ad & manifest RSpecs must pass rspeclint, but currently do not ||
     770
     771==== Major Issues ====
     772|| ID || Priority || Status || Category || Description ||
     773|| 5 || major  ||  || AM API || !GetVersion: The top-level (parallel to code and value) is missing a `geni_api` entry.  Without this entry AM API v1 clients can't discover why they can't talk to this aggregate. ||
     774|| 6 || major ||  || AM API || !SliverStatus: Return at least a `geni_resources` entry that is empty, if not including all contained slivers. ||
     775|| 7 || major ||  || Acceptance Test || There is no 'extensions' field listed in `geni_request_rspec_versions` and `geni_ad_rspec_version` in the return from !GetVersion. Add it, including all used extensions. ||
     776|| 8 || major ||  || Acceptance Test || !ListResources does not handle the `geni_compressed` flag properly. The return value is always compressed, even if `geni_compressed` is set to False.||
     777
     778==== Minor Issues ====
     779|| ID || Priority || Status || Category || Description ||
     780|| 9 || minor || || RSpec || Does ExoGENI honor the sliver_type or ignore it? It seems to give a VM for a `sliver_type` of "raw_pc". ||
     781|| 10 || minor ||  || AM API || !CreateSliver results in a !NullPointerException if a link references a non-existent node. {{{<link client_id="center"> <interface_ref client_id="geni1:0"/> <interface_ref client_id="geni2:0"/> </link>}}} when node {{{geni1}}} is not in the request. ||
     782|| 11 || minor || || Stitching || Test stitching schema validity by using Omni's libstitch. Multiple undocumented conventions about IDs may need to be followed. ||
     783|| 12 || minor || || RSpec || You list 1 node with a hardware_type 'vm-cloud'. What about at least listing multiple sliver_types for the different EC2 style VM types, with the appropriate type_slots value for each sliver type? ||
     784|| 13 || minor || || RSpec || You list 1 node for each of 2 or 3 different hardware types, all for the same site. Is information duplicated? Could these reasonably be combined in any way? ||
     785|| 14 || minor || || RSpec || The dataplane IP address is missing from the manifest RSpec ||
     786|| 15 || minor ||  || RSpec || Add more info to the <interface> tag in manifest: PG includes the <ip> sub-element with IP, netmask, type ||
     787|| 16 || minor || || RSpec || Change structure of component_manager_id. AM URNs should be of the form <prefix>+name+authority+am. So use prefix+<SM site name+domain name>+authority+am ||
     788|| 17 || minor || || RSpec || Ad component_manager_id values are unlike those in sample request RSpec. Shouldn't this break something? Make these match. ||
     789|| 18 || minor || || RSpec || client_id values are changed in the manifest from the request (add a UUID), making it hard for experimenters/tools to reconcile. Can these be kept the same? ||
     790|| 19 || minor || || RSpec || Add more information to the RSpec if possible: `<component_manager>` sub-element on links, indicating the 2 AMs that control the endpoints of this link. ||
     791|| 20 || minor || || RSpec || Add more information to the RSpec if possible: `<property>` sub-element on links specifying dest_id and source_id ||
     792|| 21 || minor || || RSpec || Add more info / check form of interface `component_id`. Compare to PG form: prefix+site+interface+node/switch name:port #:name of remote, e.g. emulab.net+interface+procurve=pgeni-wash:10:ion ||
     793|| 22 || minor || || RSpec || Simplify sliver ID. How about prefix+<site>+sliver+AM name + slice name+client_id? ||
     794|| 23 || minor || || RSpec || Node component_id is the site name, making all reservations part of this same node. Maybe change exclusive to False? ||
     795|| 24 || minor ||  || Question || Are the assigned `client_id` values from the RSpec available within the running Node/vm? ||
     796|| 25 || minor ||  || Stitching RSpec || Stitching extension says scheduled=true. Probably should be false. ||
     797|| 26 || minor ||  || Stitching RSpec || stitching extension has same component_id for both port and link. Seems wrong. Maybe OK? PG has the link ID being the ID of the interface of the corresponding node. So this part of the Orca ad seems like a mistake.||
     798|| 27 || minor ||  || RSpec || The starlight Ad RSpec lists hardware_type name=orca-transit-net twice. Not a real problem, just not what he intended probably. ||
     799
     800
     801== AM API additional Tests ==
     802
     803Two tests were run for individual AM API Compliance test based on GPO Dev request, their results are captured in this page.
     804
    755805=== test_ListResources_geni_compressed only ===
    756806